Merge branch 'upstream-gallium-0.1' into nouveau-gallium-0.1

Conflicts:

	configs/linux-dri
	src/mesa/pipe/Makefile

410 files changed, 5058 insertions, 2803 deletions

diff --git a/configs/beos b/configs/beos
index f07973d0c78..2b74af739d0 100644
--- a/configs/beos
+++ b/configs/beos
@@ -86,7 +86,7 @@ else
 endif
 
 # Directories
-SRC_DIRS = mesa glu glut/beos
+SRC_DIRS = gallium mesa glu glut/beos
 GLU_DIRS = sgi
 DRIVER_DIRS = beos
 PROGRAM_DIRS = beos samples redbook demos tests
diff --git a/configs/darwin b/configs/darwin
index 7826ecc605e..bba78026962 100644
--- a/configs/darwin
+++ b/configs/darwin
@@ -25,5 +25,5 @@ GLW_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXt $(TOP)/lib/GL.dylib
 APP_LIB_DEPS = -L$(TOP)/lib -l$(GLUT_LIB) -l$(GLU_LIB) -l$(GL_LIB) -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm
 
 # omit glw lib for now:
-SRC_DIRS = mesa glu glut/glx
+SRC_DIRS = gallium mesa glu glut/glx
 
diff --git a/configs/darwin-x86ppc b/configs/darwin-x86ppc
index 13172327a76..ebeb25051f1 100644
--- a/configs/darwin-x86ppc
+++ b/configs/darwin-x86ppc
@@ -29,5 +29,5 @@ GLW_LIB_DEPS = -L/usr/X11R6/lib -lX11 -lXt $(TOP)/lib/GL.dylib
 APP_LIB_DEPS = -L$(TOP)/lib -l$(GLUT_LIB) -l$(GLU_LIB) -l$(GL_LIB) -L/usr/X11R6/lib -lX11 -lXmu -lXt -lXi -lm
 
 # omit glw lib for now:
-SRC_DIRS = mesa glu glut/glx
+SRC_DIRS = gallium mesa glu glut/glx
 
diff --git a/configs/default b/configs/default
index 166205a1d31..25a87e66a1b 100644
--- a/configs/default
+++ b/configs/default
@@ -60,7 +60,7 @@ GLW_SOURCES = GLwDrawA.c
 
 # Directories to build
 LIB_DIR = lib
-SRC_DIRS = mesa glu glut/glx glw
+SRC_DIRS = gallium mesa glu glut/glx glw
 GLU_DIRS = sgi
 DRIVER_DIRS = x11 osmesa
 # Which subdirs under $(TOP)/progs/ to enter:
diff --git a/configs/freebsd-dri b/configs/freebsd-dri
index 402883d1de0..67d253b8695 100644
--- a/configs/freebsd-dri
+++ b/configs/freebsd-dri
@@ -36,7 +36,7 @@ GLW_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -L/usr/X11R6/lib -lGL -lXt -lX11
 
 
 # Directories
-SRC_DIRS = glx/x11 mesa glu glut/glx glw
+SRC_DIRS = glx/x11 gallium mesa glu glut/glx glw
 DRIVER_DIRS = dri
 PROGRAM_DIRS = 
 WINDOW_SYSTEM=dri
diff --git a/configs/linux-cell b/configs/linux-cell
index 3d874491e41..09f62fc1ff0 100644
--- a/configs/linux-cell
+++ b/configs/linux-cell
@@ -21,7 +21,7 @@ CFLAGS = $(OPT_FLAGS) -Wall -Winline -fPIC -m32 -mabi=altivec -maltivec -I. -I$(
 CXXFLAGS = $(CFLAGS)
 
 # Omitting glw here:
-SRC_DIRS = mesa glu glut/glx
+SRC_DIRS = gallium mesa glu glut/glx
 
 
 MKDEP_OPTIONS = -fdepend -Y
@@ -36,7 +36,9 @@ GL_LIB_DEPS = $(EXTRA_LIB_PATH) -lX11 -lXext -lm -lpthread \
 
 SPU_CC = spu-gcc
 
-SPU_CFLAGS = $(OPT_FLAGS) -W -Wall -Winline -Wmissing-prototypes -Wno-main -I. -I $(SDK)/spu/include -include spu_intrinsics.h -I $(TOP)/src/mesa/
+SPU_CFLAGS = $(OPT_FLAGS) -W -Wall -Winline -Wmissing-prototypes -Wno-main \
+	-I. -I$(SDK)/spu/include -I$(TOP)/src/mesa/ $(INCLUDE_DIRS) \
+	-include spu_intrinsics.h
 
 SPU_LFLAGS = -L$(SDK)/spu/lib -Wl,-N -lmisc
 
diff --git a/configs/linux-directfb b/configs/linux-directfb
index 09332f48081..dff27f78503 100644
--- a/configs/linux-directfb
+++ b/configs/linux-directfb
@@ -22,7 +22,7 @@ ifeq ($(HAVE_X86), yes)
 endif
 
 # Directories
-SRC_DIRS     = mesa glu glut/directfb
+SRC_DIRS     = gallium mesa glu glut/directfb
 GLU_DIRS     = sgi
 DRIVER_DIRS  = directfb
 PROGRAM_DIRS = demos directfb
diff --git a/configs/linux-dri b/configs/linux-dri
index c7eb7112c0e..e6135856fc1 100644
--- a/configs/linux-dri
+++ b/configs/linux-dri
@@ -54,10 +54,10 @@ USING_EGL=0
 
 # Directories
 ifeq ($(USING_EGL), 1)
-SRC_DIRS = egl glx/x11 mesa glu glut/glx glw
+SRC_DIRS = egl glx/x11 gallium mesa glu glut/glx glw
 PROGRAM_DIRS = egl
 else
-SRC_DIRS = glx/x11 mesa glu glut/glx glw
+SRC_DIRS = glx/x11 gallium mesa glu glut/glx glw
 PROGRAM_DIRS =
 endif
 
@@ -66,4 +66,4 @@ WINDOW_SYSTEM=dri
 
 # gamma are missing because they have not been converted to use the new
 # interface.
-DRI_DIRS = nouveau_winsys
+DRI_DIRS = intel 
diff --git a/configs/linux-dri-xcb b/configs/linux-dri-xcb
index aa292a13ec1..ea4bdf1864c 100644
--- a/configs/linux-dri-xcb
+++ b/configs/linux-dri-xcb
@@ -53,10 +53,10 @@ USING_EGL=0
 
 # Directories
 ifeq ($(USING_EGL), 1)
-SRC_DIRS = egl glx/x11 mesa glu glut/glx glw
+SRC_DIRS = egl glx/x11 gallium mesa glu glut/glx glw
 PROGRAM_DIRS = egl
 else
-SRC_DIRS = glx/x11 mesa glu glut/glx glw
+SRC_DIRS = glx/x11 gallium mesa glu glut/glx glw
 PROGRAM_DIRS =
 endif
 
diff --git a/configs/linux-fbdev b/configs/linux-fbdev
index e36d20a702c..1ddccb3f52b 100644
--- a/configs/linux-fbdev
+++ b/configs/linux-fbdev
@@ -6,7 +6,7 @@ CONFIG_NAME = linux-fbdev
 
 CFLAGS = -O3 -ffast-math -ansi -pedantic -fPIC -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE -D_BSD_SOURCE -DPTHREADS -DUSE_GLFBDEV_DRIVER
 
-SRC_DIRS = mesa glu glut/fbdev
+SRC_DIRS = gallium mesa glu glut/fbdev
 DRIVER_DIRS = fbdev osmesa
 PROGRAM_DIRS = fbdev demos redbook samples
 
diff --git a/configs/linux-osmesa b/configs/linux-osmesa
index cc1fbbd109a..0382a19553a 100644
--- a/configs/linux-osmesa
+++ b/configs/linux-osmesa
@@ -14,7 +14,7 @@ CXXFLAGS = -O3 -ansi -pedantic -fPIC -ffast-math -D_POSIX_SOURCE -D_POSIX_C_SOUR
 
 
 # Directories
-SRC_DIRS = mesa glu
+SRC_DIRS = gallium mesa glu
 DRIVER_DIRS = osmesa
 PROGRAM_DIRS = osdemos
 
diff --git a/configs/linux-osmesa16 b/configs/linux-osmesa16
index 1fb0186d315..9a527592f1d 100644
--- a/configs/linux-osmesa16
+++ b/configs/linux-osmesa16
@@ -17,7 +17,7 @@ OSMESA_LIB_NAME = libOSMesa16.so
 
 
 # Directories
-SRC_DIRS = mesa glu
+SRC_DIRS = gallium mesa glu
 DRIVER_DIRS = osmesa
 PROGRAM_DIRS = 
 
diff --git a/configs/linux-osmesa16-static b/configs/linux-osmesa16-static
index 6645504478e..1e6380b02e0 100644
--- a/configs/linux-osmesa16-static
+++ b/configs/linux-osmesa16-static
@@ -18,7 +18,7 @@ OSMESA_LIB_NAME = libOSMesa16.a
 
 
 # Directories
-SRC_DIRS = mesa glu
+SRC_DIRS = gallium mesa glu
 DRIVER_DIRS = osmesa
 PROGRAM_DIRS = 
 
diff --git a/configs/linux-osmesa32 b/configs/linux-osmesa32
index a1e5a358d60..f0ef1831b09 100644
--- a/configs/linux-osmesa32
+++ b/configs/linux-osmesa32
@@ -17,7 +17,7 @@ OSMESA_LIB_NAME = libOSMesa32.so
 
 
 # Directories
-SRC_DIRS = mesa glu
+SRC_DIRS = gallium mesa glu
 DRIVER_DIRS = osmesa
 PROGRAM_DIRS = 
 
diff --git a/configs/linux-solo b/configs/linux-solo
index 220fe58b9a4..d49b9722282 100644
--- a/configs/linux-solo
+++ b/configs/linux-solo
@@ -43,7 +43,7 @@ GLUT_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GLU_LIB) -l$(GL_LIB) -lm
 APP_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLU_LIB) -l$(GL_LIB) -lm -lpthread
 
 # Directories
-SRC_DIRS = glx/mini mesa glu glut/mini
+SRC_DIRS = glx/mini gallium mesa glu glut/mini
 DRIVER_DIRS = dri
 PROGRAM_DIRS = miniglx
 
diff --git a/progs/trivial/Makefile b/progs/trivial/Makefile
index 3a1caf15e71..d19abe7e865 100644
--- a/progs/trivial/Makefile
+++ b/progs/trivial/Makefile
@@ -39,6 +39,7 @@ SOURCES = \
 	linestrip-stipple-wide.c \
 	linestrip-stipple.c \
 	linestrip.c \
+	pgon-mode.c \
 	point-clip.c \
 	point-param.c \
 	point-sprite.c \
diff --git a/progs/trivial/pgon-mode.c b/progs/trivial/pgon-mode.c
new file mode 100644
index 00000000000..db586386196
--- /dev/null
+++ b/progs/trivial/pgon-mode.c
@@ -0,0 +1,136 @@
+/**
+ * Test glPolygonMode.
+ * A tri-strip w/ two tris is drawn so that the first tri is front-facing
+ * but the second tri is back-facing.
+ * Set glPolygonMode differently for the front/back faces
+ *
+ */
+
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <GL/glut.h>
+
+static int Win;
+static GLfloat Zrot = 0;
+static GLboolean FrontFillBackUnfilled = GL_TRUE;
+static GLboolean Lines = GL_TRUE;
+
+
+static void
+Draw(void)
+{
+   glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+   if (FrontFillBackUnfilled) {
+      if (Lines) {
+         printf("FrontMode = FILL, BackMode = LINE\n");
+         glPolygonMode(GL_BACK, GL_LINE);
+      }
+      else {
+         printf("FrontMode = FILL, BackMode = POINT\n");
+         glPolygonMode(GL_BACK, GL_POINT);
+      }
+      glPolygonMode(GL_FRONT, GL_FILL);
+   }
+   else {
+      if (Lines) {
+         printf("FrontMode = LINE, BackMode = FILL\n");
+         glPolygonMode(GL_FRONT, GL_LINE);
+      }
+      else {
+         printf("FrontMode = POINT, BackMode = FILL\n");
+         glPolygonMode(GL_FRONT, GL_POINT);
+      }
+      glPolygonMode(GL_BACK, GL_FILL);
+   }
+
+   glPushMatrix();
+   glRotatef(Zrot, 0, 0, 1);
+
+   glBegin(GL_TRIANGLE_STRIP);
+   glVertex2f(-1, 0);
+   glVertex2f( 1, 0);
+   glVertex2f(0,  1);
+   glVertex2f(0, -1);
+   glEnd();
+
+   glPopMatrix();
+
+   glutSwapBuffers();
+}
+
+
+static void
+Reshape(int width, int height)
+{
+   glViewport(0, 0, width, height);
+   glMatrixMode(GL_PROJECTION);
+   glLoadIdentity();
+   glFrustum(-1.0, 1.0, -1.0, 1.0, 5.0, 25.0);
+   glMatrixMode(GL_MODELVIEW);
+   glLoadIdentity();
+   glTranslatef(0.0, 0.0, -15.0);
+}
+
+
+static void
+Key(unsigned char key, int x, int y)
+{
+   const GLfloat step = 3.0;
+   (void) x;
+   (void) y;
+   switch (key) {
+   case 'p':
+      FrontFillBackUnfilled = !FrontFillBackUnfilled;
+      break;
+   case 'l':
+      Lines = !Lines;
+      break;
+   case 'z':
+      Zrot -= step;
+      break;
+   case 'Z':
+      Zrot += step;
+      break;
+   case 27:
+      glutDestroyWindow(Win);
+      exit(0);
+      break;
+   }
+   glutPostRedisplay();
+}
+
+
+static void
+Init(void)
+{
+   printf("GL_RENDERER = %s\n", (char*) glGetString(GL_RENDERER));
+
+   glLineWidth(3.0);
+   glPointSize(3.0);
+
+   glColor4f(1, 1, 1, 0.8);
+   glEnable(GL_BLEND);
+   glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
+
+   printf("Press 'p' to toggle polygon mode\n");
+}
+
+
+int
+main(int argc, char *argv[])
+{
+   glutInit(&argc, argv);
+   glutInitWindowPosition(0, 0);
+   glutInitWindowSize(400, 400);
+   glutInitDisplayMode(GLUT_RGB | GLUT_DOUBLE | GLUT_DEPTH);
+   Win = glutCreateWindow(argv[0]);
+   glutReshapeFunc(Reshape);
+   glutKeyboardFunc(Key);
+   glutDisplayFunc(Draw);
+   Init();
+   glutMainLoop();
+   return 0;
+}
diff --git a/src/gallium/Makefile b/src/gallium/Makefile
new file mode 100644
index 00000000000..89e068a4492
--- /dev/null
+++ b/src/gallium/Makefile
@@ -0,0 +1,20 @@
+TOP = ../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = auxiliary drivers
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
diff --git a/src/mesa/pipe/Makefile.template b/src/gallium/Makefile.template
index 8e84f8eb2d7..83b25f9b47c 100644
--- a/src/mesa/pipe/Makefile.template
+++ b/src/gallium/Makefile.template
@@ -15,7 +15,10 @@ OBJECTS = $(C_SOURCES:.c=.o) \
 ### Include directories
 INCLUDES = \
 	-I. \
-	-I$(TOP)/src/mesa/pipe \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/include/pipe \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
 	-I$(TOP)/src/mesa \
 	-I$(TOP)/include \
         $(DRIVER_INCLUDES)
@@ -38,7 +41,7 @@ INCLUDES = \
 default: depend symlinks $(LIBNAME)
 
 
-$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template
+$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template
 	$(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS)
 
 
diff --git a/src/mesa/pipe/README.portability b/src/gallium/README.portability
index c70ca774da0..c70ca774da0 100644
--- a/src/mesa/pipe/README.portability
+++ b/src/gallium/README.portability
diff --git a/src/mesa/pipe/SConscript b/src/gallium/SConscript
index d9c20e01007..d9c20e01007 100644
--- a/src/mesa/pipe/SConscript
+++ b/src/gallium/SConscript
diff --git a/src/mesa/pipe/Makefile b/src/gallium/auxiliary/Makefile
index 2b03b5c2446..da68498aa1f 100644
--- a/src/mesa/pipe/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -2,16 +2,11 @@ TOP = ../../..
 include $(TOP)/configs/current
 
 
-ifeq ($(CONFIG_NAME), linux-cell)
-CELL_DIR = cell
-endif
-
 ifeq ($(CONFIG_NAME), linux-llvm)
 LLVM_DIR = llvm
 endif
 
-SUBDIRS = softpipe i915simple i965simple nv30 nv40 nv50 failover pipebuffer \
-	  $(CELL_DIR) $(LLVM_DIR)
+SUBDIRS = pipebuffer $(LLVM_DIR)
 
 
 default: subdirs
diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c
index 9e77e0774dd..9e77e0774dd 100644
--- a/src/mesa/pipe/cso_cache/cso_cache.c
+++ b/src/gallium/auxiliary/cso_cache/cso_cache.c
diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h
index 116e2eaa2c8..116e2eaa2c8 100644
--- a/src/mesa/pipe/cso_cache/cso_cache.h
+++ b/src/gallium/auxiliary/cso_cache/cso_cache.h
diff --git a/src/mesa/pipe/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c
index 0338cb3b474..0338cb3b474 100644
--- a/src/mesa/pipe/cso_cache/cso_hash.c
+++ b/src/gallium/auxiliary/cso_cache/cso_hash.c
diff --git a/src/mesa/pipe/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h
index b4aa111860f..b4aa111860f 100644
--- a/src/mesa/pipe/cso_cache/cso_hash.h
+++ b/src/gallium/auxiliary/cso_cache/cso_hash.h
diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile
new file mode 100644
index 00000000000..fe9b150f304
--- /dev/null
+++ b/src/gallium/auxiliary/draw/Makefile
@@ -0,0 +1,2 @@
+default:
+	cd ../../../mesa ; make
diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c
index 61130c5600a..e3051507eaf 100644
--- a/src/mesa/pipe/draw/draw_clip.c
+++ b/src/gallium/auxiliary/draw/draw_clip.c
@@ -459,7 +459,7 @@ static void clip_reset_stipple_counter( struct draw_stage *stage )
 
 static void clip_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -472,7 +472,7 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw )
 {
    struct clipper *clipper = CALLOC_STRUCT(clipper);
 
-   draw_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES+1 );
+   draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 );
 
    clipper->stage.draw = draw;
    clipper->stage.point = clip_point;
diff --git a/src/mesa/pipe/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index b15f57c8248..4be38303169 100644
--- a/src/mesa/pipe/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -242,7 +242,7 @@ draw_convert_wide_lines(struct draw_context *draw, boolean enable)
 /**
  * Allocate space for temporary post-transform vertices, such as for clipping.
  */
-void draw_alloc_tmps( struct draw_stage *stage, unsigned nr )
+void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr )
 {
    assert(!stage->tmp);
 
@@ -260,7 +260,7 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr )
 }
 
 
-void draw_free_tmps( struct draw_stage *stage )
+void draw_free_temp_verts( struct draw_stage *stage )
 {
    if (stage->tmp) {
       FREE( stage->tmp[0] );
diff --git a/src/mesa/pipe/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index ddeb184497a..ddeb184497a 100644
--- a/src/mesa/pipe/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/gallium/auxiliary/draw/draw_cull.c
index 05c274e4dc2..8177b0ac86e 100644
--- a/src/mesa/pipe/draw/draw_cull.c
+++ b/src/gallium/auxiliary/draw/draw_cull.c
@@ -123,7 +123,7 @@ static void cull_reset_stipple_counter( struct draw_stage *stage )
 
 static void cull_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -135,7 +135,7 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw )
 {
    struct cull_stage *cull = CALLOC_STRUCT(cull_stage);
 
-   draw_alloc_tmps( &cull->stage, 0 );
+   draw_alloc_temp_verts( &cull->stage, 0 );
 
    cull->stage.draw = draw;
    cull->stage.next = NULL;
diff --git a/src/mesa/pipe/draw/draw_debug.c b/src/gallium/auxiliary/draw/draw_debug.c
index d6220b5f62c..d6220b5f62c 100644
--- a/src/mesa/pipe/draw/draw_debug.c
+++ b/src/gallium/auxiliary/draw/draw_debug.c
diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/gallium/auxiliary/draw/draw_flatshade.c
index 1419f287d2c..4398abbc60c 100644
--- a/src/mesa/pipe/draw/draw_flatshade.c
+++ b/src/gallium/auxiliary/draw/draw_flatshade.c
@@ -176,7 +176,7 @@ static void flatshade_reset_stipple_counter( struct draw_stage *stage )
 
 static void flatshade_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -188,7 +188,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw )
 {
    struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage);
 
-   draw_alloc_tmps( &flatshade->stage, 2 );
+   draw_alloc_temp_verts( &flatshade->stage, 2 );
 
    flatshade->stage.draw = draw;
    flatshade->stage.next = NULL;
diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/gallium/auxiliary/draw/draw_offset.c
index a2990ee8a88..dbc676deae4 100644
--- a/src/mesa/pipe/draw/draw_offset.c
+++ b/src/gallium/auxiliary/draw/draw_offset.c
@@ -159,7 +159,7 @@ static void offset_reset_stipple_counter( struct draw_stage *stage )
 
 static void offset_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -171,7 +171,7 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw )
 {
    struct offset_stage *offset = CALLOC_STRUCT(offset_stage);
 
-   draw_alloc_tmps( &offset->stage, 3 );
+   draw_alloc_temp_verts( &offset->stage, 3 );
 
    offset->stage.draw = draw;
    offset->stage.next = NULL;
diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c
index 51e2242719a..51e2242719a 100644
--- a/src/mesa/pipe/draw/draw_prim.c
+++ b/src/gallium/auxiliary/draw/draw_prim.c
diff --git a/src/mesa/pipe/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 7782db04770..bc11259cb21 100644
--- a/src/mesa/pipe/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -45,7 +45,7 @@
 #include "pipe/p_defines.h"
 
 #include "x86/rtasm/x86sse.h"
-#include "pipe/tgsi/exec/tgsi_exec.h"
+#include "tgsi/exec/tgsi_exec.h"
 
 
 struct gallivm_prog;
@@ -128,13 +128,25 @@ struct draw_stage
  * Private version of the compiled vertex_shader
  */
 struct draw_vertex_shader {
+
+   /* This member will disappear shortly:
+    */
    const struct pipe_shader_state   *state;
-#if defined(__i386__) || defined(__386__)
-   struct x86_function              sse2_program;
-#endif
-#ifdef MESA_LLVM
-   struct gallivm_prog *llvm_prog;
-#endif
+
+   void (*prepare)( struct draw_vertex_shader *shader,
+		    struct draw_context *draw );
+
+   /* Run the shader - this interface will get cleaned up in the
+    * future:
+    */
+   void (*run)( struct draw_vertex_shader *shader,
+		struct draw_context *draw,
+		const unsigned *elts, 
+		unsigned count,
+		struct vertex_header *vOut[] );
+		    
+
+   void (*delete)( struct draw_vertex_shader * );
 };
 
 
@@ -176,7 +188,7 @@ struct draw_context
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX];
    struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX];
-   const struct draw_vertex_shader *vertex_shader;
+   struct draw_vertex_shader *vertex_shader;
 
    uint num_vs_outputs;  /**< convenience, from vertex_shader */
 
@@ -201,6 +213,7 @@ struct draw_context
 
    boolean convert_wide_points; /**< convert wide points to tris? */
    boolean convert_wide_lines;  /**< convert side lines to tris? */
+   boolean use_sse;
 
    unsigned reduced_prim;
 
@@ -255,11 +268,10 @@ struct draw_context
       unsigned queue_nr;
    } pq;
 
-   int use_sse : 1;
-#ifdef MESA_LLVM
-   struct gallivm_cpu_engine *engine;
-#endif
-   
+
+   /* This (and the tgsi_exec_machine struct) probably need to be moved somewhere private.
+    */
+   struct gallivm_cpu_engine *engine;   
    void *driver_private;
 };
 
@@ -276,9 +288,9 @@ extern struct draw_stage *draw_wide_stage( struct draw_context *context );
 extern struct draw_stage *draw_validate_stage( struct draw_context *context );
 
 
-extern void draw_free_tmps( struct draw_stage *stage );
-extern void draw_reset_tmps( struct draw_stage *stage );
-extern void draw_alloc_tmps( struct draw_stage *stage, unsigned nr );
+extern void draw_free_temp_verts( struct draw_stage *stage );
+
+extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr );
 
 extern void draw_reset_vertex_ids( struct draw_context *draw );
 
@@ -290,11 +302,7 @@ extern void draw_vertex_cache_invalidate( struct draw_context *draw );
 extern void draw_vertex_cache_unreference( struct draw_context *draw );
 extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw );
 
-
 extern void draw_vertex_shader_queue_flush( struct draw_context *draw );
-#ifdef MESA_LLVM
-extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw );
-#endif
 
 struct tgsi_exec_machine;
 
diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/gallium/auxiliary/draw/draw_stipple.c
index 90291019165..506f33512c8 100644
--- a/src/mesa/pipe/draw/draw_stipple.c
+++ b/src/gallium/auxiliary/draw/draw_stipple.c
@@ -212,6 +212,7 @@ passthrough_tri(struct draw_stage *stage, struct prim_header *header)
 static void 
 stipple_destroy( struct draw_stage *stage )
 {
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -223,7 +224,7 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw )
 {
    struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage);
 
-   draw_alloc_tmps( &stipple->stage, 2 );
+   draw_alloc_temp_verts( &stipple->stage, 2 );
 
    stipple->stage.draw = draw;
    stipple->stage.next = NULL;
diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/gallium/auxiliary/draw/draw_twoside.c
index ad2aaf10bb9..1c389579871 100644
--- a/src/mesa/pipe/draw/draw_twoside.c
+++ b/src/gallium/auxiliary/draw/draw_twoside.c
@@ -176,7 +176,7 @@ static void twoside_reset_stipple_counter( struct draw_stage *stage )
 
 static void twoside_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -188,7 +188,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw )
 {
    struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage);
 
-   draw_alloc_tmps( &twoside->stage, 3 );
+   draw_alloc_temp_verts( &twoside->stage, 3 );
 
    twoside->stage.draw = draw;
    twoside->stage.next = NULL;
diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/gallium/auxiliary/draw/draw_unfilled.c
index 364bda8b791..8777cfdfc86 100644
--- a/src/mesa/pipe/draw/draw_unfilled.c
+++ b/src/gallium/auxiliary/draw/draw_unfilled.c
@@ -165,6 +165,8 @@ static void unfilled_flush( struct draw_stage *stage,
 			    unsigned flags )
 {
    stage->next->flush( stage->next, flags );
+
+   stage->tri = unfilled_first_tri;
 }
 
 
@@ -176,7 +178,7 @@ static void unfilled_reset_stipple_counter( struct draw_stage *stage )
 
 static void unfilled_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -188,7 +190,7 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw )
 {
    struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage);
 
-   draw_alloc_tmps( &unfilled->stage, 0 );
+   draw_alloc_temp_verts( &unfilled->stage, 0 );
 
    unfilled->stage.draw = draw;
    unfilled->stage.next = NULL;
diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c
index 4375ebabbc4..4375ebabbc4 100644
--- a/src/mesa/pipe/draw/draw_validate.c
+++ b/src/gallium/auxiliary/draw/draw_validate.c
diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c
index be96c8fdeb1..71ac73912b8 100644
--- a/src/mesa/pipe/draw/draw_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_vbuf.c
@@ -523,6 +523,9 @@ static void vbuf_destroy( struct draw_stage *stage )
    if(vbuf->vf)
       draw_vf_destroy( vbuf->vf );
 
+   if (vbuf->render)
+      vbuf->render->destroy( vbuf->render );
+
    FREE( stage );
 }
 
diff --git a/src/mesa/pipe/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h
index cfd2b9820c1..cfd2b9820c1 100644
--- a/src/mesa/pipe/draw/draw_vbuf.h
+++ b/src/gallium/auxiliary/draw/draw_vbuf.h
diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c
index 2d6592150fc..daf1ef4b80f 100644
--- a/src/mesa/pipe/draw/draw_vertex.c
+++ b/src/gallium/auxiliary/draw/draw_vertex.c
@@ -34,8 +34,8 @@
  */
 
 
-#include "pipe/draw/draw_private.h"
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vertex.h"
 
 
 /**
diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h
index dfc637b19b1..267c74203bd 100644
--- a/src/mesa/pipe/draw/draw_vertex.h
+++ b/src/gallium/auxiliary/draw/draw_vertex.h
@@ -25,7 +25,12 @@
  * 
  **************************************************************************/
 
-/* Author:
+/**
+ * Post-transform vertex format info.  The vertex_info struct is used by
+ * the draw_vbuf code to emit hardware-specific vertex layouts into hw
+ * vertex buffers.
+ *
+ * Author:
  *    Brian Paul
  */
 
diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c
index 44427999ccf..44427999ccf 100644
--- a/src/mesa/pipe/draw/draw_vertex_cache.c
+++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c
diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c
index e13df04605c..e13df04605c 100644
--- a/src/mesa/pipe/draw/draw_vertex_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c
diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c
new file mode 100644
index 00000000000..f68f6e32440
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c
@@ -0,0 +1,126 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "pipe/p_util.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw_private.h"
+#include "draw_context.h"
+#include "draw_vs.h"
+
+
+
+/**
+ * Run the vertex shader on all vertices in the vertex queue.
+ * Called by the draw module when the vertx cache needs to be flushed.
+ */
+void
+draw_vertex_shader_queue_flush(struct draw_context *draw)
+{
+   struct draw_vertex_shader *shader = draw->vertex_shader;
+   unsigned i;
+
+   assert(draw->vs.queue_nr != 0);
+
+   /* XXX: do this on statechange: 
+    */
+   shader->prepare( shader, draw );
+
+//   fprintf(stderr, " q(%d) ", draw->vs.queue_nr );
+
+   /* run vertex shader on vertex cache entries, four per invokation */
+   for (i = 0; i < draw->vs.queue_nr; i += 4) {
+      struct vertex_header *dests[4];
+      unsigned elts[4];
+      int j, n = MIN2(4, draw->vs.queue_nr - i);
+
+      for (j = 0; j < n; j++) {
+         elts[j] = draw->vs.queue[i + j].elt;
+         dests[j] = draw->vs.queue[i + j].dest;
+      }
+
+      for ( ; j < 4; j++) {
+	 elts[j] = elts[0];
+	 dests[j] = dests[0];
+      }
+
+      assert(n > 0);
+      assert(n <= 4);
+
+      shader->run(shader, draw, elts, n, dests);
+   }
+
+   draw->vs.queue_nr = 0;
+}
+
+
+struct draw_vertex_shader *
+draw_create_vertex_shader(struct draw_context *draw,
+                          const struct pipe_shader_state *shader)
+{
+   struct draw_vertex_shader *vs;
+
+   vs = draw_create_vs_llvm( draw, shader );
+   if (vs)
+      return vs;
+
+   vs = draw_create_vs_sse( draw, shader );
+   if (vs)
+      return vs;
+
+   vs = draw_create_vs_exec( draw, shader );
+   assert(vs);
+   return vs;
+}
+
+
+void
+draw_bind_vertex_shader(struct draw_context *draw,
+                        struct draw_vertex_shader *dvs)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+
+   draw->vertex_shader = dvs;
+   draw->num_vs_outputs = dvs->state->num_outputs;
+
+   tgsi_exec_machine_init(&draw->machine);
+
+   dvs->prepare( dvs, draw );
+}
+
+
+void
+draw_delete_vertex_shader(struct draw_context *draw,
+                          struct draw_vertex_shader *dvs)
+{
+   dvs->delete( dvs );
+}
diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c
index f23d7fcec5c..dc3a5ecd219 100644
--- a/src/mesa/pipe/draw/draw_vf.c
+++ b/src/gallium/auxiliary/draw/draw_vf.c
@@ -370,7 +370,7 @@ void draw_vf_emit_vertex( struct draw_vertex_fetch *vf,
    unsigned j;
    
    for (j = 0; j < vf->attr_count; j++) {
-      if(!a[j].isconst) {
+      if (!a[j].isconst) {
 	 a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0];
 	 a[j].inputstride = 0; /* XXX: one-vertex-max ATM */
       }
diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/gallium/auxiliary/draw/draw_vf.h
index e694b986754..011c8f0ff1c 100644
--- a/src/mesa/pipe/draw/draw_vf.h
+++ b/src/gallium/auxiliary/draw/draw_vf.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2003 Tungsten Graphics, inc.
+ * Copyright 2008 Tungsten Graphics, inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -20,6 +20,17 @@
  * 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.
+ */
+
+
+/**
+ * Vertex fetch/store/convert code.  This functionality is used in two places:
+ * 1. Vertex fetch/convert - to grab vertex data from incoming vertex
+ *    arrays and convert to format needed by vertex shaders.
+ * 2. Vertex store/emit - to convert simple float[][4] vertex attributes
+ *    (which is the organization used throughout the draw/prim pipeline) to
+ *    hardware-specific formats and emit into hardware vertex buffers.
+ *
  *
  * Authors:
  *    Keith Whitwell <[email protected]>
@@ -33,7 +44,7 @@
 #include "pipe/p_state.h"
 
 #include "draw_vertex.h"
-#include "draw_private.h" // for vertex_header
+#include "draw_private.h" /* for vertex_header */
 
 
 enum draw_vf_attr_format {
@@ -181,6 +192,7 @@ struct draw_vf_attr_type {
    unsigned offset;
 };
 
+/** XXX this could be moved into draw_vf.c */
 struct draw_vf_fastpath {
    unsigned vertex_stride;
    unsigned attr_count;
@@ -209,6 +221,7 @@ void
 draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf );
 
 
+/** XXX this type and function could probably be moved into draw_vf.c */
 struct draw_vf_format_info {
    const char *name;
    draw_vf_insert_func insert[4];
diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/gallium/auxiliary/draw/draw_vf_generic.c
index 7a60a9db9c3..7a60a9db9c3 100644
--- a/src/mesa/pipe/draw/draw_vf_generic.c
+++ b/src/gallium/auxiliary/draw/draw_vf_generic.c
diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c
index 1ad2ae756dd..1ad2ae756dd 100644
--- a/src/mesa/pipe/draw/draw_vf_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vf_sse.c
diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h
new file mode 100644
index 00000000000..4ee7e705e93
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs.h
@@ -0,0 +1,50 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <[email protected]>
+ */
+
+#ifndef DRAW_VS_H
+#define DRAW_VS_H
+
+struct draw_vertex_shader;
+struct draw_context;
+struct pipe_shader_state;
+
+struct draw_vertex_shader *
+draw_create_vs_exec(struct draw_context *draw,
+		    const struct pipe_shader_state *templ);
+
+struct draw_vertex_shader *
+draw_create_vs_sse(struct draw_context *draw,
+		   const struct pipe_shader_state *templ);
+
+struct draw_vertex_shader *
+draw_create_vs_llvm(struct draw_context *draw,
+		    const struct pipe_shader_state *templ);
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
new file mode 100644
index 00000000000..8588879400a
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -0,0 +1,186 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "pipe/p_util.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+#include "draw_vs.h"
+
+
+static INLINE unsigned
+compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
+{
+   unsigned mask = 0;
+   unsigned i;
+
+   /* Do the hardwired planes first:
+    */
+   if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT;
+   if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT;
+   if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT;
+   if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT;
+   if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT;
+   if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT;
+
+   /* Followed by any remaining ones:
+    */
+   for (i = 6; i < nr; i++) {
+      if (dot4(clip, plane[i]) < 0) 
+         mask |= (1<<i);
+   }
+
+   return mask;
+}
+
+
+static void
+vs_exec_prepare( struct draw_vertex_shader *shader,
+		 struct draw_context *draw )
+{
+   /* specify the vertex program to interpret/execute */
+   tgsi_exec_machine_bind_shader(&draw->machine,
+				 shader->state->tokens,
+				 PIPE_MAX_SAMPLERS,
+				 NULL /*samplers*/ );
+
+   draw_update_vertex_fetch( draw );
+}
+
+
+/**
+ * Transform vertices with the current vertex program/shader
+ * Up to four vertices can be shaded at a time.
+ * \param vbuffer  the input vertex data
+ * \param elts  indexes of four input vertices
+ * \param count  number of vertices to shade [1..4]
+ * \param vOut  array of pointers to four output vertices
+ */
+static void
+vs_exec_run( struct draw_vertex_shader *shader,
+	     struct draw_context *draw,
+	     const unsigned *elts, 
+	     unsigned count,
+	     struct vertex_header *vOut[] )
+{
+   struct tgsi_exec_machine *machine = &draw->machine;
+   unsigned int j;
+
+   ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX);
+   ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX);
+   const float *scale = draw->viewport.scale;
+   const float *trans = draw->viewport.translate;
+
+   assert(count <= 4);
+   assert(draw->vertex_shader->state->output_semantic_name[0]
+          == TGSI_SEMANTIC_POSITION);
+
+   machine->Consts = (float (*)[4]) draw->user.constants;
+   machine->Inputs = ALIGN16_ASSIGN(inputs);
+   machine->Outputs = ALIGN16_ASSIGN(outputs);
+
+   draw->vertex_fetch.fetch_func( draw, machine, elts, count );
+
+   /* run interpreter */
+   tgsi_exec_machine_run( machine );
+
+
+   /* store machine results */
+   for (j = 0; j < count; j++) {
+      unsigned slot;
+      float x, y, z, w;
+
+      /* Handle attr[0] (position) specially:
+       *
+       * XXX: Computing the clipmask should be done in the vertex
+       * program as a set of DP4 instructions appended to the
+       * user-provided code.
+       */
+      x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j];
+      y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j];
+      z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j];
+      w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j];
+
+      vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes);
+      vOut[j]->edgeflag = 1;
+
+      /* divide by w */
+      w = 1.0f / w;
+      x *= w;
+      y *= w;
+      z *= w;
+
+      /* Viewport mapping */
+      vOut[j]->data[0][0] = x * scale[0] + trans[0];
+      vOut[j]->data[0][1] = y * scale[1] + trans[1];
+      vOut[j]->data[0][2] = z * scale[2] + trans[2];
+      vOut[j]->data[0][3] = w;
+
+      /* Remaining attributes are packed into sequential post-transform
+       * vertex attrib slots.
+       */
+      for (slot = 1; slot < draw->num_vs_outputs; slot++) {
+         vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+         vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+         vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+         vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+      }
+   } /* loop over vertices */
+}
+
+
+
+static void
+vs_exec_delete( struct draw_vertex_shader *dvs )
+{
+   FREE( dvs );
+}
+
+
+struct draw_vertex_shader *
+draw_create_vs_exec(struct draw_context *draw,
+		    const struct pipe_shader_state *state)
+{
+   struct draw_vertex_shader *vs = CALLOC_STRUCT( draw_vertex_shader );
+
+   if (vs == NULL) 
+      return NULL;
+
+   vs->state = state;
+   vs->prepare = vs_exec_prepare;
+   vs->run = vs_exec_run;
+   vs->delete = vs_exec_delete;
+
+   return vs;
+}
diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c
index 9c31df1e3e1..44022b6e077 100644
--- a/src/mesa/pipe/draw/draw_vertex_shader.c
+++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c
@@ -27,23 +27,25 @@
 
  /*
   * Authors:
+  *   Zack Rusin
   *   Keith Whitwell <[email protected]>
   *   Brian Paul
   */
 
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#if defined(__i386__) || defined(__386__)
-#include "pipe/tgsi/exec/tgsi_sse2.h"
-#endif
 #include "draw_private.h"
 #include "draw_context.h"
+#include "draw_vs.h"
 
-#include "x86/rtasm/x86sse.h"
-#include "pipe/llvm/gallivm.h"
+#ifdef MESA_LLVM
 
+#include "llvm/gallivm.h"
 
-#define DBG_VS 0
+struct draw_llvm_vertex_shader {
+   struct draw_vertex_shader base;
+   struct gallivm_prog *llvm_prog;
+};
 
 
 static INLINE unsigned
@@ -72,11 +74,14 @@ compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
 }
 
 
-typedef void (XSTDCALL *codegen_function) (
-   const struct tgsi_exec_vector *input,
-   struct tgsi_exec_vector *output,
-   float (*constant)[4],
-   struct tgsi_exec_vector *temporary );
+
+static void
+vs_llvm_prepare( struct draw_vertex_shader *base,
+		 struct draw_context *draw )
+{
+   draw_update_vertex_fetch( draw );
+}
+
 
 
 /**
@@ -88,10 +93,15 @@ typedef void (XSTDCALL *codegen_function) (
  * \param vOut  array of pointers to four output vertices
  */
 static void
-run_vertex_program(struct draw_context *draw,
-                   unsigned elts[4], unsigned count,
-                   struct vertex_header *vOut[])
+vs_llvm_run( struct draw_vertex_shader *base,
+	     struct draw_context *draw, 
+	     const unsigned *elts, 
+	     unsigned count,
+	     struct vertex_header *vOut[] )
 {
+   struct draw_llvm_vertex_shader *shader = 
+      (struct draw_llvm_vertex_shader *)base;
+
    struct tgsi_exec_machine *machine = &draw->machine;
    unsigned int j;
 
@@ -100,6 +110,7 @@ run_vertex_program(struct draw_context *draw,
    const float *scale = draw->viewport.scale;
    const float *trans = draw->viewport.translate;
 
+
    assert(count <= 4);
    assert(draw->vertex_shader->state->output_semantic_name[0]
           == TGSI_SEMANTIC_POSITION);
@@ -113,51 +124,17 @@ run_vertex_program(struct draw_context *draw,
    draw->vertex_fetch.fetch_func( draw, machine, elts, count );
 
    /* run shader */
-#ifdef MESA_LLVM
-   if (1) {
-   struct gallivm_prog  *prog  = draw->vertex_shader->llvm_prog;
-   gallivm_cpu_vs_exec(prog,
+   gallivm_cpu_vs_exec(shader->llvm_prog,
                        machine->Inputs,
                        machine->Outputs,
-                       machine->Consts);
-   } else
-#elif defined(__i386__) || defined(__386__)
-   if (draw->use_sse) {
-      /* SSE */
-      /* cast away const */
-      struct draw_vertex_shader *shader
-         = (struct draw_vertex_shader *)draw->vertex_shader;
-      codegen_function func
-         = (codegen_function) x86_get_func( &shader->sse2_program );
-
-      if (func)
-         func(
-            machine->Inputs,
-            machine->Outputs,
-            machine->Consts,
-            machine->Temps );
-      else
-         /* interpreter */
-         tgsi_exec_machine_run( machine );
-   }
-   else
-#endif
-   {
-      /* interpreter */
-      tgsi_exec_machine_run( machine );
-   }
+                       machine->Consts,
+                       machine->Temps);
 
    /* store machine results */
    for (j = 0; j < count; j++) {
       unsigned slot;
       float x, y, z, w;
 
-      /* Handle attr[0] (position) specially:
-       *
-       * XXX: Computing the clipmask should be done in the vertex
-       * program as a set of DP4 instructions appended to the
-       * user-provided code.
-       */
       x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j];
       y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j];
       z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j];
@@ -178,13 +155,6 @@ run_vertex_program(struct draw_context *draw,
       vOut[j]->data[0][2] = z * scale[2] + trans[2];
       vOut[j]->data[0][3] = w;
 
-#if DBG_VS
-      debug_printf("output[%d]win: %f %f %f %f\n", j,
-             vOut[j]->data[0][0],
-             vOut[j]->data[0][1],
-             vOut[j]->data[0][2],
-             vOut[j]->data[0][3]);
-#endif
       /* Remaining attributes are packed into sequential post-transform
        * vertex attrib slots.
        */
@@ -193,132 +163,75 @@ run_vertex_program(struct draw_context *draw,
          vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
          vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
          vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
-#if DBG_VS
-         debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot,
-                vOut[j]->data[slot][0],
-                vOut[j]->data[slot][1],
-                vOut[j]->data[slot][2],
-                vOut[j]->data[slot][3]);
-#endif
       }
    } /* loop over vertices */
 }
 
-
-/**
- * Run the vertex shader on all vertices in the vertex queue.
- * Called by the draw module when the vertx cache needs to be flushed.
- */
-void
-draw_vertex_shader_queue_flush(struct draw_context *draw)
+static void
+vs_llvm_delete( struct draw_vertex_shader *base )
 {
-   unsigned i;
+   struct draw_llvm_vertex_shader *shader = 
+      (struct draw_llvm_vertex_shader *)base;
 
-   assert(draw->vs.queue_nr != 0);
-
-   /* XXX: do this on statechange: 
+   /* Do something to free compiled shader:
     */
-   draw_update_vertex_fetch( draw );
-
-//   fprintf(stderr, " q(%d) ", draw->vs.queue_nr );
 
-   /* run vertex shader on vertex cache entries, four per invokation */
-   for (i = 0; i < draw->vs.queue_nr; i += 4) {
-      struct vertex_header *dests[4];
-      unsigned elts[4];
-      int j, n = MIN2(4, draw->vs.queue_nr - i);
-
-      for (j = 0; j < n; j++) {
-         elts[j] = draw->vs.queue[i + j].elt;
-         dests[j] = draw->vs.queue[i + j].dest;
-      }
-
-      for ( ; j < 4; j++) {
-	 elts[j] = elts[0];
-	 dests[j] = dests[0];
-      }
+   FREE( shader );
+}
 
-      assert(n > 0);
-      assert(n <= 4);
 
-      run_vertex_program(draw, elts, n, dests);
-   }
-
-   draw->vs.queue_nr = 0;
-}
 
 
 struct draw_vertex_shader *
-draw_create_vertex_shader(struct draw_context *draw,
-                          const struct pipe_shader_state *shader)
+draw_create_vs_llvm(struct draw_context *draw,
+		    const struct pipe_shader_state *templ)
 {
-   struct draw_vertex_shader *vs;
+   struct draw_llvm_vertex_shader *vs;
 
-   vs = CALLOC_STRUCT( draw_vertex_shader );
-   if (vs == NULL) {
+   vs = CALLOC_STRUCT( draw_llvm_vertex_shader );
+   if (vs == NULL) 
       return NULL;
-   }
 
-   vs->state = shader;
+   vs->base.state = templ;
+   vs->base.prepare = vs_llvm_prepare;
+   vs->base.run = vs_llvm_run;
+   vs->base.delete = vs_llvm_delete;
 
-#ifdef MESA_LLVM
-   struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS);
-   gallivm_ir_set_layout(ir, GALLIVM_SOA);
-   gallivm_ir_set_components(ir, 4);
-   gallivm_ir_fill_from_tgsi(ir, shader->tokens);
-   vs->llvm_prog = gallivm_ir_compile(ir);
-   gallivm_ir_delete(ir);
+   {
+      struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS);
+      gallivm_ir_set_layout(ir, GALLIVM_SOA);
+      gallivm_ir_set_components(ir, 4);
+      gallivm_ir_fill_from_tgsi(ir, vs->base.state->tokens);
+      vs->llvm_prog = gallivm_ir_compile(ir);
+      gallivm_ir_delete(ir);
+   }
 
    draw->engine = gallivm_global_cpu_engine();
+
+   /* XXX: Why are there two versions of this?  Shouldn't creating the
+    *      engine be a separate operation to compiling a shader?
+    */
    if (!draw->engine) {
       draw->engine = gallivm_cpu_engine_create(vs->llvm_prog);
    }
    else {
       gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog);
    }
-#elif defined(__i386__) || defined(__386__)
-   if (draw->use_sse) {
-      /* cast-away const */
-      struct pipe_shader_state *sh = (struct pipe_shader_state *) shader;
-
-      x86_init_func( &vs->sse2_program );
-      if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens,
-                           &vs->sse2_program )) {
-         x86_release_func( (struct x86_function *) &vs->sse2_program );
-	 fprintf(stdout /*err*/,
-		 "tgsi_emit_sse2() failed, falling back to interpreter\n");
-      }
-   }
-#endif
 
-   return vs;
+   return &vs->base;
 }
 
 
-void
-draw_bind_vertex_shader(struct draw_context *draw,
-                        struct draw_vertex_shader *dvs)
-{
-   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
 
-   draw->vertex_shader = dvs;
-   draw->num_vs_outputs = dvs->state->num_outputs;
 
-   /* specify the vertex program to interpret/execute */
-   tgsi_exec_machine_init(&draw->machine,
-                          draw->vertex_shader->state->tokens,
-                          PIPE_MAX_SAMPLERS,
-                          NULL /*samplers*/ );
-}
 
+#else
 
-void
-draw_delete_vertex_shader(struct draw_context *draw,
-                          struct draw_vertex_shader *dvs)
+struct draw_vertex_shader *
+draw_create_vs_llvm(struct draw_context *draw,
+                          const struct pipe_shader_state *shader)
 {
-#if defined(__i386__) || defined(__386__)
-   x86_release_func( (struct x86_function *) &dvs->sse2_program );
-#endif
-
-   FREE( dvs );
+   return NULL;
 }
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
new file mode 100644
index 00000000000..27bc66812c3
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -0,0 +1,251 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "draw_vs.h"
+
+#if defined(__i386__) || defined(__386__)
+
+#include "pipe/p_util.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+
+#include "x86/rtasm/x86sse.h"
+#include "tgsi/exec/tgsi_sse2.h"
+
+
+typedef void (XSTDCALL *codegen_function) (
+   const struct tgsi_exec_vector *input,
+   struct tgsi_exec_vector *output,
+   float (*constant)[4],
+   struct tgsi_exec_vector *temporary );
+
+
+struct draw_sse_vertex_shader {
+   struct draw_vertex_shader base;
+   struct x86_function sse2_program;
+   codegen_function func;
+};
+
+
+/* Should be part of the generated shader:
+ */
+static INLINE unsigned
+compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
+{
+   unsigned mask = 0;
+   unsigned i;
+
+   /* Do the hardwired planes first:
+    */
+   if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT;
+   if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT;
+   if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT;
+   if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT;
+   if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT;
+   if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT;
+
+   /* Followed by any remaining ones:
+    */
+   for (i = 6; i < nr; i++) {
+      if (dot4(clip, plane[i]) < 0) 
+         mask |= (1<<i);
+   }
+
+   return mask;
+}
+
+
+static void
+vs_sse_prepare( struct draw_vertex_shader *base,
+		struct draw_context *draw )
+{
+   draw_update_vertex_fetch( draw );
+}
+
+/**
+ * Transform vertices with the current vertex program/shader
+ * Up to four vertices can be shaded at a time.
+ * \param vbuffer  the input vertex data
+ * \param elts  indexes of four input vertices
+ * \param count  number of vertices to shade [1..4]
+ * \param vOut  array of pointers to four output vertices
+ */
+static void
+vs_sse_run( struct draw_vertex_shader *base,
+	    struct draw_context *draw, 
+	    const unsigned *elts, 
+	    unsigned count,
+	    struct vertex_header *vOut[] )
+{
+   struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base;
+   struct tgsi_exec_machine *machine = &draw->machine;
+   unsigned int j;
+
+   ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX);
+   ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX);
+   const float *scale = draw->viewport.scale;
+   const float *trans = draw->viewport.translate;
+
+   assert(count <= 4);
+   assert(draw->vertex_shader->state->output_semantic_name[0]
+          == TGSI_SEMANTIC_POSITION);
+
+   /* Consts does not require 16 byte alignment. */
+   machine->Consts = (float (*)[4]) draw->user.constants;
+   machine->Inputs = ALIGN16_ASSIGN(inputs);
+   machine->Outputs = ALIGN16_ASSIGN(outputs);
+
+
+   /* Fetch vertices.  This may at some point be integrated into the
+    * compiled shader -- that would require a reorganization where
+    * multiple versions of the compiled shader might exist,
+    * specialized for each fetch state.
+    */
+   draw->vertex_fetch.fetch_func( draw, machine, elts, count );
+
+
+   /* run compiled shader
+    */   
+   shader->func(
+      machine->Inputs,
+      machine->Outputs,
+      machine->Consts,
+      machine->Temps );
+
+
+   /* XXX: Computing the clipmask and emitting results should be done
+    *      in the vertex program as a set of instructions appended to
+    *      the user-provided code.
+    */
+   for (j = 0; j < count; j++) {
+      unsigned slot;
+      float x, y, z, w;
+
+      x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j];
+      y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j];
+      z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j];
+      w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j];
+
+      vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes);
+      vOut[j]->edgeflag = 1;
+
+      /* divide by w */
+      w = 1.0f / w;
+      x *= w;
+      y *= w;
+      z *= w;
+
+      /* Viewport mapping */
+      vOut[j]->data[0][0] = x * scale[0] + trans[0];
+      vOut[j]->data[0][1] = y * scale[1] + trans[1];
+      vOut[j]->data[0][2] = z * scale[2] + trans[2];
+      vOut[j]->data[0][3] = w;
+
+      /* Remaining attributes are packed into sequential post-transform
+       * vertex attrib slots.
+       */
+      for (slot = 1; slot < draw->num_vs_outputs; slot++) {
+         vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+         vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+         vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+         vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+      }
+   } 
+}
+
+
+
+static void
+vs_sse_delete( struct draw_vertex_shader *base )
+{
+   struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base;
+   
+   x86_release_func( &shader->sse2_program );
+
+   FREE( shader );
+}
+
+
+struct draw_vertex_shader *
+draw_create_vs_sse(struct draw_context *draw,
+                          const struct pipe_shader_state *templ)
+{
+   struct draw_sse_vertex_shader *vs;
+
+   if (!draw->use_sse) 
+      return NULL;
+
+   vs = CALLOC_STRUCT( draw_sse_vertex_shader );
+   if (vs == NULL) 
+      return NULL;
+
+   vs->base.state = templ;
+   vs->base.prepare = vs_sse_prepare;
+   vs->base.run = vs_sse_run;
+   vs->base.delete = vs_sse_delete;
+   
+   x86_init_func( &vs->sse2_program );
+
+   if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state->tokens,
+			&vs->sse2_program )) 
+      goto fail;
+      
+   vs->func = (codegen_function) x86_get_func( &vs->sse2_program );
+   
+   return &vs->base;
+
+fail:
+   fprintf(stderr, "tgsi_emit_sse2() failed, falling back to interpreter\n");
+
+   x86_release_func( &vs->sse2_program );
+   
+   FREE(vs);
+   return NULL;
+}
+
+
+
+#else
+
+struct draw_vertex_shader *
+draw_create_vs_sse( struct draw_context *draw,
+		    const struct pipe_shader_state *templ )
+{
+   return (void *) 0;
+}
+
+
+#endif
+
diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c
index 4c7e279b20f..655774b155f 100644
--- a/src/mesa/pipe/draw/draw_wide_prims.c
+++ b/src/gallium/auxiliary/draw/draw_wide_prims.c
@@ -175,7 +175,7 @@ static void wide_line_aa(struct draw_stage *stage,
    float *pos;
    float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0];
    float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1];
-   const float len = sqrt(dx * dx + dy * dy);
+   const float len = (float) sqrt(dx * dx + dy * dy);
    uint i;
 
    dx = dx * half_width / len;
@@ -400,7 +400,7 @@ static void wide_flush( struct draw_stage *stage, unsigned flags )
 }
 
 
-static void draw_reset_stipple_counter( struct draw_stage *stage )
+static void wide_reset_stipple_counter( struct draw_stage *stage )
 {
    stage->next->reset_stipple_counter( stage->next );
 }
@@ -408,7 +408,7 @@ static void draw_reset_stipple_counter( struct draw_stage *stage )
 
 static void wide_destroy( struct draw_stage *stage )
 {
-   draw_free_tmps( stage );
+   draw_free_temp_verts( stage );
    FREE( stage );
 }
 
@@ -417,7 +417,7 @@ struct draw_stage *draw_wide_stage( struct draw_context *draw )
 {
    struct wide_stage *wide = CALLOC_STRUCT(wide_stage);
 
-   draw_alloc_tmps( &wide->stage, 4 );
+   draw_alloc_temp_verts( &wide->stage, 4 );
 
    wide->stage.draw = draw;
    wide->stage.next = NULL;
@@ -425,7 +425,7 @@ struct draw_stage *draw_wide_stage( struct draw_context *draw )
    wide->stage.line = wide_first_line;
    wide->stage.tri = passthrough_tri;
    wide->stage.flush = wide_flush;
-   wide->stage.reset_stipple_counter = draw_reset_stipple_counter;
+   wide->stage.reset_stipple_counter = wide_reset_stipple_counter;
    wide->stage.destroy = wide_destroy;
 
    return &wide->stage;
diff --git a/src/mesa/pipe/llvm/Makefile b/src/gallium/auxiliary/llvm/Makefile
index a0494ba966e..e0abf860c17 100644
--- a/src/mesa/pipe/llvm/Makefile
+++ b/src/gallium/auxiliary/llvm/Makefile
@@ -15,7 +15,7 @@ GALLIVM_SOURCES = \
         storagesoa.cpp \
         instructionssoa.cpp
 
-INC_SOURCES = gallivm_builtins.cpp llvm_base_shader.cpp
+INC_SOURCES = gallivm_builtins.cpp
 
 CPP_SOURCES = \
 	$(GALLIVM_SOURCES)
@@ -30,7 +30,9 @@ OBJECTS = $(C_SOURCES:.c=.o) \
 ### Include directories
 INCLUDES = \
 	-I. \
-	-I$(TOP)/src/mesa/pipe \
+	-I$(TOP)/src/gallium/drivers
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/include \
 	-I$(TOP)/src/mesa \
 	-I$(TOP)/include
 
@@ -65,8 +67,6 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES)
 gallivm_builtins.cpp: llvm_builtins.c
 	clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins
 
-llvm_base_shader.cpp: llvm_entry.c
-	clang --emit-llvm $< |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -o=$@ -f -funcname=createBaseShader
 
 # Emacs tags
 tags:
@@ -77,7 +77,7 @@ tags:
 clean:
 	-rm -f *.o */*.o *~ *.so *~ server/*.o
 	-rm -f depend depend.bak
-	-rm -f gallivm_builtins.cpp llvm_base_shader.cpp
+	-rm -f gallivm_builtins.cpp
 
 symlinks:
 
diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/gallium/auxiliary/llvm/gallivm.cpp
index b99dc6db5b8..d14bb3b99a8 100644
--- a/src/mesa/pipe/llvm/gallivm.cpp
+++ b/src/gallium/auxiliary/llvm/gallivm.cpp
@@ -42,8 +42,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_shader_tokens.h"
 
-#include "pipe/tgsi/exec/tgsi_exec.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
+#include "tgsi/exec/tgsi_exec.h"
+#include "tgsi/util/tgsi_dump.h"
 
 #include <llvm/Module.h>
 #include <llvm/CallingConv.h>
@@ -289,9 +289,9 @@ void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir,
    tgsi_dump(tokens, 0);
 
 
-   llvm::Module *irmod = tgsi_to_llvmir(ir, tokens);
+   llvm::Module *mod = tgsi_to_llvmir(ir, tokens);
 
-   llvm::Module *mod = tgsi_to_llvm(ir, tokens);
+   //llvm::Module *mod = tgsi_to_llvm(ir, tokens);
    ir->module = mod;
    gallivm_ir_dump(ir, 0);
 }
@@ -318,6 +318,9 @@ struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir)
    passes.run(*mod);
    prog->module = mod;
 
+   std::cout << "After optimizations:"<<std::endl;
+   mod->dump();
+
    return prog;
 }
 
diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/gallium/auxiliary/llvm/gallivm.h
index b104520cb7f..92da4bca7f4 100644
--- a/src/mesa/pipe/llvm/gallivm.h
+++ b/src/gallium/auxiliary/llvm/gallivm.h
@@ -82,7 +82,8 @@ struct gallivm_cpu_engine *gallivm_global_cpu_engine();
 int gallivm_cpu_vs_exec(struct gallivm_prog *prog,
                         struct tgsi_exec_vector       *inputs,
                         struct tgsi_exec_vector       *dests,
-                        float (*consts)[4]);
+                        float (*consts)[4],
+                        struct tgsi_exec_vector       *temps);
 int gallivm_cpu_fs_exec(struct gallivm_prog *prog,
                         float x, float y,
                         float (*dests)[PIPE_MAX_SHADER_INPUTS][4],
diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/gallium/auxiliary/llvm/gallivm_builtins.cpp
index 1796f0a1772..1796f0a1772 100644
--- a/src/mesa/pipe/llvm/gallivm_builtins.cpp
+++ b/src/gallium/auxiliary/llvm/gallivm_builtins.cpp
diff --git a/src/mesa/pipe/llvm/gallivm_cpu.cpp b/src/gallium/auxiliary/llvm/gallivm_cpu.cpp
index 5f1268bf4f8..8f9830d0b1e 100644
--- a/src/mesa/pipe/llvm/gallivm_cpu.cpp
+++ b/src/gallium/auxiliary/llvm/gallivm_cpu.cpp
@@ -42,8 +42,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_shader_tokens.h"
 
-#include "pipe/tgsi/exec/tgsi_exec.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
+#include "tgsi/exec/tgsi_exec.h"
+#include "tgsi/util/tgsi_dump.h"
 
 #include <llvm/Module.h>
 #include <llvm/CallingConv.h>
@@ -105,10 +105,10 @@ static inline llvm::Function *func_for_shader(struct gallivm_prog *prog)
 
    switch (prog->type) {
    case GALLIVM_VS:
-      func = mod->getFunction("run_vertex_shader");
+      func = mod->getFunction("vs_shader");
       break;
    case GALLIVM_FS:
-      func = mod->getFunction("run_fragment_shader");
+      func = mod->getFunction("fs_shader");
       break;
    default:
       assert(!"Unknown shader type!");
@@ -177,10 +177,7 @@ struct gallivm_cpu_engine * gallivm_global_cpu_engine()
 typedef void (*vertex_shader_runner)(void *ainputs,
                                      void *dests,
                                      float (*aconsts)[4],
-                                     int num_vertices,
-                                     int num_inputs,
-                                     int num_attribs,
-                                     int num_consts);
+                                     void *temps);
 
 
 /*!
@@ -191,12 +188,13 @@ typedef void (*vertex_shader_runner)(void *ainputs,
 int gallivm_cpu_vs_exec(struct gallivm_prog *prog,
                         struct tgsi_exec_vector       *inputs,
                         struct tgsi_exec_vector       *dests,
-                         float (*consts)[4])
+                        float (*consts)[4],
+                        struct tgsi_exec_vector       *temps)
 {
    vertex_shader_runner runner = reinterpret_cast<vertex_shader_runner>(prog->function);
    assert(runner);
    /*FIXME*/
-   runner(inputs, dests, consts, 4, 4, 4, prog->num_consts);
+   runner(inputs, dests, consts, temps);
 
    return 0;
 }
diff --git a/src/gallium/auxiliary/llvm/gallivm_p.h b/src/gallium/auxiliary/llvm/gallivm_p.h
new file mode 100644
index 00000000000..cfe7b1901b3
--- /dev/null
+++ b/src/gallium/auxiliary/llvm/gallivm_p.h
@@ -0,0 +1,110 @@
+#ifndef GALLIVM_P_H
+#define GALLIVM_P_H
+
+#ifdef MESA_LLVM
+
+#include "gallivm.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_compiler.h"
+
+namespace llvm {
+   class Module;
+}
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+enum gallivm_shader_type;
+enum gallivm_vector_layout;
+
+struct gallivm_interpolate {
+   int attrib;
+   int chan;
+   int type;
+};
+
+struct gallivm_ir {
+   llvm::Module *module;
+   int id;
+   enum gallivm_shader_type type;
+   enum gallivm_vector_layout layout;
+   int num_components;
+   int   num_consts;
+
+   //FIXME: this might not be enough for some shaders
+   struct gallivm_interpolate interpolators[32*4];
+   int   num_interp;
+};
+
+struct gallivm_prog {
+   llvm::Module *module;
+   void *function;
+
+   int   id;
+   enum gallivm_shader_type type;
+
+   int   num_consts;
+
+   //FIXME: this might not be enough for some shaders
+   struct gallivm_interpolate interpolators[32*4];
+   int   num_interp;
+};
+
+static INLINE void gallivm_swizzle_components(int swizzle,
+                                              int *xc, int *yc,
+                                              int *zc, int *wc)
+{
+   int x = swizzle / 1000; swizzle -= x * 1000;
+   int y = swizzle / 100;  swizzle -= y * 100;
+   int z = swizzle / 10;   swizzle -= z * 10;
+   int w = swizzle;
+
+   if (xc) *xc = x;
+   if (yc) *yc = y;
+   if (zc) *zc = z;
+   if (wc) *wc = w;
+}
+
+static INLINE boolean gallivm_is_swizzle(int swizzle)
+{
+   const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 +
+                          TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W;
+   return swizzle != NO_SWIZZLE;
+}
+
+static INLINE int gallivm_x_swizzle(int swizzle)
+{
+   int x;
+   gallivm_swizzle_components(swizzle, &x, 0, 0, 0);
+   return x;
+}
+
+static INLINE int gallivm_y_swizzle(int swizzle)
+{
+   int y;
+   gallivm_swizzle_components(swizzle, 0, &y, 0, 0);
+   return y;
+}
+
+static INLINE int gallivm_z_swizzle(int swizzle)
+{
+   int z;
+   gallivm_swizzle_components(swizzle, 0, 0, &z, 0);
+   return z;
+}
+
+static INLINE int gallivm_w_swizzle(int swizzle)
+{
+   int w;
+   gallivm_swizzle_components(swizzle, 0, 0, 0, &w);
+   return w;
+}
+
+#endif /* MESA_LLVM */
+
+#if defined __cplusplus
+} // extern "C"
+#endif
+
+#endif
diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/gallium/auxiliary/llvm/instructions.cpp
index 55d39fa5f12..55d39fa5f12 100644
--- a/src/mesa/pipe/llvm/instructions.cpp
+++ b/src/gallium/auxiliary/llvm/instructions.cpp
diff --git a/src/mesa/pipe/llvm/instructions.h b/src/gallium/auxiliary/llvm/instructions.h
index 9ebc17dd8ec..9ebc17dd8ec 100644
--- a/src/mesa/pipe/llvm/instructions.h
+++ b/src/gallium/auxiliary/llvm/instructions.h
diff --git a/src/gallium/auxiliary/llvm/instructionssoa.cpp b/src/gallium/auxiliary/llvm/instructionssoa.cpp
new file mode 100644
index 00000000000..a4d50466373
--- /dev/null
+++ b/src/gallium/auxiliary/llvm/instructionssoa.cpp
@@ -0,0 +1,121 @@
+#include "instructionssoa.h"
+
+#include "storagesoa.h"
+
+#include <llvm/Constants.h>
+
+using namespace llvm;
+
+InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func,
+                                 llvm::BasicBlock *block, StorageSoa *storage)
+   : m_builder(block),
+     m_storage(storage),
+     m_idx(0)
+{
+}
+
+const char * InstructionsSoa::name(const char *prefix) const
+{
+   ++m_idx;
+   snprintf(m_name, 32, "%s%d", prefix, m_idx);
+   return m_name;
+}
+
+llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y,
+                                              llvm::Value *z, llvm::Value *w)
+{
+   VectorType  *vectorType = VectorType::get(Type::FloatTy, 4);
+   Constant *constVector = Constant::getNullValue(vectorType);
+   Value *res = m_builder.CreateInsertElement(constVector, x,
+                                              m_storage->constantInt(0),
+                                              name("vecx"));
+   res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1),
+                               name("vecxy"));
+   res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2),
+                               name("vecxyz"));
+   if (w)
+      res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3),
+                                          name("vecxyzw"));
+   return res;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::arl(const std::vector<llvm::Value*> in)
+{
+   std::vector<llvm::Value*> res(4);
+
+   //Extract x's
+   llvm::Value *x1 = m_builder.CreateExtractElement(in[0],
+                                                    m_storage->constantInt(0),
+                                                    name("extractX"));
+   //cast it to an unsigned int
+   x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast"));
+
+   res[0] = x1;//vectorFromVals(x1, x2, x3, x4);
+   //only x is valid. the others shouldn't be necessary
+   /*
+   res[1] = Constant::getNullValue(m_floatVecType);
+   res[2] = Constant::getNullValue(m_floatVecType);
+   res[3] = Constant::getNullValue(m_floatVecType);
+   */
+
+   return res;
+}
+
+
+std::vector<llvm::Value*> InstructionsSoa::add(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx"));
+   res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy"));
+   res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz"));
+   res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw"));
+
+   return res;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::mul(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx"));
+   res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly"));
+   res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz"));
+   res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw"));
+
+   return res;
+}
+
+void InstructionsSoa::end()
+{
+   m_builder.CreateRetVoid();
+}
+
+std::vector<llvm::Value*> InstructionsSoa::madd(const std::vector<llvm::Value*> in1,
+                                                const std::vector<llvm::Value*> in2,
+                                                const std::vector<llvm::Value*> in3)
+{
+   std::vector<llvm::Value*> res = mul(in1, in2);
+   return add(res, in3);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::extractVector(llvm::Value *vector)
+{
+   std::vector<llvm::Value*> res(4);
+   res[0] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(0),
+                                           name("extract1X"));
+   res[1] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(1),
+                                           name("extract2X"));
+   res[2] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(2),
+                                           name("extract3X"));
+   res[3] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(3),
+                                           name("extract4X"));
+
+   return res;
+}
diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/gallium/auxiliary/llvm/instructionssoa.h
index 0b6b41cf056..4169dcbb2eb 100644
--- a/src/mesa/pipe/llvm/instructionssoa.h
+++ b/src/gallium/auxiliary/llvm/instructionssoa.h
@@ -28,6 +28,8 @@
 #ifndef INSTRUCTIONSSOA_H
 #define INSTRUCTIONSSOA_H
 
+#include <llvm/Support/LLVMBuilder.h>
+
 #include <vector>
 
 namespace llvm {
@@ -44,11 +46,28 @@ public:
    InstructionsSoa(llvm::Module *mod, llvm::Function *func,
                    llvm::BasicBlock *block, StorageSoa *storage);
 
+   std::vector<llvm::Value*> arl(const std::vector<llvm::Value*> in);
+
    std::vector<llvm::Value*> add(const std::vector<llvm::Value*> in1,
                                  const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> madd(const std::vector<llvm::Value*> in1,
+                                  const std::vector<llvm::Value*> in2,
+                                  const std::vector<llvm::Value*> in3);
    std::vector<llvm::Value*> mul(const std::vector<llvm::Value*> in1,
                                  const std::vector<llvm::Value*> in2);
    void         end();
+
+   std::vector<llvm::Value*> extractVector(llvm::Value *vector);
+private:
+   const char * name(const char *prefix) const;
+   llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y,
+                               llvm::Value *z, llvm::Value *w);
+private:
+   llvm::LLVMFoldingBuilder  m_builder;
+   StorageSoa *m_storage;
+private:
+   mutable int  m_idx;
+   mutable char m_name[32];
 };
 
 
diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/gallium/auxiliary/llvm/llvm_builtins.c
index 4f98d754baa..4f98d754baa 100644
--- a/src/mesa/pipe/llvm/llvm_builtins.c
+++ b/src/gallium/auxiliary/llvm/llvm_builtins.c
diff --git a/src/mesa/pipe/llvm/loweringpass.cpp b/src/gallium/auxiliary/llvm/loweringpass.cpp
index 556dbec3661..556dbec3661 100644
--- a/src/mesa/pipe/llvm/loweringpass.cpp
+++ b/src/gallium/auxiliary/llvm/loweringpass.cpp
diff --git a/src/mesa/pipe/llvm/loweringpass.h b/src/gallium/auxiliary/llvm/loweringpass.h
index f62dcf6ba73..f62dcf6ba73 100644
--- a/src/mesa/pipe/llvm/loweringpass.h
+++ b/src/gallium/auxiliary/llvm/loweringpass.h
diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/gallium/auxiliary/llvm/storage.cpp
index 08b9d29a24a..c4326de8c53 100644
--- a/src/mesa/pipe/llvm/storage.cpp
+++ b/src/gallium/auxiliary/llvm/storage.cpp
@@ -33,6 +33,8 @@
 
 #include "storage.h"
 
+#include "gallivm_p.h"
+
 #include "pipe/p_shader_tokens.h"
 #include <llvm/BasicBlock.h>
 #include <llvm/Module.h>
@@ -82,10 +84,10 @@ llvm::Constant *Storage::shuffleMask(int vec)
    if (origVec == 0) {
       const_vec = Constant::getNullValue(m_intVecType);
    } else {
-      int x = vec / 1000; vec -= x * 1000;
-      int y = vec / 100;  vec -= y * 100;
-      int z = vec / 10;   vec -= z * 10;
-      int w = vec;
+      int x = gallivm_x_swizzle(vec);
+      int y = gallivm_y_swizzle(vec);
+      int z = gallivm_z_swizzle(vec);
+      int w = gallivm_w_swizzle(vec);
       std::vector<Constant*> elems;
       elems.push_back(constantInt(x));
       elems.push_back(constantInt(y));
diff --git a/src/mesa/pipe/llvm/storage.h b/src/gallium/auxiliary/llvm/storage.h
index 8574f7554e3..8574f7554e3 100644
--- a/src/mesa/pipe/llvm/storage.h
+++ b/src/gallium/auxiliary/llvm/storage.h
diff --git a/src/gallium/auxiliary/llvm/storagesoa.cpp b/src/gallium/auxiliary/llvm/storagesoa.cpp
new file mode 100644
index 00000000000..ed0674a96f0
--- /dev/null
+++ b/src/gallium/auxiliary/llvm/storagesoa.cpp
@@ -0,0 +1,389 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "storagesoa.h"
+
+#include "gallivm_p.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_debug.h"
+
+#include <llvm/BasicBlock.h>
+#include <llvm/Module.h>
+#include <llvm/Value.h>
+
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/InstrTypes.h>
+#include <llvm/Instructions.h>
+
+using namespace llvm;
+
+
+StorageSoa::StorageSoa(llvm::BasicBlock *block,
+                       llvm::Value *input,
+                       llvm::Value *output,
+                       llvm::Value *consts,
+                       llvm::Value *temps)
+   : m_block(block),
+     m_input(input),
+     m_output(output),
+     m_consts(consts),
+     m_temps(temps),
+     m_immediates(0),
+     m_idx(0)
+{
+}
+
+void StorageSoa::addImmediate(float *vec)
+{
+   std::vector<float> vals(4);
+   vals[0] = vec[0];
+   vals[1] = vec[1];
+   vals[2] = vec[2];
+   vals[3] = vec[3];
+   m_immediatesToFlush.push_back(vals);
+}
+
+void StorageSoa::declareImmediates()
+{
+   if (m_immediatesToFlush.empty())
+      return;
+
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   ArrayType  *vectorChannels = ArrayType::get(vectorType, 4);
+   ArrayType  *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size());
+
+   m_immediates = new GlobalVariable(
+      /*Type=*/arrayType,
+      /*isConstant=*/false,
+      /*Linkage=*/GlobalValue::ExternalLinkage,
+      /*Initializer=*/0, // has initializer, specified below
+      /*Name=*/name("immediates"),
+      currentModule());
+
+   std::vector<Constant*> arrayVals;
+   for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) {
+      std::vector<float> vec = m_immediatesToFlush[i];
+      std::vector<float> vals(4);
+      std::vector<Constant*> channelArray;
+
+      vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0];
+      llvm::Constant *xChannel = createConstGlobalVector(vals);
+
+      vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1];
+      llvm::Constant *yChannel = createConstGlobalVector(vals);
+
+      vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2];
+      llvm::Constant *zChannel = createConstGlobalVector(vals);
+
+      vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3];
+      llvm::Constant *wChannel = createConstGlobalVector(vals);
+      channelArray.push_back(xChannel);
+      channelArray.push_back(yChannel);
+      channelArray.push_back(zChannel);
+      channelArray.push_back(wChannel);
+      Constant *constChannels = ConstantArray::get(vectorChannels,
+                                                   channelArray);
+      arrayVals.push_back(constChannels);
+   }
+   Constant *constArray = ConstantArray::get(arrayType, arrayVals);
+   m_immediates->setInitializer(constArray);
+
+   m_immediatesToFlush.clear();
+}
+
+llvm::Value *StorageSoa::addrElement(int idx) const
+{
+   std::map<int, llvm::Value*>::const_iterator itr = m_addresses.find(idx);
+   if (itr == m_addresses.end()) {
+      debug_printf("Trying to access invalid shader 'address'\n");
+      return 0;
+   }
+   llvm::Value * res = (*itr).second;
+
+   res = new LoadInst(res, name("addr"), false, m_block);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::inputElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_input, idx, 0);
+   res[1] = element(m_input, idx, 1);
+   res[2] = element(m_input, idx, 2);
+   res[3] = element(m_input, idx, 3);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::constElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+   llvm::Value *xChannel, *yChannel, *zChannel, *wChannel;
+
+   xChannel = elementPointer(m_consts, idx, 0);
+   yChannel = elementPointer(m_consts, idx, 1);
+   zChannel = elementPointer(m_consts, idx, 2);
+   wChannel = elementPointer(m_consts, idx, 3);
+
+   res[0] = alignedArrayLoad(xChannel);
+   res[1] = alignedArrayLoad(yChannel);
+   res[2] = alignedArrayLoad(zChannel);
+   res[3] = alignedArrayLoad(wChannel);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::outputElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_output, idx, 0);
+   res[1] = element(m_output, idx, 1);
+   res[2] = element(m_output, idx, 2);
+   res[3] = element(m_output, idx, 3);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::tempElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_temps, idx, 0);
+   res[1] = element(m_temps, idx, 1);
+   res[2] = element(m_temps, idx, 2);
+   res[3] = element(m_temps, idx, 3);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::immediateElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_immediates, idx, 0);
+   res[1] = element(m_immediates, idx, 1);
+   res[2] = element(m_immediates, idx, 2);
+   res[3] = element(m_immediates, idx, 3);
+
+   return res;
+}
+
+llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index,
+                                         int channel) const
+{
+   std::vector<Value*> indices;
+   if (m_immediates == ptr)
+      indices.push_back(constantInt(0));
+   indices.push_back(index);
+   indices.push_back(constantInt(channel));
+
+   GetElementPtrInst *getElem = new GetElementPtrInst(ptr,
+                                                      indices.begin(),
+                                                      indices.end(),
+                                                      name("ptr"),
+                                                      m_block);
+   return getElem;
+}
+
+llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index,
+                                  int channel) const
+{
+   llvm::Value *res = elementPointer(ptr, index, channel);
+   LoadInst *load = new LoadInst(res, name("element"), false, m_block);
+   //load->setAlignment(8);
+   return load;
+}
+
+const char * StorageSoa::name(const char *prefix) const
+{
+   ++m_idx;
+   snprintf(m_name, 32, "%s%d", prefix, m_idx);
+   return m_name;
+}
+
+llvm::ConstantInt * StorageSoa::constantInt(int idx) const
+{
+   if (m_constInts.find(idx) != m_constInts.end()) {
+      return m_constInts[idx];
+   }
+   ConstantInt *constInt = ConstantInt::get(APInt(32,  idx));
+   m_constInts[idx] = constInt;
+   return constInt;
+}
+
+llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val)
+{
+   VectorType  *vectorType = VectorType::get(Type::FloatTy, 4);
+   PointerType *vectorPtr  = PointerType::get(vectorType, 0);
+
+   CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block);
+   LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block);
+   load->setAlignment(8);
+   return load;
+}
+
+llvm::Module * StorageSoa::currentModule() const
+{
+    if (!m_block || !m_block->getParent())
+       return 0;
+
+    return m_block->getParent()->getParent();
+}
+
+llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector<float> &vec)
+{
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   std::vector<Constant*> immValues;
+   ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0]));
+   ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1]));
+   ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2]));
+   ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3]));
+   immValues.push_back(constx);
+   immValues.push_back(consty);
+   immValues.push_back(constz);
+   immValues.push_back(constw);
+   Constant  *constVector = ConstantVector::get(vectorType, immValues);
+
+   return constVector;
+}
+
+std::vector<llvm::Value*> StorageSoa::load(Argument type, int idx, int swizzle,
+                                           llvm::Value *indIdx)
+{
+   std::vector<llvm::Value*> val(4);
+
+   //if we have an indirect index, always use that
+   //   if not use the integer offset to create one
+   llvm::Value *realIndex = 0;
+   if (indIdx)
+      realIndex = indIdx;
+   else
+      realIndex = constantInt(idx);
+   debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx);
+
+   switch(type) {
+   case Input:
+      val = inputElement(realIndex);
+      break;
+   case Output:
+      val = outputElement(realIndex);
+      break;
+   case Temp:
+      val = tempElement(realIndex);
+      break;
+   case Const:
+      val = constElement(realIndex);
+      break;
+   case Immediate:
+      val = immediateElement(realIndex);
+      break;
+   case Address:
+      debug_printf("Address not handled in the load phase!\n");
+      assert(0);
+      break;
+   }
+   if (!gallivm_is_swizzle(swizzle))
+      return val;
+
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = val[gallivm_x_swizzle(swizzle)];
+   res[1] = val[gallivm_y_swizzle(swizzle)];
+   res[2] = val[gallivm_z_swizzle(swizzle)];
+   res[3] = val[gallivm_w_swizzle(swizzle)];
+   return res;
+}
+
+void StorageSoa::store(Argument type, int idx, const std::vector<llvm::Value*> &val,
+                       int mask)
+{
+   llvm::Value *out = 0;
+   switch(type) {
+   case Output:
+      out = m_output;
+      break;
+   case Temp:
+      out = m_temps;
+      break;
+   case Input:
+      out = m_input;
+      break;
+   case Address: {
+      llvm::Value *addr = m_addresses[idx];
+      if (!addr) {
+         addAddress(idx);
+         addr = m_addresses[idx];
+         assert(addr);
+      }
+      new StoreInst(val[0], addr, false, m_block);
+      return;
+      break;
+   }
+   default:
+      debug_printf("Can't save output of this type: %d !\n", type);
+      assert(0);
+      break;
+   }
+   llvm::Value *realIndex = constantInt(idx);
+   if ((mask & TGSI_WRITEMASK_X)) {
+      llvm::Value *xChannel = elementPointer(out, realIndex, 0);
+      new StoreInst(val[0], xChannel, false, m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_Y)) {
+      llvm::Value *yChannel = elementPointer(out, realIndex, 1);
+      new StoreInst(val[1], yChannel, false, m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_Z)) {
+      llvm::Value *zChannel = elementPointer(out, realIndex, 2);
+      new StoreInst(val[2], zChannel, false, m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_W)) {
+      llvm::Value *wChannel = elementPointer(out, realIndex, 3);
+      new StoreInst(val[3], wChannel, false, m_block);
+   }
+}
+
+void StorageSoa::addAddress(int idx)
+{
+   GlobalVariable *val = new GlobalVariable(
+      /*Type=*/IntegerType::get(32),
+      /*isConstant=*/false,
+      /*Linkage=*/GlobalValue::ExternalLinkage,
+      /*Initializer=*/0, // has initializer, specified below
+      /*Name=*/name("address"),
+      currentModule());
+   val->setInitializer(Constant::getNullValue(IntegerType::get(32)));
+
+   debug_printf("adding to %d\n", idx);
+   m_addresses[idx] = val;
+}
diff --git a/src/gallium/auxiliary/llvm/storagesoa.h b/src/gallium/auxiliary/llvm/storagesoa.h
new file mode 100644
index 00000000000..6443351f270
--- /dev/null
+++ b/src/gallium/auxiliary/llvm/storagesoa.h
@@ -0,0 +1,111 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STORAGESOA_H
+#define STORAGESOA_H
+
+#include <vector>
+#include <list>
+#include <map>
+
+namespace llvm {
+   class BasicBlock;
+   class Constant;
+   class ConstantInt;
+   class GlobalVariable;
+   class LoadInst;
+   class Value;
+   class VectorType;
+   class Module;
+}
+
+class StorageSoa
+{
+public:
+   enum Argument {
+      Input,
+      Output,
+      Temp,
+      Const,
+      Immediate,
+      Address
+   };
+public:
+   StorageSoa(llvm::BasicBlock *block,
+              llvm::Value *input,
+              llvm::Value *output,
+              llvm::Value *consts,
+              llvm::Value *temps);
+
+
+   std::vector<llvm::Value*> load(Argument type, int idx, int swizzle, 
+                                  llvm::Value *indIdx =0);
+   void store(Argument type, int idx, const std::vector<llvm::Value*> &val,
+              int mask);
+
+   void addImmediate(float *vec);
+   void declareImmediates();
+
+   void addAddress(int idx);
+
+   llvm::Value  * addrElement(int idx) const;
+
+   llvm::ConstantInt *constantInt(int) const;
+private:
+   llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx,
+                               int channel) const;
+   llvm::Value *element(llvm::Value *ptr, llvm::Value *idx,
+                        int channel) const;
+   const char *name(const char *prefix) const;
+   llvm::Value  *alignedArrayLoad(llvm::Value *val);
+   llvm::Module *currentModule() const;
+   llvm::Constant  *createConstGlobalVector(const std::vector<float> &vec);
+
+   std::vector<llvm::Value*> inputElement(llvm::Value *indIdx);
+   std::vector<llvm::Value*> constElement(llvm::Value *indIdx);
+   std::vector<llvm::Value*> outputElement(llvm::Value *indIdx);
+   std::vector<llvm::Value*> tempElement(llvm::Value *indIdx);
+   std::vector<llvm::Value*> immediateElement(llvm::Value *indIdx);
+private:
+   llvm::BasicBlock *m_block;
+
+   llvm::Value *m_input;
+   llvm::Value *m_output;
+   llvm::Value *m_consts;
+   llvm::Value *m_temps;
+   llvm::GlobalVariable *m_immediates;
+
+   std::map<int, llvm::Value*> m_addresses;
+
+   std::vector<std::vector<float> > m_immediatesToFlush;
+
+   mutable std::map<int, llvm::ConstantInt*> m_constInts;
+   mutable char        m_name[32];
+   mutable int         m_idx;
+};
+
+#endif
diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/gallium/auxiliary/llvm/tgsitollvm.cpp
index bc4df610717..2cb4acce32f 100644
--- a/src/mesa/pipe/llvm/tgsitollvm.cpp
+++ b/src/gallium/auxiliary/llvm/tgsitollvm.cpp
@@ -10,11 +10,11 @@
 
 #include "pipe/p_shader_tokens.h"
 
-#include "pipe/tgsi/util/tgsi_parse.h"
-#include "pipe/tgsi/exec/tgsi_exec.h"
-#include "pipe/tgsi/util/tgsi_util.h"
-#include "pipe/tgsi/util/tgsi_build.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
+#include "tgsi/util/tgsi_parse.h"
+#include "tgsi/exec/tgsi_exec.h"
+#include "tgsi/util/tgsi_util.h"
+#include "tgsi/util/tgsi_build.h"
+#include "tgsi/util/tgsi_dump.h"
 
 
 #include <llvm/Module.h>
@@ -44,7 +44,38 @@
 #include <iostream>
 
 using namespace llvm;
-#include "llvm_base_shader.cpp"
+
+static inline FunctionType *vertexShaderFunctionType()
+{
+   //Function takes three arguments,
+   // the calling code has to make sure the types it will
+   // pass are castable to the following:
+   // [4 x <4 x float>] inputs,
+   // [4 x <4 x float>] output,
+   // [4 x [4 x float]] consts,
+   // [4 x <4 x float>] temps
+
+   std::vector<const Type*> funcArgs;
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   ArrayType *vectorArray = ArrayType::get(vectorType, 4);
+   PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0);
+
+   ArrayType   *floatArray     = ArrayType::get(Type::FloatTy, 4);
+   ArrayType   *constsArray    = ArrayType::get(floatArray, 4);
+   PointerType *constsArrayPtr = PointerType::get(constsArray, 0);
+
+   funcArgs.push_back(vectorArrayPtr);//inputs
+   funcArgs.push_back(vectorArrayPtr);//output
+   funcArgs.push_back(constsArrayPtr);//consts
+   funcArgs.push_back(vectorArrayPtr);//temps
+
+   FunctionType *functionType = FunctionType::get(
+      /*Result=*/Type::VoidTy,
+      /*Params=*/funcArgs,
+      /*isVarArg=*/false);
+
+   return functionType;
+}
 
 static inline void
 add_interpolator(struct gallivm_ir *ir,
@@ -117,10 +148,14 @@ translate_declaration(struct gallivm_ir *prog,
 static void
 translate_declarationir(struct gallivm_ir *,
                       llvm::Module *,
-                      StorageSoa *,
-                      struct tgsi_full_declaration *,
+                      StorageSoa *storage,
+                      struct tgsi_full_declaration *decl,
                       struct tgsi_full_declaration *)
 {
+   if (decl->Declaration.File == TGSI_FILE_ADDRESS) {
+      int idx = decl->u.DeclarationRange.First;
+      storage->addAddress(idx);
+   }
 }
 
 static void
@@ -178,9 +213,8 @@ swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src,
               Storage *storage)
 {
    int swizzle = swizzleInt(src);
-   const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 +
-                          TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W;
-   if (swizzle != NO_SWIZZLE) {
+
+   if (gallivm_is_swizzle(swizzle)) {
       /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/
       val = storage->shuffleVector(val, swizzle);
    }
@@ -673,18 +707,22 @@ translate_instructionir(llvm::Module *module,
 
       if (src->SrcRegister.Indirect) {
          indIdx = storage->addrElement(src->SrcRegisterInd.Index);
-         indIdx = storage->extractIndex(indIdx);
       }
       if (src->SrcRegister.File == TGSI_FILE_CONSTANT) {
-         val = storage->constElement(src->SrcRegister.Index, swizzle, indIdx);
+         val = storage->load(StorageSoa::Const,
+                             src->SrcRegister.Index, swizzle, indIdx);
       } else if (src->SrcRegister.File == TGSI_FILE_INPUT) {
-         val = storage->inputElement(src->SrcRegister.Index, swizzle, indIdx);
+         val = storage->load(StorageSoa::Input,
+                             src->SrcRegister.Index, swizzle, indIdx);
       } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) {
-         val = storage->tempElement(src->SrcRegister.Index, swizzle);
+         val = storage->load(StorageSoa::Temp,
+                             src->SrcRegister.Index, swizzle, indIdx);
       } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) {
-         val = storage->outputElement(src->SrcRegister.Index, swizzle, indIdx);
+         val = storage->load(StorageSoa::Output,
+                             src->SrcRegister.Index, swizzle, indIdx);
       } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) {
-         val = storage->immediateElement(src->SrcRegister.Index, swizzle);
+         val = storage->load(StorageSoa::Immediate,
+                             src->SrcRegister.Index, swizzle, indIdx);
       } else {
          fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File);
          return;
@@ -696,6 +734,7 @@ translate_instructionir(llvm::Module *module,
    std::vector<llvm::Value*> out(4);
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_ARL: {
+      out = instr->arl(inputs[0]);
    }
       break;
    case TGSI_OPCODE_MOV: {
@@ -745,6 +784,7 @@ translate_instructionir(llvm::Module *module,
    }
       break;
    case TGSI_OPCODE_MAD: {
+      out = instr->madd(inputs[0], inputs[1], inputs[2]);
    }
       break;
    case TGSI_OPCODE_SUB: {
@@ -1028,11 +1068,14 @@ translate_instructionir(llvm::Module *module,
       struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
 
       if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
-         storage->storeOutput(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+         storage->store(StorageSoa::Output,
+                        dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
       } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) {
-         storage->storeTemp(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+         storage->store(StorageSoa::Temp,
+                        dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
       } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) {
-         storage->storeAddress(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+         storage->store(StorageSoa::Address,
+                        dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
       } else {
          fprintf(stderr, "ERROR: unsupported LLVM destination!");
          assert(!"wrong destination");
@@ -1043,7 +1086,7 @@ translate_instructionir(llvm::Module *module,
 llvm::Module *
 tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens)
 {
-   llvm::Module *mod = createBaseShader();
+   llvm::Module *mod = new Module("shader");
    struct tgsi_parse_context parse;
    struct tgsi_full_instruction fi;
    struct tgsi_full_declaration fd;
@@ -1107,29 +1150,32 @@ tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens)
 llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir,
                               const struct tgsi_token *tokens)
 {
-   llvm::Module *mod = createBaseShader();
+   llvm::Module *mod = new Module("shader");
    struct tgsi_parse_context parse;
    struct tgsi_full_instruction fi;
    struct tgsi_full_declaration fd;
    unsigned instno = 0;
-   Function* shader = mod->getFunction("execute_shader");
    std::ostringstream stream;
    if (ir->type == GALLIVM_VS) {
       stream << "vs_shader";
    } else {
       stream << "fs_shader";
    }
-   stream << ir->id;
+   //stream << ir->id;
    std::string func_name = stream.str();
-   shader->setName(func_name.c_str());
+   Function *shader = llvm::cast<Function>(mod->getOrInsertFunction(
+                                              func_name.c_str(),
+                                              vertexShaderFunctionType()));
 
    Function::arg_iterator args = shader->arg_begin();
    Value *input = args++;
-   input->setName("input");
+   input->setName("inputs");
    Value *output = args++;
-   output->setName("output");
+   output->setName("outputs");
    Value *consts = args++;
    consts->setName("consts");
+   Value *temps = args++;
+   temps->setName("temps");
 
    BasicBlock *label_entry = new BasicBlock("entry", shader, 0);
 
@@ -1138,7 +1184,7 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir,
    fi = tgsi_default_full_instruction();
    fd = tgsi_default_full_declaration();
 
-   StorageSoa storage(label_entry, input, output, consts);
+   StorageSoa storage(label_entry, input, output, consts, temps);
    InstructionsSoa instr(mod, shader, label_entry, &storage);
 
    while(!tgsi_parse_end_of_tokens(&parse)) {
@@ -1157,6 +1203,7 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir,
          break;
 
       case TGSI_TOKEN_TYPE_INSTRUCTION:
+         storage.declareImmediates();
          translate_instructionir(mod, &storage, &instr,
                                  &parse.FullToken.FullInstruction,
                                  &fi, instno);
diff --git a/src/mesa/pipe/llvm/tgsitollvm.h b/src/gallium/auxiliary/llvm/tgsitollvm.h
index 7ada04d6299..7ada04d6299 100644
--- a/src/mesa/pipe/llvm/tgsitollvm.h
+++ b/src/gallium/auxiliary/llvm/tgsitollvm.h
diff --git a/src/mesa/pipe/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile
index 75764a9a188..588629e8701 100644
--- a/src/mesa/pipe/pipebuffer/Makefile
+++ b/src/gallium/auxiliary/pipebuffer/Makefile
@@ -17,7 +17,7 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-include ../Makefile.template
+include ../../Makefile.template
 
 symlinks:
 
diff --git a/src/mesa/pipe/pipebuffer/linked_list.h b/src/gallium/auxiliary/pipebuffer/linked_list.h
index e99817fb130..e99817fb130 100644
--- a/src/mesa/pipe/pipebuffer/linked_list.h
+++ b/src/gallium/auxiliary/pipebuffer/linked_list.h
diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
index 97beb5f72a9..97beb5f72a9 100644
--- a/src/mesa/pipe/pipebuffer/pb_buffer.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
index f4fc3f6d714..f4fc3f6d714 100644
--- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
index c40b9c75e18..c40b9c75e18 100644
--- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
index c1b7759874c..9e8244f909b 100644
--- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
@@ -110,7 +110,7 @@ pb_malloc_buffer_create(size_t size,
    buf = CALLOC_STRUCT(malloc_buffer);
    if(!buf)
       return NULL;
-   
+
    buf->base.base.refcount = 1;
    buf->base.base.alignment = desc->alignment;
    buf->base.base.usage = desc->usage;
@@ -119,7 +119,7 @@ pb_malloc_buffer_create(size_t size,
 
    buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment);
    if(!buf->data) {
-      FREE(buf);
+      align_free(buf);
       return NULL;
    }
 
diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
index 1ddf784c978..1ddf784c978 100644
--- a/src/mesa/pipe/pipebuffer/pb_bufmgr.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c
index c535d3276c9..c535d3276c9 100644
--- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c
diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c
index 8b1b51c0e28..8b1b51c0e28 100644
--- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c
diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c
index 04477a865a5..04477a865a5 100644
--- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c
diff --git a/src/mesa/pipe/pipebuffer/pb_winsys.c b/src/gallium/auxiliary/pipebuffer/pb_winsys.c
index 978944091fd..978944091fd 100644
--- a/src/mesa/pipe/pipebuffer/pb_winsys.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_winsys.c
diff --git a/src/mesa/pipe/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile
index 12a8bd0409e..12a8bd0409e 100644
--- a/src/mesa/pipe/tgsi/Makefile
+++ b/src/gallium/auxiliary/tgsi/Makefile
diff --git a/src/mesa/pipe/tgsi/exec/Makefile b/src/gallium/auxiliary/tgsi/exec/Makefile
index eb8b14e0e89..eb8b14e0e89 100644
--- a/src/mesa/pipe/tgsi/exec/Makefile
+++ b/src/gallium/auxiliary/tgsi/exec/Makefile
diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
index 336ae1c8b6d..d7b18dc9c59 100644
--- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
@@ -54,8 +54,8 @@
 #include "pipe/p_state.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
-#include "pipe/tgsi/util/tgsi_util.h"
+#include "tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_util.h"
 #include "tgsi_exec.h"
 
 #define TILE_TOP_LEFT     0
@@ -120,18 +120,41 @@
 
 
 
-static void
-tgsi_exec_prepare( struct tgsi_exec_machine *mach )
+/**
+ * Initialize machine state by expanding tokens to full instructions,
+ * allocating temporary storage, setting up constants, etc.
+ * After this, we can call tgsi_exec_machine_run() many times.
+ */
+void 
+tgsi_exec_machine_bind_shader(
+   struct tgsi_exec_machine *mach,
+   const struct tgsi_token *tokens,
+   uint numSamplers,
+   struct tgsi_sampler *samplers)
 {
-   struct tgsi_exec_labels *labels = &mach->Labels;
+   uint k;
    struct tgsi_parse_context parse;
+   struct tgsi_exec_labels *labels = &mach->Labels;
    struct tgsi_full_instruction *instructions;
    struct tgsi_full_declaration *declarations;
    uint maxInstructions = 10, numInstructions = 0;
    uint maxDeclarations = 10, numDeclarations = 0;
-   uint k;
    uint instno = 0;
 
+#if 0
+   tgsi_dump(tokens, 0);
+#endif
+
+   mach->Tokens = tokens;
+   mach->Samplers = samplers;
+
+   k = tgsi_parse_init (&parse, mach->Tokens);
+   if (k != TGSI_PARSE_OK) {
+      debug_printf( "Problem parsing!\n" );
+      return;
+   }
+
+   mach->Processor = parse.FullHeader.Processor.Processor;
    mach->ImmLimit = 0;
    labels->count = 0;
 
@@ -141,11 +164,6 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach )
    instructions = (struct tgsi_full_instruction *)
       MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) );
 
-   k = tgsi_parse_init( &parse, mach->Tokens );
-   if (k != TGSI_PARSE_OK) {
-      debug_printf("Problem parsing!\n");
-      return;
-   }
 
    while( !tgsi_parse_end_of_tokens( &parse ) ) {
       uint pointer = parse.Position;
@@ -158,13 +176,13 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach )
          if (numDeclarations == maxDeclarations) {
             declarations = REALLOC(declarations,
                                    maxDeclarations
-                                   * sizeof(struct tgsi_full_instruction),
+                                   * sizeof(struct tgsi_full_declaration),
                                    (maxDeclarations + 10)
-                                   * sizeof(struct tgsi_full_instruction));
+                                   * sizeof(struct tgsi_full_declaration));
             maxDeclarations += 10;
          }
          memcpy(declarations + numDeclarations,
-                &parse.FullToken.FullInstruction,
+                &parse.FullToken.FullDeclaration,
                 sizeof(declarations[0]));
          numDeclarations++;
          break;
@@ -176,7 +194,8 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach )
             assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES );
 
             for( i = 0; i < size; i++ ) {
-               mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+               mach->Imms[mach->ImmLimit + i / 4][i % 4] = 
+		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
             }
             mach->ImmLimit += size / 4;
          }
@@ -224,37 +243,11 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach )
 }
 
 
-/**
- * Initialize machine state by expanding tokens to full instructions,
- * allocating temporary storage, setting up constants, etc.
- * After this, we can call tgsi_exec_machine_run() many times.
- */
 void
 tgsi_exec_machine_init(
-   struct tgsi_exec_machine *mach,
-   const struct tgsi_token *tokens,
-   uint numSamplers,
-   struct tgsi_sampler *samplers)
+   struct tgsi_exec_machine *mach )
 {
-   uint i, k;
-   struct tgsi_parse_context parse;
-
-#if 0
-   tgsi_dump(tokens, 0);
-#endif
-
-   mach->Tokens = tokens;
-
-   mach->Samplers = samplers;
-
-   k = tgsi_parse_init (&parse, mach->Tokens);
-   if (k != TGSI_PARSE_OK) {
-      debug_printf( "Problem parsing!\n" );
-      return;
-   }
-
-   mach->Processor = parse.FullHeader.Processor.Processor;
-   tgsi_parse_free (&parse);
+   uint i;
 
    mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps);
    mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS];
@@ -270,8 +263,6 @@ tgsi_exec_machine_init(
       mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f;
       mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f;
    }
-
-   tgsi_exec_prepare( mach );
 }
 
 
@@ -1346,9 +1337,12 @@ exec_tex(struct tgsi_exec_machine *mach,
 }
 
 
-
+/**
+ * Evaluate a constant-valued coefficient at the position of the
+ * current quad.
+ */
 static void
-constant_interpolation(
+eval_constant_coef(
    struct tgsi_exec_machine *mach,
    unsigned attrib,
    unsigned chan )
@@ -1360,8 +1354,12 @@ constant_interpolation(
    }
 }
 
+/**
+ * Evaluate a linear-valued coefficient at the position of the
+ * current quad.
+ */
 static void
-linear_interpolation(
+eval_linear_coef(
    struct tgsi_exec_machine *mach,
    unsigned attrib,
    unsigned chan )
@@ -1377,8 +1375,12 @@ linear_interpolation(
    mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady;
 }
 
+/**
+ * Evaluate a perspective-valued coefficient at the position of the
+ * current quad.
+ */
 static void
-perspective_interpolation(
+eval_perspective_coef(
    struct tgsi_exec_machine *mach,
    unsigned attrib,
    unsigned chan )
@@ -1397,7 +1399,7 @@ perspective_interpolation(
 }
 
 
-typedef void (* interpolation_func)(
+typedef void (* eval_coef_func)(
    struct tgsi_exec_machine *mach,
    unsigned attrib,
    unsigned chan );
@@ -1410,7 +1412,7 @@ exec_declaration(
    if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) {
       if( decl->Declaration.File == TGSI_FILE_INPUT ) {
          unsigned first, last, mask;
-         interpolation_func interp;
+         eval_coef_func eval;
 
          assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE );
 
@@ -1420,15 +1422,15 @@ exec_declaration(
 
          switch( decl->Interpolation.Interpolate ) {
          case TGSI_INTERPOLATE_CONSTANT:
-            interp = constant_interpolation;
+            eval = eval_constant_coef;
             break;
 
          case TGSI_INTERPOLATE_LINEAR:
-            interp = linear_interpolation;
+            eval = eval_linear_coef;
             break;
 
          case TGSI_INTERPOLATE_PERSPECTIVE:
-            interp = perspective_interpolation;
+            eval = eval_perspective_coef;
             break;
 
          default:
@@ -1440,7 +1442,7 @@ exec_declaration(
 
             for( i = first; i <= last; i++ ) {
                for( j = 0; j < NUM_CHANNELS; j++ ) {
-                  interp( mach, i, j );
+                  eval( mach, i, j );
                }
             }
          }
@@ -1450,7 +1452,7 @@ exec_declaration(
             for( j = 0; j < NUM_CHANNELS; j++ ) {
                if( mask & (1 << j) ) {
                   for( i = first; i <= last; i++ ) {
-                     interp( mach, i, j );
+                     eval( mach, i, j );
                   }
                }
             }
diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h
index 1fb66ee960f..45c49dd007c 100644
--- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h
@@ -215,12 +215,16 @@ struct tgsi_exec_machine
    struct tgsi_exec_labels Labels;
 };
 
-
 void
 tgsi_exec_machine_init(
+   struct tgsi_exec_machine *mach );
+
+
+void 
+tgsi_exec_machine_bind_shader(
    struct tgsi_exec_machine *mach,
    const struct tgsi_token *tokens,
-   unsigned numSamplers,
+   uint numSamplers,
    struct tgsi_sampler *samplers);
 
 uint
diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
index 40bacf85526..62c6a69c63f 100755
--- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
@@ -27,8 +27,8 @@
 
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
-#include "pipe/tgsi/util/tgsi_util.h"
+#include "tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_util.h"
 #include "tgsi_exec.h"
 #include "tgsi_sse2.h"
 
@@ -44,6 +44,9 @@ static void
 _print_reg(
    struct x86_reg reg )
 {
+   if (reg.mod != mod_REG) 
+      debug_printf( "[" );
+      
    switch( reg.file ) {
    case file_REG32:
       switch( reg.idx ) {
@@ -83,6 +86,13 @@ _print_reg(
       assert( 0 );
       break;
    }
+
+   if (reg.mod == mod_DISP8 ||
+       reg.mod == mod_DISP32)
+      debug_printf("+%d", reg.disp);
+
+   if (reg.mod != mod_REG) 
+      debug_printf( "]" );
 }
 
 static void
@@ -318,8 +328,11 @@ emit_call(
    struct x86_function  *func,
    void                 (* addr)() )
 {
+   struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
+
    DUMP_I( "CALL", addr );
-   x86_call( func, addr );
+   x86_mov_reg_imm( func, ecx, (unsigned long) addr );
+   x86_call( func, ecx );
 }
 
 static void
@@ -1426,7 +1439,7 @@ emit_cmp(
    }
 }
 
-static void
+static int
 emit_instruction(
    struct x86_function *func,
    struct tgsi_full_instruction *inst )
@@ -1537,11 +1550,11 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_EXP:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_LOG:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_MUL:
@@ -1696,24 +1709,24 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_CND:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_CND0:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_DOT2ADD:
    /* TGSI_OPCODE_DP2A */
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_INDEX:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_NEGATE:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_FRAC:
@@ -1726,7 +1739,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_CLAMP:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_FLOOR:
@@ -1739,7 +1752,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_ROUND:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_EXPBASE2:
@@ -1814,7 +1827,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_MULTIPLYMATRIX:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ABS:
@@ -1827,7 +1840,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_RCC:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_DPH:
@@ -1858,11 +1871,11 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_DDX:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_DDY:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_KIL:
@@ -1870,35 +1883,35 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_PK2H:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_PK2US:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_PK4B:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_PK4UB:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_RFL:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SEQ:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SFL:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SGT:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SIN:
@@ -1910,66 +1923,73 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_SLE:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SNE:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_STR:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_TEX:
-      emit_tempf(
-         func,
-         0,
-         TGSI_EXEC_TEMP_ONE_I,
-         TGSI_EXEC_TEMP_ONE_C );
-      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( func, *inst, 0, 0, chan_index );
+      if (0) {
+	 /* Disable dummy texture code: 
+	  */
+	 emit_tempf(
+	    func,
+	    0,
+	    TGSI_EXEC_TEMP_ONE_I,
+	    TGSI_EXEC_TEMP_ONE_C );
+	 FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+	    STORE( func, *inst, 0, 0, chan_index );
+	 }
+      }
+      else {
+	 return 0;
       }
       break;
 
    case TGSI_OPCODE_TXD:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_UP2H:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_UP2US:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_UP4B:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_UP4UB:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_X2D:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ARA:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ARR:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_BRA:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_CAL:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_RET:
@@ -1982,7 +2002,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_SSG:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_CMP:
@@ -2011,132 +2031,134 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXB:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_NRM:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_DIV:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_DP2:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_TXL:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_BRK:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_IF:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_LOOP:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_REP:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ELSE:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ENDIF:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ENDLOOP:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ENDREP:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_PUSHA:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_POPA:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_CEIL:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_I2F:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_NOT:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_TRUNC:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SHL:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SHR:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_AND:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_OR:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_MOD:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_XOR:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_SAD:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_TXF:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_TXQ:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_CONT:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_EMIT:
-      assert( 0 );
+      return 0;
       break;
 
    case TGSI_OPCODE_ENDPRIM:
-      assert( 0 );
+      return 0;
       break;
 
    default:
-      assert( 0 );
+      return 0;
    }
+   
+   return 1;
 }
 
 static void
@@ -2154,14 +2176,6 @@ emit_declaration(
       last = decl->u.DeclarationRange.Last;
       mask = decl->Declaration.UsageMask;
 
-      /* Do not touch WPOS.xy */
-      if( first == 0 ) {
-         mask &= ~TGSI_WRITEMASK_XY;
-         if( mask == TGSI_WRITEMASK_NONE ) {
-            first++;
-         }
-      }
-
       for( i = first; i <= last; i++ ) {
          for( j = 0; j < NUM_CHANNELS; j++ ) {
             if( mask & (1 << j) ) {
@@ -2172,9 +2186,9 @@ emit_declaration(
                   break;
 
                case TGSI_INTERPOLATE_LINEAR:
-                  emit_inputf( func, 0, 0, TGSI_SWIZZLE_X );
+                  emit_tempf( func, 0, 0, TGSI_SWIZZLE_X );
                   emit_coef_dadx( func, 1, i, j );
-                  emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y );
+                  emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y );
                   emit_coef_dady( func, 3, i, j );
                   emit_mul( func, 0, 1 );    /* x * dadx */
                   emit_coef_a0( func, 4, i, j );
@@ -2185,12 +2199,12 @@ emit_declaration(
                   break;
 
                case TGSI_INTERPOLATE_PERSPECTIVE:
-                  emit_inputf( func, 0, 0, TGSI_SWIZZLE_X );
+                  emit_tempf( func, 0, 0, TGSI_SWIZZLE_X );
                   emit_coef_dadx( func, 1, i, j );
-                  emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y );
+                  emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y );
                   emit_coef_dady( func, 3, i, j );
                   emit_mul( func, 0, 1 );    /* x * dadx */
-                  emit_inputf( func, 4, 0, TGSI_SWIZZLE_W );
+                  emit_tempf( func, 4, 0, TGSI_SWIZZLE_W );
                   emit_coef_a0( func, 5, i, j );
                   emit_rcp( func, 4, 4 );    /* 1.0 / w */
                   emit_mul( func, 2, 3 );    /* y * dady */
@@ -2202,6 +2216,7 @@ emit_declaration(
 
                default:
                   assert( 0 );
+		  break;
                }
             }
          }
@@ -2215,6 +2230,7 @@ tgsi_emit_sse2(
    struct x86_function *func )
 {
    struct tgsi_parse_context parse;
+   unsigned ok = 1;
 
    DUMP_START();
 
@@ -2239,7 +2255,7 @@ tgsi_emit_sse2(
 
    tgsi_parse_init( &parse, tokens );
 
-   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+   while( !tgsi_parse_end_of_tokens( &parse ) && ok ) {
       tgsi_parse_token( &parse );
 
       switch( parse.FullToken.Token.Type ) {
@@ -2247,17 +2263,26 @@ tgsi_emit_sse2(
          break;
 
       case TGSI_TOKEN_TYPE_INSTRUCTION:
-         emit_instruction(
-            func,
-            &parse.FullToken.FullInstruction );
+         ok = emit_instruction(
+	    func,
+	    &parse.FullToken.FullInstruction );
+
+	 if (!ok) {
+	    debug_printf("failed to translate tgsi opcode %d\n", 
+			 parse.FullToken.FullInstruction.Instruction.Opcode );
+	 }
          break;
 
       case TGSI_TOKEN_TYPE_IMMEDIATE:
          /* XXX implement this */
-         return 0;
+	 ok = 0;
+	 debug_printf("failed to emit immediate value\n");
+	 break;
 
       default:
          assert( 0 );
+	 ok = 0;
+	 break;
       }
    }
 
@@ -2265,7 +2290,7 @@ tgsi_emit_sse2(
 
    DUMP_END();
 
-   return 1;
+   return ok;
 }
 
 /**
@@ -2283,6 +2308,7 @@ tgsi_emit_sse2_fs(
 {
    struct tgsi_parse_context parse;
    boolean instruction_phase = FALSE;
+   unsigned ok = 1;
 
    DUMP_START();
 
@@ -2308,7 +2334,7 @@ tgsi_emit_sse2_fs(
 
    tgsi_parse_init( &parse, tokens );
 
-   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+   while( !tgsi_parse_end_of_tokens( &parse ) && ok ) {
       tgsi_parse_token( &parse );
 
       switch( parse.FullToken.Token.Type ) {
@@ -2327,17 +2353,18 @@ tgsi_emit_sse2_fs(
                get_output_base(),
                get_argument( 1 ) );
          }
-         emit_instruction(
+         ok = emit_instruction(
             func,
             &parse.FullToken.FullInstruction );
          break;
 
       case TGSI_TOKEN_TYPE_IMMEDIATE:
          /* XXX implement this */
-         assert(0);
+	 ok = 0;
          break;
 
       default:
+	 ok = 0;
          assert( 0 );
       }
    }
@@ -2346,7 +2373,7 @@ tgsi_emit_sse2_fs(
 
    DUMP_END();
 
-   return 1;
+   return ok;
 }
 
 #endif /* i386 */
diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h
index 9bee3717665..9bee3717665 100755
--- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h
diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c
index a00ff1c2a5f..a00ff1c2a5f 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_build.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.c
diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h
index 116c78abf34..116c78abf34 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_build.h
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.h
diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c
index b5c54847e0b..b5c54847e0b 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c
diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h
index 1adc9db2519..1adc9db2519 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_dump.h
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h
diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c
index bf6b89ce564..bf6b89ce564 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_parse.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c
diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h
index 9372da8d5d1..9372da8d5d1 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_parse.h
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h
diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_transform.c b/src/gallium/auxiliary/tgsi/util/tgsi_transform.c
new file mode 100644
index 00000000000..357f77b05a6
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_transform.c
@@ -0,0 +1,199 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+/**
+ * TGSI program transformation utility.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "tgsi_transform.h"
+
+
+
+static void
+emit_instruction(struct tgsi_transform_context *ctx,
+                 const struct tgsi_full_instruction *inst)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_instruction(inst,
+                                     ctx->tokens_out + ti,
+                                     ctx->header,
+                                     ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
+
+static void
+emit_declaration(struct tgsi_transform_context *ctx,
+                 const struct tgsi_full_declaration *decl)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_declaration(decl,
+                                     ctx->tokens_out + ti,
+                                     ctx->header,
+                                     ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
+
+static void
+emit_immediate(struct tgsi_transform_context *ctx,
+               const struct tgsi_full_immediate *imm)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_immediate(imm,
+                                   ctx->tokens_out + ti,
+                                   ctx->header,
+                                   ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
+
+
+/**
+ * Apply user-defined transformations to the input shader to produce
+ * the output shader.
+ * For example, a register search-and-replace operation could be applied
+ * by defining a transform_instruction() callback that examined and changed
+ * the instruction src/dest regs.
+ *
+ * \return number of tokens emitted
+ */
+int
+tgsi_transform_shader(const struct tgsi_token *tokens_in,
+                      struct tgsi_token *tokens_out,
+                      uint max_tokens_out,
+                      struct tgsi_transform_context *ctx)
+{
+   uint procType;
+
+   /* input shader */
+   struct tgsi_parse_context parse;
+
+   /* output shader */
+   struct tgsi_processor *processor;
+
+
+   /**
+    ** callback context init
+    **/
+   ctx->emit_instruction = emit_instruction;
+   ctx->emit_declaration = emit_declaration;
+   ctx->emit_immediate = emit_immediate;
+   ctx->tokens_out = tokens_out;
+   ctx->max_tokens_out = max_tokens_out;
+
+
+   /**
+    ** Setup to begin parsing input shader
+    **/
+   if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) {
+      debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n");
+      return -1;
+   }
+   procType = parse.FullHeader.Processor.Processor;
+   assert(procType == TGSI_PROCESSOR_FRAGMENT ||
+          procType == TGSI_PROCESSOR_VERTEX ||
+          procType == TGSI_PROCESSOR_GEOMETRY);
+
+
+   /**
+    **  Setup output shader
+    **/
+   *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version();
+
+   ctx->header = (struct tgsi_header *) (tokens_out + 1);
+   *ctx->header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) (tokens_out + 2);
+   *processor = tgsi_build_processor( procType, ctx->header );
+
+   ctx->ti = 3;
+
+
+   /**
+    ** Loop over incoming program tokens/instructions
+    */
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         {
+            struct tgsi_full_instruction *fullinst
+               = &parse.FullToken.FullInstruction;
+
+            if (ctx->transform_instruction)
+               ctx->transform_instruction(ctx, fullinst);
+            else
+               ctx->emit_instruction(ctx, fullinst);
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         {
+            struct tgsi_full_declaration *fulldecl
+               = &parse.FullToken.FullDeclaration;
+
+            if (ctx->transform_declaration)
+               ctx->transform_declaration(ctx, fulldecl);
+            else
+               ctx->emit_declaration(ctx, fulldecl);
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         {
+            struct tgsi_full_immediate *fullimm
+               = &parse.FullToken.FullImmediate;
+
+            if (ctx->transform_immediate)
+               ctx->transform_immediate(ctx, fullimm);
+            else
+               ctx->emit_immediate(ctx, fullimm);
+         }
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+
+   if (ctx->epilog) {
+      ctx->epilog(ctx);
+   }
+
+   tgsi_parse_free (&parse);
+
+   return ctx->ti;
+}
diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_transform.h b/src/gallium/auxiliary/tgsi/util/tgsi_transform.h
new file mode 100644
index 00000000000..fcf85d603be
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_transform.h
@@ -0,0 +1,93 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_TRANSFORM_H
+#define TGSI_TRANSFORM_H
+
+
+#include "pipe/p_util.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_build.h"
+
+
+
+/**
+ * Subclass this to add caller-specific data
+ */
+struct tgsi_transform_context
+{
+/**** PUBLIC ***/
+
+   /**
+    * User-defined callbacks invoked per instruction.
+    */
+   void (*transform_instruction)(struct tgsi_transform_context *ctx,
+                                 struct tgsi_full_instruction *inst);
+
+   void (*transform_declaration)(struct tgsi_transform_context *ctx,
+                                 struct tgsi_full_declaration *decl);
+
+   void (*transform_immediate)(struct tgsi_transform_context *ctx,
+                               struct tgsi_full_immediate *imm);
+
+   /**
+    * Called at end of input program to allow caller to append extra
+    * instructions.  Return number of tokens emitted.
+    */
+   void (*epilog)(struct tgsi_transform_context *ctx);
+
+
+/*** PRIVATE ***/
+
+   /**
+    * These are setup by tgsi_transform_shader() and cannot be overridden.
+    * Meant to be called from in the above user callback functions.
+    */
+   void (*emit_instruction)(struct tgsi_transform_context *ctx,
+                            const struct tgsi_full_instruction *inst);
+   void (*emit_declaration)(struct tgsi_transform_context *ctx,
+                            const struct tgsi_full_declaration *decl);
+   void (*emit_immediate)(struct tgsi_transform_context *ctx,
+                          const struct tgsi_full_immediate *imm);
+
+   struct tgsi_header *header;
+   uint max_tokens_out;
+   struct tgsi_token *tokens_out;
+   uint ti;
+};
+
+
+
+extern int
+tgsi_transform_shader(const struct tgsi_token *tokens_in,
+                      struct tgsi_token *tokens_out,
+                      uint max_tokens_out,
+                      struct tgsi_transform_context *ctx);
+
+
+#endif /* TGSI_TRANSFORM_H */
diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.c b/src/gallium/auxiliary/tgsi/util/tgsi_util.c
index 4cdd89182ad..4cdd89182ad 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_util.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.c
diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.h b/src/gallium/auxiliary/tgsi/util/tgsi_util.h
index ef14446f0e4..ef14446f0e4 100644
--- a/src/mesa/pipe/tgsi/util/tgsi_util.h
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.h
diff --git a/src/mesa/pipe/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c
index b9607a6ba7f..b9607a6ba7f 100644
--- a/src/mesa/pipe/util/p_debug.c
+++ b/src/gallium/auxiliary/util/p_debug.c
diff --git a/src/mesa/pipe/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c
index 3f795a38989..3f795a38989 100644
--- a/src/mesa/pipe/util/p_tile.c
+++ b/src/gallium/auxiliary/util/p_tile.c
diff --git a/src/mesa/pipe/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h
index cd8124bf11f..cd8124bf11f 100644
--- a/src/mesa/pipe/util/p_tile.h
+++ b/src/gallium/auxiliary/util/p_tile.h
diff --git a/src/mesa/pipe/util/p_util.c b/src/gallium/auxiliary/util/p_util.c
index c4882b77d2d..2a92f8e408b 100644
--- a/src/mesa/pipe/util/p_util.c
+++ b/src/gallium/auxiliary/util/p_util.c
@@ -47,9 +47,9 @@ pipe_copy_rect(ubyte * dst,
                unsigned width,
                unsigned height,
                const ubyte * src,
-               unsigned src_pitch,
+               int src_pitch,
                unsigned src_x, 
-               unsigned src_y)
+               int src_y)
 {
    unsigned i;
 
diff --git a/src/gallium/drivers/Makefile b/src/gallium/drivers/Makefile
new file mode 100644
index 00000000000..c0345a9cb54
--- /dev/null
+++ b/src/gallium/drivers/Makefile
@@ -0,0 +1,24 @@
+TOP = ../../..
+include $(TOP)/configs/current
+
+
+ifeq ($(CONFIG_NAME), linux-cell)
+CELL_DIR = cell
+endif
+
+SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR)
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
diff --git a/src/mesa/pipe/cell/Makefile b/src/gallium/drivers/cell/Makefile
index 47aef7b05f6..47aef7b05f6 100644
--- a/src/mesa/pipe/cell/Makefile
+++ b/src/gallium/drivers/cell/Makefile
diff --git a/src/mesa/pipe/cell/common.h b/src/gallium/drivers/cell/common.h
index 4de514c3586..74b131fbefc 100644
--- a/src/mesa/pipe/cell/common.h
+++ b/src/gallium/drivers/cell/common.h
@@ -90,6 +90,7 @@
 #define CELL_CMD_STATE_VS_ARRAY_INFO 16
 #define CELL_CMD_STATE_BLEND         17
 #define CELL_CMD_VS_EXECUTE          18
+#define CELL_CMD_STATE_ATTRIB_FETCH  19
 
 
 #define CELL_NUM_BUFFERS 4
@@ -128,13 +129,19 @@ struct cell_command_clear_surface
  */
 struct cell_array_info
 {
-    uint64_t base;      /**< Base address of the 0th element. */
-    uint attr;          /**< Attribute that this state is for. */
-    uint pitch;         /**< Byte pitch from one entry to the next. */
-    uint format;        /**< Pipe format of each entry. */
+   uint64_t base;      /**< Base address of the 0th element. */
+   uint attr;          /**< Attribute that this state is for. */
+   uint pitch;         /**< Byte pitch from one entry to the next. */
+   uint size;
+   uint function_offset;
 } ALIGN16_ATTRIB;
 
 
+struct cell_attribute_fetch_code {
+   uint64_t base;
+   uint size;
+};
+
 struct cell_shader_info
 {
    unsigned num_outputs;
diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile
index 50060f5cd30..196ab777f54 100644
--- a/src/mesa/pipe/cell/ppu/Makefile
+++ b/src/gallium/drivers/cell/ppu/Makefile
@@ -34,14 +34,18 @@ SOURCES = \
 	cell_surface.c \
 	cell_texture.c \
 	cell_vbuf.c \
+	cell_vertex_fetch.c \
 	cell_vertex_shader.c \
 	cell_winsys.c
 
 
 OBJECTS = $(SOURCES:.c=.o) \
 
-INCLUDE_DIRS = -I$(TOP)/src/mesa
-
+INCLUDE_DIRS = \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers
 
 .c.o:
 	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/gallium/drivers/cell/ppu/cell_batch.c
index f45e5f25b64..f45e5f25b64 100644
--- a/src/mesa/pipe/cell/ppu/cell_batch.c
+++ b/src/gallium/drivers/cell/ppu/cell_batch.c
diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/gallium/drivers/cell/ppu/cell_batch.h
index a6eee0a8b18..a6eee0a8b18 100644
--- a/src/mesa/pipe/cell/ppu/cell_batch.h
+++ b/src/gallium/drivers/cell/ppu/cell_batch.h
diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c
index 07b908eec59..e588a30d5bc 100644
--- a/src/mesa/pipe/cell/ppu/cell_clear.c
+++ b/src/gallium/drivers/cell/ppu/cell_clear.c
@@ -35,7 +35,7 @@
 #include <stdint.h>
 #include "pipe/p_inlines.h"
 #include "pipe/p_util.h"
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 #include "cell_clear.h"
 #include "cell_context.h"
 #include "cell_batch.h"
diff --git a/src/mesa/pipe/cell/ppu/cell_clear.h b/src/gallium/drivers/cell/ppu/cell_clear.h
index ff47d43f4cd..ff47d43f4cd 100644
--- a/src/mesa/pipe/cell/ppu/cell_clear.h
+++ b/src/gallium/drivers/cell/ppu/cell_clear.h
diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c
index bbe1fd7a111..e1eb22f4685 100644
--- a/src/mesa/pipe/cell/ppu/cell_context.c
+++ b/src/gallium/drivers/cell/ppu/cell_context.c
@@ -37,9 +37,9 @@
 #include "pipe/p_format.h"
 #include "pipe/p_util.h"
 #include "pipe/p_winsys.h"
-#include "pipe/cell/common.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_private.h"
+#include "cell/common.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
 #include "cell_clear.h"
 #include "cell_context.h"
 #include "cell_draw_arrays.h"
diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h
index 3b63419b5eb..91f8e542a25 100644
--- a/src/mesa/pipe/cell/ppu/cell_context.h
+++ b/src/gallium/drivers/cell/ppu/cell_context.h
@@ -32,10 +32,11 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
-#include "pipe/draw/draw_vertex.h"
-#include "pipe/draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_vbuf.h"
 #include "cell_winsys.h"
-#include "pipe/cell/common.h"
+#include "cell/common.h"
+#include "ppc/rtasm/spe_asm.h"
 
 
 struct cell_vbuf_render;
@@ -111,6 +112,9 @@ struct cell_context
    /** [4] to ensure 16-byte alignment for each status word */
    uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB;
 
+
+   struct spe_function attrib_fetch;
+   unsigned attrib_fetch_offsets[PIPE_ATTRIB_MAX];
 };
 
 
diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
index 717cd8370f9..f12613649b9 100644
--- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
@@ -39,7 +39,7 @@
 #include "cell_draw_arrays.h"
 #include "cell_state.h"
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 
diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h
index d5df4aa05f8..d5df4aa05f8 100644
--- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h
diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c
index f62bc4650ce..20f27531fce 100644
--- a/src/mesa/pipe/cell/ppu/cell_flush.c
+++ b/src/gallium/drivers/cell/ppu/cell_flush.c
@@ -31,7 +31,7 @@
 #include "cell_flush.h"
 #include "cell_spu.h"
 #include "cell_render.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 void
diff --git a/src/mesa/pipe/cell/ppu/cell_flush.h b/src/gallium/drivers/cell/ppu/cell_flush.h
index eda351b1cbc..eda351b1cbc 100644
--- a/src/mesa/pipe/cell/ppu/cell_flush.h
+++ b/src/gallium/drivers/cell/ppu/cell_flush.h
diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/gallium/drivers/cell/ppu/cell_render.c
index 4ab277a4b24..b663b376222 100644
--- a/src/mesa/pipe/cell/ppu/cell_render.c
+++ b/src/gallium/drivers/cell/ppu/cell_render.c
@@ -34,7 +34,7 @@
 #include "cell_render.h"
 #include "cell_spu.h"
 #include "pipe/p_util.h"
-#include "pipe/draw/draw_private.h"
+#include "draw/draw_private.h"
 
 
 struct render_stage {
diff --git a/src/mesa/pipe/cell/ppu/cell_render.h b/src/gallium/drivers/cell/ppu/cell_render.h
index 826dcbafeba..826dcbafeba 100644
--- a/src/mesa/pipe/cell/ppu/cell_render.h
+++ b/src/gallium/drivers/cell/ppu/cell_render.h
diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/gallium/drivers/cell/ppu/cell_spu.c
index 7c83a47e574..419e74dc402 100644
--- a/src/mesa/pipe/cell/ppu/cell_spu.c
+++ b/src/gallium/drivers/cell/ppu/cell_spu.c
@@ -31,7 +31,7 @@
 #include "cell_spu.h"
 #include "pipe/p_format.h"
 #include "pipe/p_state.h"
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 
 
 /*
diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/gallium/drivers/cell/ppu/cell_spu.h
index 19eff94f967..137f26612e4 100644
--- a/src/mesa/pipe/cell/ppu/cell_spu.h
+++ b/src/gallium/drivers/cell/ppu/cell_spu.h
@@ -31,7 +31,7 @@
 
 #include <libspe2.h>
 #include <libmisc.h>
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 
 #include "cell_context.h"
 
diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h
index 3a71ba14fa8..3a71ba14fa8 100644
--- a/src/mesa/pipe/cell/ppu/cell_state.h
+++ b/src/gallium/drivers/cell/ppu/cell_state.h
diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/gallium/drivers/cell/ppu/cell_state_blend.c
index 4fc60548c85..b6d6d71f0ce 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_blend.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_blend.c
@@ -29,7 +29,7 @@
  */
 
 #include "pipe/p_util.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "cell_context.h"
 #include "cell_state.h"
 
diff --git a/src/mesa/pipe/cell/ppu/cell_state_clip.c b/src/gallium/drivers/cell/ppu/cell_state_clip.c
index 4f43665941d..0482f87e889 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_clip.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_clip.c
@@ -30,7 +30,7 @@
 
 #include "cell_context.h"
 #include "cell_state.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 void cell_set_clip_state( struct pipe_context *pipe,
diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c
index 56daf5dfde0..0c468292584 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_derived.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c
@@ -27,8 +27,8 @@
 
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
 #include "cell_context.h"
 #include "cell_batch.h"
 #include "cell_state.h"
diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c
index 5d2a7864493..5d2a7864493 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_emit.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c
diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.h b/src/gallium/drivers/cell/ppu/cell_state_emit.h
index 59f8affe8d3..59f8affe8d3 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_emit.h
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.h
diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/gallium/drivers/cell/ppu/cell_state_fs.c
index 3f46a87d189..b2ed699a5be 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_fs.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_fs.c
@@ -29,12 +29,12 @@
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #if 0
 #include "pipe/p_shader_tokens.h"
-#include "pipe/llvm/gallivm.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
-#include "pipe/tgsi/exec/tgsi_sse2.h"
+#include "llvm/gallivm.h"
+#include "tgsi/util/tgsi_dump.h"
+#include "tgsi/exec/tgsi_sse2.h"
 #endif
 
 #include "cell_context.h"
diff --git a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c
index d8128ece54d..7eca5b57656 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c
@@ -27,7 +27,7 @@
 
 #include "pipe/p_defines.h"
 #include "pipe/p_util.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "cell_context.h"
 #include "cell_state.h"
 
diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/gallium/drivers/cell/ppu/cell_state_sampler.c
index ade6cc8338e..a33421a4ad0 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_sampler.c
@@ -30,7 +30,7 @@
  */
 
 #include "pipe/p_util.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "cell_context.h"
 #include "cell_state.h"
 #include "cell_texture.h"
diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/gallium/drivers/cell/ppu/cell_state_surface.c
index 287610b76b9..287610b76b9 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_surface.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_surface.c
diff --git a/src/mesa/pipe/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
index 0f01e920f95..563831b62db 100644
--- a/src/mesa/pipe/cell/ppu/cell_state_vertex.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
@@ -32,7 +32,7 @@
 #include "cell_context.h"
 #include "cell_state.h"
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 void
diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c
index 6b7b9181282..a35db0ef991 100644
--- a/src/mesa/pipe/cell/ppu/cell_surface.c
+++ b/src/gallium/drivers/cell/ppu/cell_surface.c
@@ -29,7 +29,7 @@
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 #include "cell_context.h"
 #include "cell_surface.h"
 
@@ -60,6 +60,7 @@ cell_surface_data(struct pipe_context *pipe,
 
 static void
 cell_surface_copy(struct pipe_context *pipe,
+                  unsigned do_flip,
                   struct pipe_surface *dst,
                   unsigned dstx, unsigned dsty,
                   struct pipe_surface *src,
@@ -74,8 +75,8 @@ cell_surface_copy(struct pipe_context *pipe,
                   dstx, dsty,
                   width, height,
                   pipe_surface_map(src),
-                  src->pitch,
-                  srcx, srcy);
+                  do_flip ? -src->pitch : src->pitch,
+                  srcx, do_flip ? 1 - srcy - height : srcy);
 
    pipe_surface_unmap(src);
    pipe_surface_unmap(dst);
diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/gallium/drivers/cell/ppu/cell_surface.h
index 9e58f329443..9e58f329443 100644
--- a/src/mesa/pipe/cell/ppu/cell_surface.h
+++ b/src/gallium/drivers/cell/ppu/cell_surface.h
diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index df178d9ca24..c8ef36002f7 100644
--- a/src/mesa/pipe/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -61,7 +61,7 @@ cell_texture_layout(struct cell_texture * spt)
 
    spt->buffer_size = 0;
 
-   for ( level = pt->first_level ; level <= pt->last_level ; level++ ) {
+   for ( level = 0 ; level <= pt->last_level ; level++ ) {
       pt->width[level] = width;
       pt->height[level] = height;
       pt->depth[level] = depth;
diff --git a/src/mesa/pipe/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h
index 0264fed88e6..0264fed88e6 100644
--- a/src/mesa/pipe/cell/ppu/cell_texture.h
+++ b/src/gallium/drivers/cell/ppu/cell_texture.h
diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c
index e9fafe492ee..cc727ff4edb 100644
--- a/src/mesa/pipe/cell/ppu/cell_vbuf.c
+++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c
@@ -36,7 +36,7 @@
 #include "cell_flush.h"
 #include "cell_spu.h"
 #include "cell_vbuf.h"
-#include "pipe/draw/draw_vbuf.h"
+#include "draw/draw_vbuf.h"
 
 
 /** Allow vertex data to be inlined after RENDER command */
diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.h b/src/gallium/drivers/cell/ppu/cell_vbuf.h
index d265cbf7701..d265cbf7701 100644
--- a/src/mesa/pipe/cell/ppu/cell_vbuf.h
+++ b/src/gallium/drivers/cell/ppu/cell_vbuf.h
diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
new file mode 100644
index 00000000000..f10689a959e
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
@@ -0,0 +1,383 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <inttypes.h>
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_format.h"
+
+#include "../auxiliary/draw/draw_context.h"
+#include "../auxiliary/draw/draw_private.h"
+
+#include "cell_context.h"
+#include "ppc/rtasm/spe_asm.h"
+
+typedef uint64_t register_mask;
+
+int allocate_available_register(register_mask *m)
+{
+   unsigned i;
+   for (i = 0; i < 64; i++) {
+      const uint64_t mask = (1ULL << i);
+
+      if ((m[0] & mask) != 0) {
+	 m[0] &= ~mask;
+	 return i;
+      }
+   }
+
+   return -1;
+}
+
+
+int allocate_register(register_mask *m, unsigned reg)
+{
+   assert((m[0] & (1ULL << reg)) != 0);
+
+   m[0] &= ~(1ULL << reg);
+   return reg;
+}
+
+
+void release_register(register_mask *m, unsigned reg)
+{
+   assert((m[0] & (1ULL << reg)) == 0);
+
+   m[0] |= (1ULL << reg);
+}
+
+
+/**
+ * Emit a 4x4 matrix transpose operation
+ *
+ * \param p         Function that the transpose operation is to be appended to
+ * \param m         Live register mask
+ * \param row0      Register containing row 0 of the source matrix
+ * \param row1      Register containing row 1 of the source matrix
+ * \param row2      Register containing row 2 of the source matrix
+ * \param row3      Register containing row 3 of the source matrix
+ * \param dest_ptr  Register containing the address of the destination matrix
+ * \param shuf_ptr  Register containing the address of the shuffled data
+ * \param count     Number of colums to actually be written to the destination
+ *
+ * \note
+ * This function assumes that the registers named by \c row0, \c row1,
+ * \c row2, and \c row3 are scratch and can be modified by the generated code.
+ * Furthermore, these registers will be released, via calls to
+ * \c release_register, by this function.
+ * 
+ * \note
+ * This function requires that four temporary are available on entry.
+ */
+static void
+emit_matrix_transpose(struct spe_function *p, register_mask *m,
+		      unsigned row0, unsigned row1, unsigned row2,
+		      unsigned row3, unsigned dest_ptr,
+		      unsigned shuf_ptr, unsigned count)
+{
+   int shuf_hi = allocate_available_register(m);
+   int shuf_lo = allocate_available_register(m);
+   int t1 = allocate_available_register(m);
+   int t2 = allocate_available_register(m);
+   int t3;
+   int t4;
+   int col0;
+   int col1;
+   int col2;
+   int col3;
+
+
+   spe_lqd(p, shuf_hi, shuf_ptr, 3);
+   spe_lqd(p, shuf_lo, shuf_ptr, 4);
+   spe_shufb(p, t1, row0, row2, shuf_hi);
+   spe_shufb(p, t2, row0, row2, shuf_lo);
+
+
+   /* row0 and row2 are now no longer needed.  Re-use those registers as
+    * temporaries.
+    */
+   t3 = row0;
+   t4 = row2;
+
+   spe_shufb(p, t3, row1, row3, shuf_hi);
+   spe_shufb(p, t4, row1, row3, shuf_lo);
+
+
+   /* row1 and row3 are now no longer needed.  Re-use those registers as
+    * temporaries.
+    */
+   col0 = row1;
+   col1 = row3;
+
+   spe_shufb(p, col0, t1, t3, shuf_hi);
+   if (count > 1) {
+      spe_shufb(p, col1, t1, t3, shuf_lo);
+   }
+
+   /* t1 and t3 are now no longer needed.  Re-use those registers as
+    * temporaries.
+    */
+   col2 = t1;
+   col3 = t3;
+
+   if (count > 2) {
+      spe_shufb(p, col2, t2, t4, shuf_hi);
+   }
+
+   if (count > 3) {
+      spe_shufb(p, col3, t2, t4, shuf_lo);
+   }
+
+
+   /* Store the results.  Remember that the stqd instruction is encoded using
+    * the qword offset (stand-alone assemblers to the byte-offset to
+    * qword-offset conversion for you), so the byte-offset needs be divided by
+    * 16.
+    */
+   switch (count) {
+   case 4:
+      spe_stqd(p, col3, dest_ptr, 3);
+   case 3:
+      spe_stqd(p, col2, dest_ptr, 2);
+   case 2:
+      spe_stqd(p, col1, dest_ptr, 1);
+   case 1:
+      spe_stqd(p, col0, dest_ptr, 0);
+   }
+
+
+   /* Release all of the temporary registers used.
+    */
+   release_register(m, col0);
+   release_register(m, col1);
+   release_register(m, col2);
+   release_register(m, col3);
+   release_register(m, shuf_hi);
+   release_register(m, shuf_lo);
+   release_register(m, t2);
+   release_register(m, t4);
+}
+
+
+static void
+emit_fetch(struct spe_function *p, register_mask *m,
+	   unsigned in_ptr, unsigned *offset,
+	   unsigned out_ptr, unsigned shuf_ptr,
+	   enum pipe_format format)
+{
+   const unsigned count = (pf_size_x(format) != 0) + (pf_size_y(format) != 0)
+       + (pf_size_z(format) != 0) + (pf_size_w(format) != 0);
+   const unsigned type = pf_type(format);
+   const unsigned bytes = pf_size_x(format);
+
+   int v0 = allocate_available_register(m);
+   int v1 = allocate_available_register(m);
+   int v2 = allocate_available_register(m);
+   int v3 = allocate_available_register(m);
+   int tmp = allocate_available_register(m);
+   int float_zero = -1;
+   int float_one = -1;
+   float scale_signed = 0.0;
+   float scale_unsigned = 0.0;
+
+   spe_lqd(p, v0, in_ptr, 0 + offset[0]);
+   spe_lqd(p, v1, in_ptr, 1 + offset[0]);
+   spe_lqd(p, v2, in_ptr, 2 + offset[0]);
+   spe_lqd(p, v3, in_ptr, 3 + offset[0]);
+   offset[0] += 4;
+   
+   switch (bytes) {
+   case 1:
+      scale_signed = 1.0f / 127.0f;
+      scale_unsigned = 1.0f / 255.0f;
+      spe_lqd(p, tmp, shuf_ptr, 1);
+      spe_shufb(p, v0, v0, v0, tmp);
+      spe_shufb(p, v1, v1, v1, tmp);
+      spe_shufb(p, v2, v2, v2, tmp);
+      spe_shufb(p, v3, v3, v3, tmp);
+      break;
+   case 2:
+      scale_signed = 1.0f / 32767.0f;
+      scale_unsigned = 1.0f / 65535.0f;
+      spe_lqd(p, tmp, shuf_ptr, 2);
+      spe_shufb(p, v0, v0, v0, tmp);
+      spe_shufb(p, v1, v1, v1, tmp);
+      spe_shufb(p, v2, v2, v2, tmp);
+      spe_shufb(p, v3, v3, v3, tmp);
+      break;
+   case 4:
+      scale_signed = 1.0f / 2147483647.0f;
+      scale_unsigned = 1.0f / 4294967295.0f;
+      break;
+   default:
+      assert(0);
+      break;
+   }
+
+   switch (type) {
+   case PIPE_FORMAT_TYPE_FLOAT:
+      break;
+   case PIPE_FORMAT_TYPE_UNORM:
+      spe_ilhu(p, tmp, ((unsigned) scale_unsigned) >> 16);
+      spe_iohl(p, tmp, ((unsigned) scale_unsigned) & 0x0ffff);
+      spe_cuflt(p, v0, v0, 0);
+      spe_fm(p, v0, v0, tmp);
+      break;
+   case PIPE_FORMAT_TYPE_SNORM:
+      spe_ilhu(p, tmp, ((unsigned) scale_signed) >> 16);
+      spe_iohl(p, tmp, ((unsigned) scale_signed) & 0x0ffff);
+      spe_csflt(p, v0, v0, 0);
+      spe_fm(p, v0, v0, tmp);
+      break;
+   case PIPE_FORMAT_TYPE_USCALED:
+      spe_cuflt(p, v0, v0, 0);
+      break;
+   case PIPE_FORMAT_TYPE_SSCALED:
+      spe_csflt(p, v0, v0, 0);
+      break;
+   }
+
+
+   if (count < 4) {
+      float_one = allocate_available_register(m);
+      spe_il(p, float_one, 1);
+      spe_cuflt(p, float_one, float_one, 0);
+      
+      if (count < 3) {
+	 float_zero = allocate_available_register(m);
+	 spe_il(p, float_zero, 0);
+      }
+   }
+
+   release_register(m, tmp);
+
+   emit_matrix_transpose(p, m, v0, v1, v2, v3, out_ptr, shuf_ptr, count);
+
+   switch (count) {
+   case 1:
+      spe_stqd(p, float_zero, out_ptr, 1);
+   case 2:
+      spe_stqd(p, float_zero, out_ptr, 2);
+   case 3:
+      spe_stqd(p, float_one, out_ptr, 3);
+   }
+
+   if (float_zero != -1) {
+      release_register(m, float_zero);
+   }
+
+   if (float_one != -1) {
+      release_register(m, float_one);
+   }
+}
+
+
+void cell_update_vertex_fetch(struct draw_context *draw)
+{
+   struct cell_context *const cell =
+       (struct cell_context *) draw->driver_private;
+   register_mask m = ~0;
+   struct spe_function *p = &cell->attrib_fetch;
+   unsigned function_index[PIPE_ATTRIB_MAX];
+   unsigned unique_attr_formats;
+   int out_ptr;
+   int in_ptr;
+   int shuf_ptr;
+   unsigned i;
+   unsigned j;
+
+
+   /* Determine how many unique input attribute formats there are.  At the
+    * same time, store the index of the lowest numbered attribute that has
+    * the same format as any non-unique format.
+    */
+   unique_attr_formats = 1;
+   function_index[0] = 0;
+   for (i = 1; i < draw->vertex_fetch.nr_attrs; i++) {
+      const enum pipe_format curr_fmt = draw->vertex_element[i].src_format;
+
+      for (j = 0; j < i; j++) {
+	 if (curr_fmt == draw->vertex_element[j].src_format) {
+	    break;
+	 }
+      }
+      
+      if (j == i) {
+	 unique_attr_formats++;
+      }
+
+      function_index[i] = j;
+   }
+
+
+   /* Each fetch function can be a maximum of 34 instructions (note: this is
+    * actually a slight over-estimate).  That means (34 * 4) = 136 bytes
+    * each maximum.
+    */
+   spe_init_func(p, 136 * unique_attr_formats);
+
+
+   /* Registers 0, 1, and 2 are reserved by the ABI.
+    */
+   allocate_register(&m, 0);
+   allocate_register(&m, 1);
+   allocate_register(&m, 2);
+
+
+   /* Allocate registers for the function's input parameters.
+    */
+   out_ptr = allocate_register(&m, 3);
+   in_ptr = allocate_register(&m, 4);
+   shuf_ptr = allocate_register(&m, 5);
+
+
+   /* Generate code for the individual attribute fetch functions.
+    */
+   for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) {
+      unsigned offset;
+
+      if (function_index[i] == i) {
+	 cell->attrib_fetch_offsets[i] = (unsigned) ((void *) p->csr 
+						     - (void *) p->store);
+
+	 offset = 0;
+	 emit_fetch(p, & m, in_ptr, &offset, out_ptr, shuf_ptr,
+		    draw->vertex_element[i].src_format);
+	 spe_bi(p, 0, 0, 0);
+
+	 /* Round up to the next 16-byte boundary.
+	  */
+	 if ((((unsigned) p->store) & 0x0f) != 0) {
+	    const unsigned align = ((unsigned) p->store) & 0x0f;
+	    p->store = (uint32_t *) (((void *) p->store) + align);
+	 }
+      } else {
+	 /* Use the same function entry-point as a previously seen attribute
+	  * with the same format.
+	  */
+	 cell->attrib_fetch_offsets[i] = 
+	     cell->attrib_fetch_offsets[function_index[i]];
+      }
+   }
+}
diff --git a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
index 80dd500b345..6a1d3bc20a1 100644
--- a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c
+++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
@@ -38,9 +38,9 @@
 #include "cell_spu.h"
 #include "cell_batch.h"
 
-#include "pipe/cell/common.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_private.h"
+#include "cell/common.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
 
 /**
  * Run the vertex shader on all vertices in the vertex queue.
@@ -55,14 +55,32 @@ cell_vertex_shader_queue_flush(struct draw_context *draw)
    uint64_t *batch;
    struct cell_array_info *array_info;
    unsigned i, j;
+   struct cell_attribute_fetch_code *cf;
 
    assert(draw->vs.queue_nr != 0);
 
    /* XXX: do this on statechange: 
     */
    draw_update_vertex_fetch(draw);
+   cell_update_vertex_fetch(draw);
+
+
+   batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*cf));
+   batch[0] = CELL_CMD_STATE_ATTRIB_FETCH;
+   cf = (struct cell_attribute_fetch_code *) (&batch[1]);
+   cf->base = cell->attrib_fetch.store;
+   cf->size = ROUNDUP16((unsigned)((void *) cell->attrib_fetch.csr 
+				   - (void *) cell->attrib_fetch.store));
+
 
    for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) {
+      const enum pipe_format format = draw->vertex_element[i].src_format;
+      const unsigned count = ((pf_size_x(format) != 0)
+			      + (pf_size_y(format) != 0)
+			      + (pf_size_z(format) != 0)
+			      + (pf_size_w(format) != 0));
+      const unsigned size = pf_size_x(format) * count;
+
       batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*array_info));
 
       batch[0] = CELL_CMD_STATE_VS_ARRAY_INFO;
@@ -72,7 +90,8 @@ cell_vertex_shader_queue_flush(struct draw_context *draw)
       array_info->base = (uintptr_t) draw->vertex_fetch.src_ptr[i];
       array_info->attr = i;
       array_info->pitch = draw->vertex_fetch.pitch[i];
-      array_info->format = draw->vertex_element[i].src_format;
+      array_info->size = size;
+      array_info->function_offset = cell->attrib_fetch_offsets[i];
    }
 
    batch = cell_batch_alloc(cell, sizeof(batch[0])
diff --git a/src/mesa/pipe/cell/ppu/cell_winsys.c b/src/gallium/drivers/cell/ppu/cell_winsys.c
index ebabce3c8f5..ebabce3c8f5 100644
--- a/src/mesa/pipe/cell/ppu/cell_winsys.c
+++ b/src/gallium/drivers/cell/ppu/cell_winsys.c
diff --git a/src/mesa/pipe/cell/ppu/cell_winsys.h b/src/gallium/drivers/cell/ppu/cell_winsys.h
index ae2af5696b5..ae2af5696b5 100644
--- a/src/mesa/pipe/cell/ppu/cell_winsys.h
+++ b/src/gallium/drivers/cell/ppu/cell_winsys.h
diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile
index f202971d738..30ef2450ece 100644
--- a/src/mesa/pipe/cell/spu/Makefile
+++ b/src/gallium/drivers/cell/spu/Makefile
@@ -31,7 +31,11 @@ SPU_OBJECTS = $(SOURCES:.c=.o) \
 
 SPU_ASM_OUT = $(SOURCES:.c=.s) \
 
-INCLUDE_DIRS = -I$(TOP)/src/mesa
+INCLUDE_DIRS = \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers
 
 
 .c.o:
diff --git a/src/mesa/pipe/cell/spu/spu_blend.c b/src/gallium/drivers/cell/spu/spu_blend.c
index 23ec0eeb451..23ec0eeb451 100644
--- a/src/mesa/pipe/cell/spu/spu_blend.c
+++ b/src/gallium/drivers/cell/spu/spu_blend.c
diff --git a/src/mesa/pipe/cell/spu/spu_blend.h b/src/gallium/drivers/cell/spu/spu_blend.h
index 2b594b578b4..2b594b578b4 100644
--- a/src/mesa/pipe/cell/spu/spu_blend.h
+++ b/src/gallium/drivers/cell/spu/spu_blend.h
diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/gallium/drivers/cell/spu/spu_colorpack.h
index e9fee8a3a61..e9fee8a3a61 100644
--- a/src/mesa/pipe/cell/spu/spu_colorpack.h
+++ b/src/gallium/drivers/cell/spu/spu_colorpack.h
diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c
index e51008b9b3c..109540b1f7b 100644
--- a/src/mesa/pipe/cell/spu/spu_exec.c
+++ b/src/gallium/drivers/cell/spu/spu_exec.c
@@ -67,8 +67,8 @@
 #include "pipe/p_state.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
-#include "pipe/tgsi/util/tgsi_util.h"
+#include "tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_util.h"
 #include "spu_exec.h"
 #include "spu_main.h"
 #include "spu_vertex_shader.h"
diff --git a/src/mesa/pipe/cell/spu/spu_exec.h b/src/gallium/drivers/cell/spu/spu_exec.h
index b4c7661ef67..3e17c490d20 100644
--- a/src/mesa/pipe/cell/spu/spu_exec.h
+++ b/src/gallium/drivers/cell/spu/spu_exec.h
@@ -29,7 +29,7 @@
 #define SPU_EXEC_H
 
 #include "pipe/p_compiler.h"
-#include "pipe/tgsi/exec/tgsi_exec.h"
+#include "tgsi/exec/tgsi_exec.h"
 
 #if defined __cplusplus
 extern "C" {
diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c
index e375197fe60..fcbf0f841e6 100644
--- a/src/mesa/pipe/cell/spu/spu_main.c
+++ b/src/gallium/drivers/cell/spu/spu_main.c
@@ -38,7 +38,7 @@
 #include "spu_tile.h"
 //#include "spu_test.h"
 #include "spu_vertex_shader.h"
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 #include "pipe/p_defines.h"
 
 
@@ -54,6 +54,9 @@ struct spu_global spu;
 
 struct spu_vs_context draw;
 
+static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX]
+    ALIGN16_ATTRIB;
+
 /**
  * Tell the PPU that this SPU has finished copying a buffer to
  * local store and that it may be reused by the PPU.
@@ -306,7 +309,8 @@ cmd_state_vs_array_info(const struct cell_array_info *vs_info)
    ASSERT(attr < PIPE_ATTRIB_MAX);
    draw.vertex_fetch.src_ptr[attr] = vs_info->base;
    draw.vertex_fetch.pitch[attr] = vs_info->pitch;
-   draw.vertex_fetch.format[attr] = vs_info->format;
+   draw.vertex_fetch.size[attr] = vs_info->size;
+   draw.vertex_fetch.code_offset[attr] = vs_info->function_offset;
    draw.vertex_fetch.dirty = 1;
 }
 
@@ -433,6 +437,22 @@ cmd_batch(uint opcode)
          cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]);
          pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8);
          break;
+      case CELL_CMD_STATE_ATTRIB_FETCH: {
+         struct cell_attribute_fetch_code *code =
+             (struct cell_attribute_fetch_code *) &buffer[pos+1];
+
+              mfc_get(attribute_fetch_code_buffer,
+                      (unsigned int) code->base,  /* src */
+                      code->size,
+                      TAG_BATCH_BUFFER,
+                      0, /* tid */
+                      0  /* rid */);
+         wait_on_mask(1 << TAG_BATCH_BUFFER);
+
+         draw.vertex_fetch.code = attribute_fetch_code_buffer;
+         pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8);
+         break;
+      }
       default:
          printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]);
          ASSERT(0);
diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h
index 1710a175123..5c95d112ac1 100644
--- a/src/mesa/pipe/cell/spu/spu_main.h
+++ b/src/gallium/drivers/cell/spu/spu_main.h
@@ -31,8 +31,8 @@
 
 #include <spu_mfcio.h>
 
-#include "pipe/cell/common.h"
-#include "pipe/draw/draw_vertex.h"
+#include "cell/common.h"
+#include "draw/draw_vertex.h"
 #include "pipe/p_state.h"
 
 
diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c
index 932fb500b3f..20e77aa2e63 100644
--- a/src/mesa/pipe/cell/spu/spu_render.c
+++ b/src/gallium/drivers/cell/spu/spu_render.c
@@ -34,7 +34,7 @@
 #include "spu_render.h"
 #include "spu_tri.h"
 #include "spu_tile.h"
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 
 
 
diff --git a/src/mesa/pipe/cell/spu/spu_render.h b/src/gallium/drivers/cell/spu/spu_render.h
index fbcdc5ec316..493434f0878 100644
--- a/src/mesa/pipe/cell/spu/spu_render.h
+++ b/src/gallium/drivers/cell/spu/spu_render.h
@@ -29,7 +29,7 @@
 #ifndef SPU_RENDER_H
 #define SPU_RENDER_H
 
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 
 extern void
 cmd_render(const struct cell_command_render *render, uint *pos_incr);
diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c
index 3962aaa4a9b..3962aaa4a9b 100644
--- a/src/mesa/pipe/cell/spu/spu_texture.c
+++ b/src/gallium/drivers/cell/spu/spu_texture.c
diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/gallium/drivers/cell/spu/spu_texture.h
index 95eb87080f1..95eb87080f1 100644
--- a/src/mesa/pipe/cell/spu/spu_texture.h
+++ b/src/gallium/drivers/cell/spu/spu_texture.h
diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/gallium/drivers/cell/spu/spu_tile.c
index 12dc2463283..12dc2463283 100644
--- a/src/mesa/pipe/cell/spu/spu_tile.c
+++ b/src/gallium/drivers/cell/spu/spu_tile.c
diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/gallium/drivers/cell/spu/spu_tile.h
index e53340a55a4..3105b848fdc 100644
--- a/src/mesa/pipe/cell/spu/spu_tile.h
+++ b/src/gallium/drivers/cell/spu/spu_tile.h
@@ -32,7 +32,7 @@
 #include <libmisc.h>
 #include <spu_mfcio.h>
 #include "spu_main.h"
-#include "pipe/cell/common.h"
+#include "cell/common.h"
 
 
 
diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
index be9624cf7d9..be9624cf7d9 100644
--- a/src/mesa/pipe/cell/spu/spu_tri.c
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
diff --git a/src/mesa/pipe/cell/spu/spu_tri.h b/src/gallium/drivers/cell/spu/spu_tri.h
index aa694dd7c93..aa694dd7c93 100644
--- a/src/mesa/pipe/cell/spu/spu_tri.h
+++ b/src/gallium/drivers/cell/spu/spu_tri.h
diff --git a/src/mesa/pipe/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c
index ac373240c1f..ea4274a0a7b 100644
--- a/src/mesa/pipe/cell/spu/spu_util.c
+++ b/src/gallium/drivers/cell/spu/spu_util.c
@@ -1,8 +1,8 @@
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_parse.h"
 //#include "tgsi_build.h"
-#include "pipe/tgsi/util/tgsi_util.h"
+#include "tgsi/util/tgsi_util.h"
 
 unsigned
 tgsi_util_get_src_register_swizzle(
diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c
new file mode 100644
index 00000000000..55c6c287175
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c
@@ -0,0 +1,214 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Ian Romanick <[email protected]>
+  */
+
+#include <spu_mfcio.h>
+
+#include "pipe/p_util.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "spu_exec.h"
+#include "spu_vertex_shader.h"
+#include "spu_main.h"
+
+#define CACHE_NAME            attribute
+#define CACHED_TYPE           qword
+#define CACHE_TYPE            CACHE_TYPE_RO
+#define CACHE_SET_TAGID(set)  TAG_VERTEX_BUFFER
+#define CACHE_LOG2NNWAY       2
+#define CACHE_LOG2NSETS       6
+#include <cache-api.h>
+
+/* Yes folks, this is ugly.
+ */
+#undef CACHE_NWAY
+#undef CACHE_NSETS
+#define CACHE_NAME            attribute
+#define CACHE_NWAY            4
+#define CACHE_NSETS           (1U << 6)
+
+
+#define DRAW_DBG 0
+
+typedef void (*spu_fetch_func)(qword *out, const qword *in,
+			       const qword *shuffle_data);
+
+
+static const qword fetch_shuffle_data[] = {
+   /* Shuffle used by CVT_64_FLOAT
+    */
+   {
+      0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13,
+      0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
+   },
+
+   /* Shuffle used by CVT_8_USCALED and CVT_8_SSCALED
+    */
+   {
+      0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80,
+      0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80,
+   },
+   
+   /* Shuffle used by CVT_16_USCALED and CVT_16_SSCALED
+    */
+   {
+      0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80,
+      0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80,
+   },
+   
+   /* High value shuffle used by trans4x4.
+    */
+   {
+      0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13,
+      0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17
+   },
+
+   /* Low value shuffle used by trans4x4.
+    */
+   {
+      0x08, 0x09, 0x0A, 0x0B, 0x18, 0x19, 0x1A, 0x1B,
+      0x0C, 0x0D, 0x0E, 0x0F, 0x1C, 0x1D, 0x1E, 0x1F
+   }
+};
+
+
+/**
+ * Fetch between 1 and 32 bytes from an unaligned address
+ */
+static INLINE void
+fetch_unaligned(qword *dst, unsigned ea, unsigned size)
+{
+   qword tmp[4];
+   const int shift = ea & 0x0f;
+   const unsigned aligned_start_ea = ea & ~0x0f;
+   const unsigned aligned_end_ea = (ea + size) & ~0x0f;
+   const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1;
+   unsigned i;
+
+
+   if (shift == 0) {
+      /* Data is already aligned.  Fetch directly into the destination buffer.
+       */
+      for (i = 0; i < num_entries; i++) {
+	 dst[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16));
+      }
+   } else {
+      /* Fetch data from the cache to the local buffer.
+       */
+      for (i = 0; i < num_entries; i++) {
+	 tmp[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16));
+      }
+
+
+      /* Fix the alignment of the data and write to the destination buffer.
+       */
+      for (i = 0; i < ((size + 15) / 16); i++) {
+	 dst[i] = si_or((qword) spu_slqwbyte(tmp[i], shift),
+			(qword) spu_rlmaskqwbyte(tmp[i + 1], shift - 16));
+      }
+   }
+}
+
+
+/**
+ * Fetch vertex attributes for 'count' vertices.
+ */
+static void generic_vertex_fetch(struct spu_vs_context *draw,
+                                 struct spu_exec_machine *machine,
+                                 const unsigned *elts,
+                                 unsigned count)
+{
+   unsigned nr_attrs = draw->vertex_fetch.nr_attrs;
+   unsigned attr;
+
+   assert(count <= 4);
+
+#if DRAW_DBG
+   printf("SPU: %s count = %u, nr_attrs = %u\n", 
+          __FUNCTION__, count, nr_attrs);
+#endif
+
+   /* loop over vertex attributes (vertex shader inputs)
+    */
+   for (attr = 0; attr < nr_attrs; attr++) {
+      const unsigned pitch = draw->vertex_fetch.pitch[attr];
+      const uint64_t src = draw->vertex_fetch.src_ptr[attr];
+      const spu_fetch_func fetch = (spu_fetch_func)
+	  (draw->vertex_fetch.code + draw->vertex_fetch.code_offset[attr]);
+      unsigned i;
+      unsigned idx;
+      const unsigned bytes_per_entry = draw->vertex_fetch.size[attr];
+      const unsigned quads_per_entry = (bytes_per_entry + 15) / 16;
+      qword in[2 * 4];
+
+
+      /* Fetch four attributes for four vertices.  
+       */
+      idx = 0;
+      for (i = 0; i < count; i++) {
+         const uint64_t addr = src + (elts[i] * pitch);
+
+#if DRAW_DBG
+         printf("SPU: fetching = 0x%llx\n", addr);
+#endif
+
+         fetch_unaligned(& in[idx], addr, bytes_per_entry);
+         idx += quads_per_entry;
+      }
+
+      /* Be nice and zero out any missing vertices.
+       */
+      (void) memset(& in[idx], 0, (8 - idx) * sizeof(qword));
+
+
+      /* Convert all 4 vertices to vectors of float.
+       */
+      (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, fetch_shuffle_data);
+   }
+}
+
+
+void spu_update_vertex_fetch( struct spu_vs_context *draw )
+{
+   unsigned i;
+
+   
+   /* Invalidate the vertex cache.
+    */
+   for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) {
+      CACHELINE_CLEARVALID(i);
+   }
+
+
+   draw->vertex_fetch.fetch_func = generic_vertex_fetch;
+}
diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c
index c1cbbb6d1e9..3f5bf41aa2f 100644
--- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c
+++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c
@@ -39,9 +39,9 @@
 #include "pipe/p_shader_tokens.h"
 #include "spu_vertex_shader.h"
 #include "spu_exec.h"
-#include "pipe/draw/draw_private.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/cell/common.h"
+#include "draw/draw_private.h"
+#include "draw/draw_context.h"
+#include "cell/common.h"
 #include "spu_main.h"
 
 static INLINE unsigned
diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h
index c96b93ff0ac..0fb0bc28d03 100644
--- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h
+++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h
@@ -6,7 +6,6 @@
 
 struct spu_vs_context;
 
-typedef qword (*spu_fetch_func)(const void *ptr);
 typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw,
 				     struct spu_exec_machine *machine,
 				     const unsigned *elts,
@@ -18,12 +17,13 @@ struct spu_vs_context {
    struct {
       uint64_t src_ptr[PIPE_ATTRIB_MAX];
       unsigned pitch[PIPE_ATTRIB_MAX];
-      enum pipe_format format[PIPE_ATTRIB_MAX];
+      unsigned size[PIPE_ATTRIB_MAX];
+      unsigned code_offset[PIPE_ATTRIB_MAX];
       unsigned nr_attrs;
       boolean dirty;
 
-      spu_fetch_func fetch[PIPE_ATTRIB_MAX];
       spu_full_fetch_func fetch_func;
+      void *code;
    } vertex_fetch;
    
    /* Clip derived state:
diff --git a/src/mesa/pipe/cell/spu/spu_ztest.h b/src/gallium/drivers/cell/spu/spu_ztest.h
index ce8ad003393..ce8ad003393 100644
--- a/src/mesa/pipe/cell/spu/spu_ztest.h
+++ b/src/gallium/drivers/cell/spu/spu_ztest.h
diff --git a/src/mesa/pipe/failover/Makefile b/src/gallium/drivers/failover/Makefile
index 72d0895c74d..14389bd0551 100644
--- a/src/mesa/pipe/failover/Makefile
+++ b/src/gallium/drivers/failover/Makefile
@@ -15,7 +15,7 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-include ../Makefile.template
+include ../../Makefile.template
 
 symlinks:
 
diff --git a/src/mesa/pipe/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c
index 7ce4a7df17e..7ce4a7df17e 100644
--- a/src/mesa/pipe/failover/fo_context.c
+++ b/src/gallium/drivers/failover/fo_context.c
diff --git a/src/mesa/pipe/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h
index 1dc87291c9f..1dc87291c9f 100644
--- a/src/mesa/pipe/failover/fo_context.h
+++ b/src/gallium/drivers/failover/fo_context.h
diff --git a/src/mesa/pipe/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c
index 0fc5568da11..0fc5568da11 100644
--- a/src/mesa/pipe/failover/fo_state.c
+++ b/src/gallium/drivers/failover/fo_state.c
diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c
index c663dd49476..c663dd49476 100644
--- a/src/mesa/pipe/failover/fo_state_emit.c
+++ b/src/gallium/drivers/failover/fo_state_emit.c
diff --git a/src/mesa/pipe/failover/fo_winsys.h b/src/gallium/drivers/failover/fo_winsys.h
index a8ce997a1f6..a8ce997a1f6 100644
--- a/src/mesa/pipe/failover/fo_winsys.h
+++ b/src/gallium/drivers/failover/fo_winsys.h
diff --git a/src/mesa/pipe/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile
index 2f91de3afc6..ee22ba86f9b 100644
--- a/src/mesa/pipe/i915simple/Makefile
+++ b/src/gallium/drivers/i915simple/Makefile
@@ -32,7 +32,7 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-include ../Makefile.template
+include ../../Makefile.template
 
 symlinks:
 
diff --git a/src/mesa/pipe/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript
index f5fb96b995c..f5fb96b995c 100644
--- a/src/mesa/pipe/i915simple/SConscript
+++ b/src/gallium/drivers/i915simple/SConscript
diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h
index fb88cd6db0d..fb88cd6db0d 100644
--- a/src/mesa/pipe/i915simple/i915_batch.h
+++ b/src/gallium/drivers/i915simple/i915_batch.h
diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c
index d49876f9708..db4671ff553 100644
--- a/src/mesa/pipe/i915simple/i915_blit.c
+++ b/src/gallium/drivers/i915simple/i915_blit.c
@@ -85,6 +85,7 @@ i915_fill_blit(struct i915_context *i915,
 
 void
 i915_copy_blit( struct i915_context *i915,
+                  unsigned do_flip,
                   unsigned cpp,
                   short src_pitch,
                   struct pipe_buffer *src_buffer,
diff --git a/src/mesa/pipe/i915simple/i915_blit.h b/src/gallium/drivers/i915simple/i915_blit.h
index d7a66be10ae..6e5b44e1247 100644
--- a/src/mesa/pipe/i915simple/i915_blit.h
+++ b/src/gallium/drivers/i915simple/i915_blit.h
@@ -31,6 +31,7 @@
 #include "i915_context.h"
 
 extern void i915_copy_blit(struct i915_context *i915,
+                           unsigned do_flip,
 			   unsigned cpp,
 			   short src_pitch,
 			   struct pipe_buffer *src_buffer,
diff --git a/src/mesa/pipe/i915simple/i915_clear.c b/src/gallium/drivers/i915simple/i915_clear.c
index cde69daacc0..cde69daacc0 100644
--- a/src/mesa/pipe/i915simple/i915_clear.c
+++ b/src/gallium/drivers/i915simple/i915_clear.c
diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c
index 497623a7006..7f71f8fd4f5 100644
--- a/src/mesa/pipe/i915simple/i915_context.c
+++ b/src/gallium/drivers/i915simple/i915_context.c
@@ -32,7 +32,7 @@
 #include "i915_texture.h"
 #include "i915_reg.h"
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_winsys.h"
 #include "pipe/p_util.h"
diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h
index b4ea63c3e74..2d876925b2c 100644
--- a/src/mesa/pipe/i915simple/i915_context.h
+++ b/src/gallium/drivers/i915simple/i915_context.h
@@ -33,7 +33,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
 
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_vertex.h"
 
 
 #define I915_TEX_UNITS 8
diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c
index 94db44e1aaf..94db44e1aaf 100644
--- a/src/mesa/pipe/i915simple/i915_debug.c
+++ b/src/gallium/drivers/i915simple/i915_debug.c
diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/gallium/drivers/i915simple/i915_debug.h
index 0bcd0942334..0bcd0942334 100644
--- a/src/mesa/pipe/i915simple/i915_debug.h
+++ b/src/gallium/drivers/i915simple/i915_debug.h
diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c
index ebfdb3d93c5..ebfdb3d93c5 100644
--- a/src/mesa/pipe/i915simple/i915_debug_fp.c
+++ b/src/gallium/drivers/i915simple/i915_debug_fp.c
diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c
index 3c2069b8273..3c2069b8273 100644
--- a/src/mesa/pipe/i915simple/i915_flush.c
+++ b/src/gallium/drivers/i915simple/i915_flush.c
diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/gallium/drivers/i915simple/i915_fpc.h
index 8c7b68aefb5..8c7b68aefb5 100644
--- a/src/mesa/pipe/i915simple/i915_fpc.h
+++ b/src/gallium/drivers/i915simple/i915_fpc.h
diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c
index 74924ff0a1d..74924ff0a1d 100644
--- a/src/mesa/pipe/i915simple/i915_fpc_emit.c
+++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c
diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c
index 868f0c7e046..6c1524c768e 100644
--- a/src/mesa/pipe/i915simple/i915_fpc_translate.c
+++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c
@@ -33,9 +33,9 @@
 #include "i915_fpc.h"
 
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_parse.h"
 
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_vertex.h"
 
 
 /**
diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c
index c4a706c37d8..44c43259369 100644
--- a/src/mesa/pipe/i915simple/i915_prim_emit.c
+++ b/src/gallium/drivers/i915simple/i915_prim_emit.c
@@ -26,7 +26,7 @@
  **************************************************************************/
 
 
-#include "pipe/draw/draw_private.h"
+#include "draw/draw_private.h"
 #include "pipe/p_util.h"
 
 #include "i915_context.h"
diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
index e069773fd4e..c5bf6174f68 100644
--- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c
+++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
@@ -38,7 +38,7 @@
  */
 
 
-#include "pipe/draw/draw_vbuf.h"
+#include "draw/draw_vbuf.h"
 #include "pipe/p_debug.h"
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
diff --git a/src/mesa/pipe/i915simple/i915_reg.h b/src/gallium/drivers/i915simple/i915_reg.h
index 04620fec681..04620fec681 100644
--- a/src/mesa/pipe/i915simple/i915_reg.h
+++ b/src/gallium/drivers/i915simple/i915_reg.h
diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c
index abd5571b885..294e6fad035 100644
--- a/src/mesa/pipe/i915simple/i915_state.c
+++ b/src/gallium/drivers/i915simple/i915_state.c
@@ -29,7 +29,7 @@
  */
 
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "pipe/p_winsys.h"
 #include "pipe/p_util.h"
 
diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/gallium/drivers/i915simple/i915_state.h
index 86c6b0027d5..86c6b0027d5 100644
--- a/src/mesa/pipe/i915simple/i915_state.h
+++ b/src/gallium/drivers/i915simple/i915_state.h
diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c
index 653983e4a99..4767584fc60 100644
--- a/src/mesa/pipe/i915simple/i915_state_derived.c
+++ b/src/gallium/drivers/i915simple/i915_state_derived.c
@@ -27,8 +27,8 @@
 
 
 #include "pipe/p_util.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
 #include "i915_context.h"
 #include "i915_state.h"
 #include "i915_reg.h"
diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/gallium/drivers/i915simple/i915_state_dynamic.c
index 8cfbdddd19b..8cfbdddd19b 100644
--- a/src/mesa/pipe/i915simple/i915_state_dynamic.c
+++ b/src/gallium/drivers/i915simple/i915_state_dynamic.c
diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c
index 3339287f498..3339287f498 100644
--- a/src/mesa/pipe/i915simple/i915_state_emit.c
+++ b/src/gallium/drivers/i915simple/i915_state_emit.c
diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c
index 07031fc6c5b..07031fc6c5b 100644
--- a/src/mesa/pipe/i915simple/i915_state_immediate.c
+++ b/src/gallium/drivers/i915simple/i915_state_immediate.c
diff --git a/src/mesa/pipe/i915simple/i915_state_inlines.h b/src/gallium/drivers/i915simple/i915_state_inlines.h
index 0934ac79a41..0934ac79a41 100644
--- a/src/mesa/pipe/i915simple/i915_state_inlines.h
+++ b/src/gallium/drivers/i915simple/i915_state_inlines.h
diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c
index 9c1a5bbbd65..9c1a5bbbd65 100644
--- a/src/mesa/pipe/i915simple/i915_state_sampler.c
+++ b/src/gallium/drivers/i915simple/i915_state_sampler.c
diff --git a/src/mesa/pipe/i915simple/i915_strings.c b/src/gallium/drivers/i915simple/i915_strings.c
index c713bf72086..301fedea197 100644
--- a/src/mesa/pipe/i915simple/i915_strings.c
+++ b/src/gallium/drivers/i915simple/i915_strings.c
@@ -70,7 +70,7 @@ static const char *i915_get_name( struct pipe_context *pipe )
       break;
    }
 
-   sprintf(buffer, "pipe/i915 (chipset: %s)", chipset);
+   sprintf(buffer, "i915 (chipset: %s)", chipset);
    return buffer;
 }
 
diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c
index 1bdaba773f1..17fd27895a2 100644
--- a/src/mesa/pipe/i915simple/i915_surface.c
+++ b/src/gallium/drivers/i915simple/i915_surface.c
@@ -33,7 +33,7 @@
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 
 
 /*
@@ -83,6 +83,7 @@ i915_get_tex_surface(struct pipe_context *pipe,
  */
 static void
 i915_surface_copy(struct pipe_context *pipe,
+                  unsigned do_flip,
 		  struct pipe_surface *dst,
 		  unsigned dstx, unsigned dsty,
 		  struct pipe_surface *src,
@@ -98,14 +99,15 @@ i915_surface_copy(struct pipe_context *pipe,
                      dstx, dsty, 
                      width, height, 
                      pipe_surface_map(src), 
-                     src->pitch, 
-                     srcx, srcy);
+                     do_flip ? -(int) src->pitch : src->pitch, 
+                     srcx, do_flip ? 1 - srcy - height : srcy);
 
       pipe_surface_unmap(src);
       pipe_surface_unmap(dst);
    }
    else {
       i915_copy_blit( i915_context(pipe),
+                      do_flip,
 		      dst->cpp,
 		      (short) src->pitch, src->buffer, src->offset,
 		      (short) dst->pitch, dst->buffer, dst->offset,
diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c
index 6d37ae3d747..6d37ae3d747 100644
--- a/src/mesa/pipe/i915simple/i915_texture.c
+++ b/src/gallium/drivers/i915simple/i915_texture.c
diff --git a/src/mesa/pipe/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h
index 330d111dc78..330d111dc78 100644
--- a/src/mesa/pipe/i915simple/i915_texture.h
+++ b/src/gallium/drivers/i915simple/i915_texture.h
diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h
index fe49710852b..fe49710852b 100644
--- a/src/mesa/pipe/i915simple/i915_winsys.h
+++ b/src/gallium/drivers/i915simple/i915_winsys.h
diff --git a/src/mesa/pipe/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile
index 48c00ab50b8..1dec1f97495 100644
--- a/src/mesa/pipe/i965simple/Makefile
+++ b/src/gallium/drivers/i965simple/Makefile
@@ -61,6 +61,6 @@ ASM_SOURCES =
 
 DRIVER_DEFINES = -I.
 
-include ../Makefile.template
+include ../../Makefile.template
 
 symlinks:
diff --git a/src/mesa/pipe/i965simple/SConscript b/src/gallium/drivers/i965simple/SConscript
index 74621de84c9..74621de84c9 100644
--- a/src/mesa/pipe/i965simple/SConscript
+++ b/src/gallium/drivers/i965simple/SConscript
diff --git a/src/mesa/pipe/i965simple/brw_batch.h b/src/gallium/drivers/i965simple/brw_batch.h
index 5f5932a4883..5f5932a4883 100644
--- a/src/mesa/pipe/i965simple/brw_batch.h
+++ b/src/gallium/drivers/i965simple/brw_batch.h
diff --git a/src/mesa/pipe/i965simple/brw_blit.c b/src/gallium/drivers/i965simple/brw_blit.c
index bbd366294f5..8494f70493c 100644
--- a/src/mesa/pipe/i965simple/brw_blit.c
+++ b/src/gallium/drivers/i965simple/brw_blit.c
@@ -111,6 +111,7 @@ static unsigned translate_raster_op(unsigned logicop)
 /* Copy BitBlt
  */
 void brw_copy_blit(struct brw_context *brw,
+                   unsigned do_flip,
                    unsigned cpp,
                    short src_pitch,
                    struct pipe_buffer *src_buffer,
diff --git a/src/mesa/pipe/i965simple/brw_blit.h b/src/gallium/drivers/i965simple/brw_blit.h
index 7f17a701732..111c5d91d39 100644
--- a/src/mesa/pipe/i965simple/brw_blit.h
+++ b/src/gallium/drivers/i965simple/brw_blit.h
@@ -16,6 +16,7 @@ void brw_fill_blit(struct brw_context *intel,
                    short w, short h,
                    unsigned color);
 void brw_copy_blit(struct brw_context *intel,
+                   unsigned do_flip,
                    unsigned cpp,
                    short src_pitch,
                    struct pipe_buffer *src_buffer,
diff --git a/src/mesa/pipe/i965simple/brw_cc.c b/src/gallium/drivers/i965simple/brw_cc.c
index 337e4f95f69..337e4f95f69 100644
--- a/src/mesa/pipe/i965simple/brw_cc.c
+++ b/src/gallium/drivers/i965simple/brw_cc.c
diff --git a/src/mesa/pipe/i965simple/brw_clip.c b/src/gallium/drivers/i965simple/brw_clip.c
index 268124cc53f..268124cc53f 100644
--- a/src/mesa/pipe/i965simple/brw_clip.c
+++ b/src/gallium/drivers/i965simple/brw_clip.c
diff --git a/src/mesa/pipe/i965simple/brw_clip.h b/src/gallium/drivers/i965simple/brw_clip.h
index a89d08b7910..a89d08b7910 100644
--- a/src/mesa/pipe/i965simple/brw_clip.h
+++ b/src/gallium/drivers/i965simple/brw_clip.h
diff --git a/src/mesa/pipe/i965simple/brw_clip_line.c b/src/gallium/drivers/i965simple/brw_clip_line.c
index 75d9e5fcda2..75d9e5fcda2 100644
--- a/src/mesa/pipe/i965simple/brw_clip_line.c
+++ b/src/gallium/drivers/i965simple/brw_clip_line.c
diff --git a/src/mesa/pipe/i965simple/brw_clip_point.c b/src/gallium/drivers/i965simple/brw_clip_point.c
index 6fce7210d1b..6fce7210d1b 100644
--- a/src/mesa/pipe/i965simple/brw_clip_point.c
+++ b/src/gallium/drivers/i965simple/brw_clip_point.c
diff --git a/src/mesa/pipe/i965simple/brw_clip_state.c b/src/gallium/drivers/i965simple/brw_clip_state.c
index ea5c05a2796..ea5c05a2796 100644
--- a/src/mesa/pipe/i965simple/brw_clip_state.c
+++ b/src/gallium/drivers/i965simple/brw_clip_state.c
diff --git a/src/mesa/pipe/i965simple/brw_clip_tri.c b/src/gallium/drivers/i965simple/brw_clip_tri.c
index c5da7b825e0..c5da7b825e0 100644
--- a/src/mesa/pipe/i965simple/brw_clip_tri.c
+++ b/src/gallium/drivers/i965simple/brw_clip_tri.c
diff --git a/src/mesa/pipe/i965simple/brw_clip_unfilled.c b/src/gallium/drivers/i965simple/brw_clip_unfilled.c
index b774a76dd63..b774a76dd63 100644
--- a/src/mesa/pipe/i965simple/brw_clip_unfilled.c
+++ b/src/gallium/drivers/i965simple/brw_clip_unfilled.c
diff --git a/src/mesa/pipe/i965simple/brw_clip_util.c b/src/gallium/drivers/i965simple/brw_clip_util.c
index 6d58ceafff3..6d58ceafff3 100644
--- a/src/mesa/pipe/i965simple/brw_clip_util.c
+++ b/src/gallium/drivers/i965simple/brw_clip_util.c
diff --git a/src/mesa/pipe/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c
index 5e58701e91c..5e58701e91c 100644
--- a/src/mesa/pipe/i965simple/brw_context.c
+++ b/src/gallium/drivers/i965simple/brw_context.c
diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h
index 65664d853dd..65664d853dd 100644
--- a/src/mesa/pipe/i965simple/brw_context.h
+++ b/src/gallium/drivers/i965simple/brw_context.h
diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/gallium/drivers/i965simple/brw_curbe.c
index 52bbd525c16..52bbd525c16 100644
--- a/src/mesa/pipe/i965simple/brw_curbe.c
+++ b/src/gallium/drivers/i965simple/brw_curbe.c
diff --git a/src/mesa/pipe/i965simple/brw_defines.h b/src/gallium/drivers/i965simple/brw_defines.h
index 9379a397f63..9379a397f63 100644
--- a/src/mesa/pipe/i965simple/brw_defines.h
+++ b/src/gallium/drivers/i965simple/brw_defines.h
diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/gallium/drivers/i965simple/brw_draw.c
index 7598e3dc8af..7598e3dc8af 100644
--- a/src/mesa/pipe/i965simple/brw_draw.c
+++ b/src/gallium/drivers/i965simple/brw_draw.c
diff --git a/src/mesa/pipe/i965simple/brw_draw.h b/src/gallium/drivers/i965simple/brw_draw.h
index 62fe0d5d0ee..62fe0d5d0ee 100644
--- a/src/mesa/pipe/i965simple/brw_draw.h
+++ b/src/gallium/drivers/i965simple/brw_draw.h
diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/gallium/drivers/i965simple/brw_draw_upload.c
index aa85d93866a..aa85d93866a 100644
--- a/src/mesa/pipe/i965simple/brw_draw_upload.c
+++ b/src/gallium/drivers/i965simple/brw_draw_upload.c
diff --git a/src/mesa/pipe/i965simple/brw_eu.c b/src/gallium/drivers/i965simple/brw_eu.c
index e2002d1821f..e2002d1821f 100644
--- a/src/mesa/pipe/i965simple/brw_eu.c
+++ b/src/gallium/drivers/i965simple/brw_eu.c
diff --git a/src/mesa/pipe/i965simple/brw_eu.h b/src/gallium/drivers/i965simple/brw_eu.h
index 23151ae9ed6..23151ae9ed6 100644
--- a/src/mesa/pipe/i965simple/brw_eu.h
+++ b/src/gallium/drivers/i965simple/brw_eu.h
diff --git a/src/mesa/pipe/i965simple/brw_eu_debug.c b/src/gallium/drivers/i965simple/brw_eu_debug.c
index 4a94ddefa6a..4a94ddefa6a 100644
--- a/src/mesa/pipe/i965simple/brw_eu_debug.c
+++ b/src/gallium/drivers/i965simple/brw_eu_debug.c
diff --git a/src/mesa/pipe/i965simple/brw_eu_emit.c b/src/gallium/drivers/i965simple/brw_eu_emit.c
index 400a80b6fba..400a80b6fba 100644
--- a/src/mesa/pipe/i965simple/brw_eu_emit.c
+++ b/src/gallium/drivers/i965simple/brw_eu_emit.c
diff --git a/src/mesa/pipe/i965simple/brw_eu_util.c b/src/gallium/drivers/i965simple/brw_eu_util.c
index 3a65b141f07..3a65b141f07 100644
--- a/src/mesa/pipe/i965simple/brw_eu_util.c
+++ b/src/gallium/drivers/i965simple/brw_eu_util.c
diff --git a/src/mesa/pipe/i965simple/brw_flush.c b/src/gallium/drivers/i965simple/brw_flush.c
index 5216c680cf6..5216c680cf6 100644
--- a/src/mesa/pipe/i965simple/brw_flush.c
+++ b/src/gallium/drivers/i965simple/brw_flush.c
diff --git a/src/mesa/pipe/i965simple/brw_gs.c b/src/gallium/drivers/i965simple/brw_gs.c
index de60868ccca..de60868ccca 100644
--- a/src/mesa/pipe/i965simple/brw_gs.c
+++ b/src/gallium/drivers/i965simple/brw_gs.c
diff --git a/src/mesa/pipe/i965simple/brw_gs.h b/src/gallium/drivers/i965simple/brw_gs.h
index f09141c6aa1..f09141c6aa1 100644
--- a/src/mesa/pipe/i965simple/brw_gs.h
+++ b/src/gallium/drivers/i965simple/brw_gs.h
diff --git a/src/mesa/pipe/i965simple/brw_gs_emit.c b/src/gallium/drivers/i965simple/brw_gs_emit.c
index c3cc90b10f8..c3cc90b10f8 100644
--- a/src/mesa/pipe/i965simple/brw_gs_emit.c
+++ b/src/gallium/drivers/i965simple/brw_gs_emit.c
diff --git a/src/mesa/pipe/i965simple/brw_gs_state.c b/src/gallium/drivers/i965simple/brw_gs_state.c
index 3932e9e9394..3932e9e9394 100644
--- a/src/mesa/pipe/i965simple/brw_gs_state.c
+++ b/src/gallium/drivers/i965simple/brw_gs_state.c
diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/gallium/drivers/i965simple/brw_misc_state.c
index 925049ecc19..925049ecc19 100644
--- a/src/mesa/pipe/i965simple/brw_misc_state.c
+++ b/src/gallium/drivers/i965simple/brw_misc_state.c
diff --git a/src/mesa/pipe/i965simple/brw_reg.h b/src/gallium/drivers/i965simple/brw_reg.h
index 9e885c3b3b7..9e885c3b3b7 100644
--- a/src/mesa/pipe/i965simple/brw_reg.h
+++ b/src/gallium/drivers/i965simple/brw_reg.h
diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c
index 7c83b81c858..7c83b81c858 100644
--- a/src/mesa/pipe/i965simple/brw_sf.c
+++ b/src/gallium/drivers/i965simple/brw_sf.c
diff --git a/src/mesa/pipe/i965simple/brw_sf.h b/src/gallium/drivers/i965simple/brw_sf.h
index b7ada475604..b7ada475604 100644
--- a/src/mesa/pipe/i965simple/brw_sf.h
+++ b/src/gallium/drivers/i965simple/brw_sf.h
diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/gallium/drivers/i965simple/brw_sf_emit.c
index 78d6fa5e9e5..78d6fa5e9e5 100644
--- a/src/mesa/pipe/i965simple/brw_sf_emit.c
+++ b/src/gallium/drivers/i965simple/brw_sf_emit.c
diff --git a/src/mesa/pipe/i965simple/brw_sf_state.c b/src/gallium/drivers/i965simple/brw_sf_state.c
index 9acd3ea61b2..9acd3ea61b2 100644
--- a/src/mesa/pipe/i965simple/brw_sf_state.c
+++ b/src/gallium/drivers/i965simple/brw_sf_state.c
diff --git a/src/mesa/pipe/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c
index 431b45466a6..a762a870fe9 100644
--- a/src/mesa/pipe/i965simple/brw_shader_info.c
+++ b/src/gallium/drivers/i965simple/brw_shader_info.c
@@ -3,7 +3,7 @@
 #include "brw_state.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_parse.h"
 
 
 
diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c
index 95dfce88e4a..f746d1cc57c 100644
--- a/src/mesa/pipe/i965simple/brw_state.c
+++ b/src/gallium/drivers/i965simple/brw_state.c
@@ -33,7 +33,7 @@
 #include "pipe/p_winsys.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
+#include "tgsi/util/tgsi_dump.h"
 
 #include "brw_context.h"
 #include "brw_defines.h"
diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/gallium/drivers/i965simple/brw_state.h
index 258e9a556e2..258e9a556e2 100644
--- a/src/mesa/pipe/i965simple/brw_state.h
+++ b/src/gallium/drivers/i965simple/brw_state.h
diff --git a/src/mesa/pipe/i965simple/brw_state_batch.c b/src/gallium/drivers/i965simple/brw_state_batch.c
index 35db76b5941..35db76b5941 100644
--- a/src/mesa/pipe/i965simple/brw_state_batch.c
+++ b/src/gallium/drivers/i965simple/brw_state_batch.c
diff --git a/src/mesa/pipe/i965simple/brw_state_cache.c b/src/gallium/drivers/i965simple/brw_state_cache.c
index b3a5124461d..b3a5124461d 100644
--- a/src/mesa/pipe/i965simple/brw_state_cache.c
+++ b/src/gallium/drivers/i965simple/brw_state_cache.c
diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/gallium/drivers/i965simple/brw_state_pool.c
index f3174bfe0ae..f3174bfe0ae 100644
--- a/src/mesa/pipe/i965simple/brw_state_pool.c
+++ b/src/gallium/drivers/i965simple/brw_state_pool.c
diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/gallium/drivers/i965simple/brw_state_upload.c
index e727601e1e7..e727601e1e7 100644
--- a/src/mesa/pipe/i965simple/brw_state_upload.c
+++ b/src/gallium/drivers/i965simple/brw_state_upload.c
diff --git a/src/mesa/pipe/i965simple/brw_strings.c b/src/gallium/drivers/i965simple/brw_strings.c
index 29a41ed1e9b..3d9c50961fa 100644
--- a/src/mesa/pipe/i965simple/brw_strings.c
+++ b/src/gallium/drivers/i965simple/brw_strings.c
@@ -59,7 +59,7 @@ static const char *brw_get_name( struct pipe_context *pipe )
       break;
    }
 
-   sprintf(buffer, "pipe/i965 (chipset: %s)", chipset);
+   sprintf(buffer, "i965 (chipset: %s)", chipset);
    return buffer;
 }
 
diff --git a/src/mesa/pipe/i965simple/brw_structs.h b/src/gallium/drivers/i965simple/brw_structs.h
index bbb087e95d6..bbb087e95d6 100644
--- a/src/mesa/pipe/i965simple/brw_structs.h
+++ b/src/gallium/drivers/i965simple/brw_structs.h
diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c
index eb7835836ee..376a42b1a6a 100644
--- a/src/mesa/pipe/i965simple/brw_surface.c
+++ b/src/gallium/drivers/i965simple/brw_surface.c
@@ -32,7 +32,7 @@
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 
 
 /*
@@ -103,6 +103,7 @@ brw_surface_data(struct pipe_context *pipe,
  */
 static void
 brw_surface_copy(struct pipe_context *pipe,
+                 unsigned do_flip,
                  struct pipe_surface *dst,
                  unsigned dstx, unsigned dsty,
                  struct pipe_surface *src,
@@ -118,14 +119,15 @@ brw_surface_copy(struct pipe_context *pipe,
                      dstx, dsty,
                      width, height,
                      pipe_surface_map(src) + src->offset,
-                     src->pitch,
-                     srcx, srcy);
+                     do_flip ? -src->pitch : src->pitch,
+                     srcx, do_flip ? 1 - srcy - height : srcy);
 
       pipe_surface_unmap(src);
       pipe_surface_unmap(dst);
    }
    else {
       brw_copy_blit(brw_context(pipe),
+                    do_flip,
                     dst->cpp,
                     (short) src->pitch, src->buffer, src->offset, FALSE,
                     (short) dst->pitch, dst->buffer, dst->offset, FALSE,
diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c
index 90561f1307d..90561f1307d 100644
--- a/src/mesa/pipe/i965simple/brw_tex_layout.c
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.c
diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.h b/src/gallium/drivers/i965simple/brw_tex_layout.h
index cfd6b1ef3ae..cfd6b1ef3ae 100644
--- a/src/mesa/pipe/i965simple/brw_tex_layout.h
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.h
diff --git a/src/mesa/pipe/i965simple/brw_urb.c b/src/gallium/drivers/i965simple/brw_urb.c
index 101a4367b90..101a4367b90 100644
--- a/src/mesa/pipe/i965simple/brw_urb.c
+++ b/src/gallium/drivers/i965simple/brw_urb.c
diff --git a/src/mesa/pipe/i965simple/brw_util.c b/src/gallium/drivers/i965simple/brw_util.c
index 42391d7c8c5..42391d7c8c5 100644
--- a/src/mesa/pipe/i965simple/brw_util.c
+++ b/src/gallium/drivers/i965simple/brw_util.c
diff --git a/src/mesa/pipe/i965simple/brw_util.h b/src/gallium/drivers/i965simple/brw_util.h
index d60e5934dba..d60e5934dba 100644
--- a/src/mesa/pipe/i965simple/brw_util.h
+++ b/src/gallium/drivers/i965simple/brw_util.h
diff --git a/src/mesa/pipe/i965simple/brw_vs.c b/src/gallium/drivers/i965simple/brw_vs.c
index 738c6346d5e..738c6346d5e 100644
--- a/src/mesa/pipe/i965simple/brw_vs.c
+++ b/src/gallium/drivers/i965simple/brw_vs.c
diff --git a/src/mesa/pipe/i965simple/brw_vs.h b/src/gallium/drivers/i965simple/brw_vs.h
index 0e58f043b02..0e58f043b02 100644
--- a/src/mesa/pipe/i965simple/brw_vs.h
+++ b/src/gallium/drivers/i965simple/brw_vs.h
diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c
index 98915ba1016..05df4860eda 100644
--- a/src/mesa/pipe/i965simple/brw_vs_emit.c
+++ b/src/gallium/drivers/i965simple/brw_vs_emit.c
@@ -33,7 +33,7 @@
 #include "brw_vs.h"
 
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_parse.h"
 
 struct brw_prog_info {
    unsigned num_temps;
diff --git a/src/mesa/pipe/i965simple/brw_vs_state.c b/src/gallium/drivers/i965simple/brw_vs_state.c
index c73469929ce..c73469929ce 100644
--- a/src/mesa/pipe/i965simple/brw_vs_state.c
+++ b/src/gallium/drivers/i965simple/brw_vs_state.c
diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/gallium/drivers/i965simple/brw_winsys.h
index 3523a58614f..3523a58614f 100644
--- a/src/mesa/pipe/i965simple/brw_winsys.h
+++ b/src/gallium/drivers/i965simple/brw_winsys.h
diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c
index 539b1707444..539b1707444 100644
--- a/src/mesa/pipe/i965simple/brw_wm.c
+++ b/src/gallium/drivers/i965simple/brw_wm.c
diff --git a/src/mesa/pipe/i965simple/brw_wm.h b/src/gallium/drivers/i965simple/brw_wm.h
index a1ac0f504a6..a1ac0f504a6 100644
--- a/src/mesa/pipe/i965simple/brw_wm.h
+++ b/src/gallium/drivers/i965simple/brw_wm.h
diff --git a/src/mesa/pipe/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c
index b45a333a2e9..97418a52e7f 100644
--- a/src/mesa/pipe/i965simple/brw_wm_decl.c
+++ b/src/gallium/drivers/i965simple/brw_wm_decl.c
@@ -4,7 +4,7 @@
 #include "brw_wm.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_parse.h"
 
 static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
 {
diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c
index d95645d1085..44f946ea748 100644
--- a/src/mesa/pipe/i965simple/brw_wm_glsl.c
+++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c
@@ -4,7 +4,7 @@
 #include "brw_wm.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_parse.h"
 
 
 
diff --git a/src/mesa/pipe/i965simple/brw_wm_iz.c b/src/gallium/drivers/i965simple/brw_wm_iz.c
index 6c5f25bf39e..6c5f25bf39e 100644
--- a/src/mesa/pipe/i965simple/brw_wm_iz.c
+++ b/src/gallium/drivers/i965simple/brw_wm_iz.c
diff --git a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c
index de42ffc5b1e..de42ffc5b1e 100644
--- a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c
+++ b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c
diff --git a/src/mesa/pipe/i965simple/brw_wm_state.c b/src/gallium/drivers/i965simple/brw_wm_state.c
index 5ccd4888423..5ccd4888423 100644
--- a/src/mesa/pipe/i965simple/brw_wm_state.c
+++ b/src/gallium/drivers/i965simple/brw_wm_state.c
diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c
index d16d919bce9..d16d919bce9 100644
--- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c
+++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c
diff --git a/src/mesa/pipe/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile
index 31438a882e6..5479daf8ea2 100644
--- a/src/mesa/pipe/softpipe/Makefile
+++ b/src/gallium/drivers/softpipe/Makefile
@@ -5,6 +5,9 @@ include $(TOP)/configs/current
 LIBNAME = softpipe
 
 DRIVER_SOURCES = \
+	sp_fs_exec.c \
+	sp_fs_sse.c \
+	sp_fs_llvm.c \
 	sp_clear.c \
 	sp_flush.c \
 	sp_query.c \
@@ -44,7 +47,7 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-include ../Makefile.template
+include ../../Makefile.template
 
 symlinks:
 
diff --git a/src/mesa/pipe/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript
index d581ee8d3ca..d581ee8d3ca 100644
--- a/src/mesa/pipe/softpipe/SConscript
+++ b/src/gallium/drivers/softpipe/SConscript
diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c
index 8d295a30ca6..8d295a30ca6 100644
--- a/src/mesa/pipe/softpipe/sp_clear.c
+++ b/src/gallium/drivers/softpipe/sp_clear.c
diff --git a/src/mesa/pipe/softpipe/sp_clear.h b/src/gallium/drivers/softpipe/sp_clear.h
index a8ed1c4ecc4..a8ed1c4ecc4 100644
--- a/src/mesa/pipe/softpipe/sp_clear.h
+++ b/src/gallium/drivers/softpipe/sp_clear.h
diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index cea6b90104f..5e98f190bbf 100644
--- a/src/mesa/pipe/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -29,7 +29,7 @@
  *    Keith Whitwell <[email protected]>
  */
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_util.h"
diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index aff8c2cc5dc..b70d4fea85a 100644
--- a/src/mesa/pipe/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -34,7 +34,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_vertex.h"
 
 #include "sp_quad.h"
 
@@ -44,8 +44,8 @@ struct softpipe_vbuf_render;
 struct draw_context;
 struct draw_stage;
 struct softpipe_tile_cache;
-struct sp_fragment_shader_state;
-struct sp_vertex_shader_state;
+struct sp_fragment_shader;
+struct sp_vertex_shader;
 
 
 struct softpipe_context {
@@ -59,8 +59,8 @@ struct softpipe_context {
    const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS];
    const struct pipe_depth_stencil_alpha_state   *depth_stencil;
    const struct pipe_rasterizer_state *rasterizer;
-   const struct sp_fragment_shader_state *fs;
-   const struct sp_vertex_shader_state *vs;
+   const struct sp_fragment_shader *fs;
+   const struct sp_vertex_shader *vs;
 
    struct pipe_blend_color blend_color;
    struct pipe_clip_state clip;
diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index 71a303a8b58..2049afda34f 100644
--- a/src/mesa/pipe/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -38,7 +38,7 @@
 #include "sp_context.h"
 #include "sp_state.h"
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 
diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c
index ec6bb4a0dca..2cbd0d7cabe 100644
--- a/src/mesa/pipe/softpipe/sp_flush.c
+++ b/src/gallium/drivers/softpipe/sp_flush.c
@@ -31,6 +31,7 @@
 
 
 #include "pipe/p_defines.h"
+#include "draw/draw_context.h"
 #include "sp_flush.h"
 #include "sp_context.h"
 #include "sp_surface.h"
@@ -50,6 +51,8 @@ softpipe_flush( struct pipe_context *pipe,
    struct softpipe_context *softpipe = softpipe_context(pipe);
    uint i;
 
+   draw_flush(softpipe->draw);
+
    /* - flush the quad pipeline
     * - flush the texture cache
     * - flush the render cache
diff --git a/src/mesa/pipe/softpipe/sp_flush.h b/src/gallium/drivers/softpipe/sp_flush.h
index 34ec617866b..34ec617866b 100644
--- a/src/mesa/pipe/softpipe/sp_flush.h
+++ b/src/gallium/drivers/softpipe/sp_flush.h
diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/gallium/drivers/softpipe/sp_fs.h
index 551b0b97342..4792ace3a33 100644
--- a/src/mesa/pipe/llvm/storagesoa.h
+++ b/src/gallium/drivers/softpipe/sp_fs.h
@@ -1,5 +1,5 @@
 /**************************************************************************
- *
+ * 
  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
  *
@@ -10,11 +10,11 @@
  * 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.
@@ -22,50 +22,33 @@
  * 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 STORAGESOA_H
-#define STORAGESOA_H
-
-#include <vector>
-
-namespace llvm {
-   class BasicBlock;
-   class Constant;
-   class ConstantInt;
-   class LoadInst;
-   class Value;
-   class VectorType;
-}
+/* Authors:  Keith Whitwell <[email protected]>
+ */
 
-class StorageSoa
-{
-public:
-   StorageSoa(llvm::BasicBlock *block,
-              llvm::Value *input,
-              llvm::Value *output,
-              llvm::Value *consts);
+#ifndef SP_FS_H
+#define SP_FS_H
 
-   void addImmediate(float *vec);
+struct sp_fragment_shader *
+softpipe_create_fs_exec(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ);
 
-   llvm::Value  * addrElement(int idx) const;
+struct sp_fragment_shader *
+softpipe_create_fs_sse(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ);
 
-   std::vector<llvm::Value*> inputElement(int idx, int swizzle, llvm::Value *indIdx =0);
-   std::vector<llvm::Value*> constElement(int idx, int swizzle, llvm::Value *indIdx =0);
-   std::vector<llvm::Value*> outputElement(int idx, int swizzle, llvm::Value *indIdx =0);
-   std::vector<llvm::Value*> tempElement(int idx, int swizzle, llvm::Value *indIdx =0);
-   std::vector<llvm::Value*> immediateElement(int idx, int swizzle);
+struct sp_fragment_shader *
+softpipe_create_fs_llvm(struct softpipe_context *softpipe,
+			const struct pipe_shader_state *templ);
 
-   llvm::Value *extractIndex(llvm::Value *vec);
+struct tgsi_interp_coef;
+struct tgsi_exec_vector;
 
-   void storeOutput(int dstIdx, const std::vector<llvm::Value*> &val,
-                         int mask);
-   void storeTemp(int idx, const std::vector<llvm::Value*> &val,
-                  int mask);
-   void storeAddress(int idx, const std::vector<llvm::Value*> &val,
-                     int mask);
+void sp_setup_pos_vector(const struct tgsi_interp_coef *coef,
+			 float x, float y,
+			 struct tgsi_exec_vector *quadpos);
 
-};
 
 #endif
diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c
new file mode 100644
index 00000000000..9ad30a76817
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs_exec.c
@@ -0,0 +1,112 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+#include "sp_headers.h"
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/exec/tgsi_exec.h"
+
+struct sp_exec_fragment_shader {
+   struct sp_fragment_shader base;
+};
+
+
+
+
+static void
+exec_prepare( struct sp_fragment_shader *base,
+	      struct tgsi_exec_machine *machine,
+	      struct tgsi_sampler *samplers )
+{
+   tgsi_exec_machine_bind_shader( machine,
+				  base->shader.tokens,
+				  PIPE_MAX_SAMPLERS,
+				  samplers );
+}
+
+
+
+
+/* TODO: hide the machine struct in here somewhere, remove from this
+ * interface:
+ */
+static unsigned 
+exec_run( struct sp_fragment_shader *base,
+	  struct tgsi_exec_machine *machine,
+	  struct quad_header *quad )
+{
+
+   /* Compute X, Y, Z, W vals for this quad */
+   sp_setup_pos_vector(quad->posCoef, 
+		       (float)quad->x0, (float)quad->y0, 
+		       &machine->QuadPos);
+   
+   return tgsi_exec_machine_run( machine );
+}
+
+
+
+static void 
+exec_delete( struct sp_fragment_shader *base )
+{
+   FREE(base);
+}
+
+
+
+
+
+struct sp_fragment_shader *
+softpipe_create_fs_exec(struct softpipe_context *softpipe,
+			const struct pipe_shader_state *templ)
+{
+   struct sp_exec_fragment_shader *shader;
+
+   /* Decide whether we'll be codegenerating this shader and if so do
+    * that now.
+    */
+
+   shader = CALLOC_STRUCT(sp_exec_fragment_shader);
+   if (!shader)
+      return NULL;
+
+   shader->base.shader = *templ;
+   shader->base.prepare = exec_prepare;
+   shader->base.run = exec_run;
+   shader->base.delete = exec_delete;
+
+   return &shader->base;
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c
new file mode 100644
index 00000000000..22da4714533
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c
@@ -0,0 +1,200 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:
+ *   Zack Rusin
+ */
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/exec/tgsi_sse2.h"
+
+#if 0
+
+struct sp_llvm_fragment_shader {
+   struct sp_fragment_shader base;
+   struct gallivm_prog *llvm_prog;
+};
+
+static void
+shade_quad_llvm(struct quad_stage *qs,
+                struct quad_header *quad)
+{
+   struct quad_shade_stage *qss = quad_shade_stage(qs);
+   struct softpipe_context *softpipe = qs->softpipe;
+   float dests[4][16][4] ALIGN16_ATTRIB;
+   float inputs[4][16][4] ALIGN16_ATTRIB;
+   const float fx = (float) quad->x0;
+   const float fy = (float) quad->y0;
+   struct gallivm_prog *llvm = qss->llvm_prog;
+
+   inputs[0][0][0] = fx;
+   inputs[1][0][0] = fx + 1.0f;
+   inputs[2][0][0] = fx;
+   inputs[3][0][0] = fx + 1.0f;
+
+   inputs[0][0][1] = fy;
+   inputs[1][0][1] = fy;
+   inputs[2][0][1] = fy + 1.0f;
+   inputs[3][0][1] = fy + 1.0f;
+
+
+   gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef);
+
+#if DLLVM
+   debug_printf("MASK = %d\n", quad->mask);
+   for (int i = 0; i < 4; ++i) {
+      for (int j = 0; j < 2; ++j) {
+         debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, 
+                inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]);
+      }
+   }
+#endif
+
+   quad->mask &=
+      gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs,
+                                   softpipe->mapped_constants[PIPE_SHADER_FRAGMENT],
+                                   qss->samplers);
+#if DLLVM
+   debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n",
+          dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], 
+          dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]);
+#endif
+
+   /* store result color */
+   if (qss->colorOutSlot >= 0) {
+      unsigned i;
+      /* XXX need to handle multiple color outputs someday */
+      allvmrt(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot]
+             == TGSI_SEMANTIC_COLOR);
+      for (i = 0; i < QUAD_SIZE; ++i) {
+         quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0];
+         quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1];
+         quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2];
+         quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3];
+      }
+   }
+#if DLLVM
+   for (int i = 0; i < QUAD_SIZE; ++i) {
+      debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot,
+             quad->outputs.color[0][i],
+             quad->outputs.color[1][i],
+             quad->outputs.color[2][i],
+             quad->outputs.color[3][i]);
+   }
+#endif
+
+   /* store result Z */
+   if (qss->depthOutSlot >= 0) {
+      /* output[slot] is new Z */
+      uint i;
+      for (i = 0; i < 4; i++) {
+         quad->outputs.depth[i] = dests[i][0][2];
+      }
+   }
+   else {
+      /* copy input Z (which was interpolated by the executor) to output Z */
+      uint i;
+      for (i = 0; i < 4; i++) {
+         quad->outputs.depth[i] = inputs[i][0][2];
+      }
+   }
+#if DLLVM
+   debug_printf("D [%f, %f, %f, %f] mask = %d\n",
+             quad->outputs.depth[0],
+             quad->outputs.depth[1],
+             quad->outputs.depth[2],
+             quad->outputs.depth[3], quad->mask);
+#endif
+
+   /* shader may cull fragments */
+   if( quad->mask ) {
+      qs->next->run( qs->next, quad );
+   }
+}
+
+
+unsigned 
+run_llvm_fs( struct sp_fragment_shader *base,
+	     struct foo *machine )
+{
+}
+
+
+void 
+delete_llvm_fs( struct sp_fragment_shader *base )
+{
+   FREE(base);
+}
+
+
+struct sp_fragment_shader *
+softpipe_create_fs_llvm(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   struct sp_llvm_fragment_shader *shader = NULL;
+
+   /* LLVM fragment shaders currently disabled:
+    */
+   state = CALLOC_STRUCT(sp_llvm_shader_state);
+   if (!state)
+      return NULL;
+
+   state->llvm_prog = 0;
+
+   if (!gallivm_global_cpu_engine()) {
+      gallivm_cpu_engine_create(state->llvm_prog);
+   }
+   else
+      gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog);
+   
+   if (shader) {
+      shader->base.run = run_llvm_fs;
+      shader->base.delete = delete_llvm_fs;
+   }
+
+   return shader;
+}
+
+
+#else
+
+struct sp_fragment_shader *
+softpipe_create_fs_llvm(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   return NULL;
+}
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c
new file mode 100644
index 00000000000..d90066e0257
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs_sse.c
@@ -0,0 +1,192 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+#include "sp_headers.h"
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/exec/tgsi_exec.h"
+#include "tgsi/exec/tgsi_sse2.h"
+
+
+#if defined(__i386__) || defined(__386__)
+
+#include "x86/rtasm/x86sse.h"
+
+/* Surely this should be defined somewhere in a tgsi header:
+ */
+typedef void (XSTDCALL *codegen_function)(
+   const struct tgsi_exec_vector *input,
+   struct tgsi_exec_vector *output,
+   float (*constant)[4],
+   struct tgsi_exec_vector *temporary,
+   const struct tgsi_interp_coef *coef
+   //, const struct tgsi_exec_vector *quadPos
+ );
+
+
+struct sp_sse_fragment_shader {
+   struct sp_fragment_shader base;
+   struct x86_function             sse2_program;
+   codegen_function func;
+};
+
+
+/**
+ * Compute quad X,Y,Z,W for the four fragments in a quad.
+ *
+ * This should really be part of the compiled shader.
+ */
+void
+sp_setup_pos_vector(const struct tgsi_interp_coef *coef,
+		    float x, float y,
+		    struct tgsi_exec_vector *quadpos)
+{
+   uint chan;
+   /* do X */
+   quadpos->xyzw[0].f[0] = x;
+   quadpos->xyzw[0].f[1] = x + 1;
+   quadpos->xyzw[0].f[2] = x;
+   quadpos->xyzw[0].f[3] = x + 1;
+
+   /* do Y */
+   quadpos->xyzw[1].f[0] = y;
+   quadpos->xyzw[1].f[1] = y;
+   quadpos->xyzw[1].f[2] = y + 1;
+   quadpos->xyzw[1].f[3] = y + 1;
+
+   /* do Z and W for all fragments in the quad */
+   for (chan = 2; chan < 4; chan++) {
+      const float dadx = coef->dadx[chan];
+      const float dady = coef->dady[chan];
+      const float a0 = coef->a0[chan] + dadx * x + dady * y;
+      quadpos->xyzw[chan].f[0] = a0;
+      quadpos->xyzw[chan].f[1] = a0 + dadx;
+      quadpos->xyzw[chan].f[2] = a0 + dady;
+      quadpos->xyzw[chan].f[3] = a0 + dadx + dady;
+   }
+}
+
+
+static void
+fs_sse_prepare( struct sp_fragment_shader *base,
+	      struct tgsi_exec_machine *machine,
+	      struct tgsi_sampler *samplers )
+{
+}
+
+
+/* TODO: codegenerate the whole run function, skip this wrapper.
+ * TODO: break dependency on tgsi_exec_machine struct
+ * TODO: push Position calculation into the generated shader
+ * TODO: process >1 quad at a time
+ */
+static unsigned 
+fs_sse_run( struct sp_fragment_shader *base,
+	 struct tgsi_exec_machine *machine,
+	 struct quad_header *quad )
+{
+   struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
+
+   /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
+   sp_setup_pos_vector(quad->posCoef, 
+		       (float)quad->x0, (float)quad->y0, 
+		       machine->Temps);
+
+   shader->func( machine->Inputs,
+		 machine->Outputs,
+		 machine->Consts,
+		 machine->Temps,
+		 machine->InterpCoefs
+		 //	 , &machine->QuadPos
+      );
+
+   return ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]);
+}
+
+
+static void 
+fs_sse_delete( struct sp_fragment_shader *base )
+{
+   struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
+
+   x86_release_func( &shader->sse2_program );
+   FREE(shader);
+}
+
+
+struct sp_fragment_shader *
+softpipe_create_fs_sse(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   struct sp_sse_fragment_shader *shader;
+
+   if (!softpipe->use_sse)
+      return NULL;
+
+   shader = CALLOC_STRUCT(sp_sse_fragment_shader);
+   if (!shader)
+      return NULL;
+
+   x86_init_func( &shader->sse2_program );
+   
+   if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program )) {
+      FREE(shader);
+      return NULL;
+   }
+
+   shader->func = (codegen_function) x86_get_func( &shader->sse2_program );
+   assert(shader->func);
+
+   shader->base.shader = *templ;
+   shader->base.prepare = fs_sse_prepare;
+   shader->base.run = fs_sse_run;
+   shader->base.delete = fs_sse_delete;
+
+   return &shader->base;
+}
+
+
+#else
+
+/* Maybe put this varient in the header file.
+ */
+struct sp_fragment_shader *
+softpipe_create_fs_sse(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   return NULL;
+}
+
+#endif
diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h
index 0ae31d87961..9cf82221333 100644
--- a/src/mesa/pipe/softpipe/sp_headers.h
+++ b/src/gallium/drivers/softpipe/sp_headers.h
@@ -31,7 +31,7 @@
 #ifndef SP_HEADERS_H
 #define SP_HEADERS_H
 
-#include "pipe/tgsi/exec/tgsi_exec.h"
+#include "tgsi/exec/tgsi_exec.h"
 
 #define PRIM_POINT 1
 #define PRIM_LINE  2
diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c
index 7478b2336b6..d73521ccbe5 100644
--- a/src/mesa/pipe/softpipe/sp_prim_setup.c
+++ b/src/gallium/drivers/softpipe/sp_prim_setup.c
@@ -38,8 +38,8 @@
 #include "sp_quad.h"
 #include "sp_state.h"
 #include "sp_prim_setup.h"
-#include "pipe/draw/draw_private.h"
-#include "pipe/draw/draw_vertex.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vertex.h"
 #include "pipe/p_util.h"
 #include "pipe/p_shader_tokens.h"
 
@@ -499,7 +499,7 @@ setup_fragcoord_coeff(struct setup_stage *setup)
    setup->coef[0].a0[2] = setup->posCoef.a0[2];
    setup->coef[0].dadx[2] = setup->posCoef.dadx[2];
    setup->coef[0].dady[2] = setup->posCoef.dady[2];
-   /*w*/
+   /*W*/
    setup->coef[0].a0[3] = setup->posCoef.a0[3];
    setup->coef[0].dadx[3] = setup->posCoef.dadx[3];
    setup->coef[0].dady[3] = setup->posCoef.dady[3];
@@ -513,8 +513,9 @@ setup_fragcoord_coeff(struct setup_stage *setup)
  */
 static void setup_tri_coefficients( struct setup_stage *setup )
 {
-   const struct softpipe_context *softpipe = setup->softpipe;
+   struct softpipe_context *softpipe = setup->softpipe;
    const struct pipe_shader_state *fs = &softpipe->fs->shader;
+   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
    uint fragSlot;
 
    /* z and w are done by linear interpolation:
@@ -525,10 +526,10 @@ static void setup_tri_coefficients( struct setup_stage *setup )
    /* setup interpolation for all the remaining attributes:
     */
    for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) {
-      const uint vertSlot = softpipe->vertex_info.src_index[fragSlot];
+      const uint vertSlot = vinfo->src_index[fragSlot];
       uint j;
 
-      switch (softpipe->vertex_info.interp_mode[fragSlot]) {
+      switch (vinfo->interp_mode[fragSlot]) {
       case INTERP_CONSTANT:
          for (j = 0; j < NUM_CHANNELS; j++)
             const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
@@ -756,8 +757,9 @@ line_persp_coeff(struct setup_stage *setup,
 static INLINE void
 setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim)
 {
-   const struct softpipe_context *softpipe = setup->softpipe;
+   struct softpipe_context *softpipe = setup->softpipe;
    const struct pipe_shader_state *fs = &setup->softpipe->fs->shader;
+   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
    uint fragSlot;
 
    /* use setup->vmin, vmax to point to vertices */
@@ -779,10 +781,10 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim)
    /* setup interpolation for all the remaining attributes:
     */
    for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) {
-      const uint vertSlot = softpipe->vertex_info.src_index[fragSlot];
+      const uint vertSlot = vinfo->src_index[fragSlot];
       uint j;
 
-      switch (softpipe->vertex_info.interp_mode[fragSlot]) {
+      switch (vinfo->interp_mode[fragSlot]) {
       case INTERP_CONSTANT:
          for (j = 0; j < NUM_CHANNELS; j++)
             const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
@@ -978,6 +980,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim)
    const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth;
    const float x = v0->data[0][0];  /* Note: data[0] is always position */
    const float y = v0->data[0][1];
+   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
    uint fragSlot;
 
    /* For points, all interpolants are constant-valued.
@@ -1003,10 +1006,10 @@ setup_point(struct draw_stage *stage, struct prim_header *prim)
    const_coeff(setup, &setup->posCoef, 0, 3);
 
    for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) {
-      const uint vertSlot = softpipe->vertex_info.src_index[fragSlot];
+      const uint vertSlot = vinfo->src_index[fragSlot];
       uint j;
 
-      switch (softpipe->vertex_info.interp_mode[fragSlot]) {
+      switch (vinfo->interp_mode[fragSlot]) {
       case INTERP_CONSTANT:
          /* fall-through */
       case INTERP_LINEAR:
diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/gallium/drivers/softpipe/sp_prim_setup.h
index f3e8a79dd9e..f3e8a79dd9e 100644
--- a/src/mesa/pipe/softpipe/sp_prim_setup.h
+++ b/src/gallium/drivers/softpipe/sp_prim_setup.h
diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
index c9089e7eb22..69bea8a8f5c 100644
--- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -37,10 +37,11 @@
 
 
 #include "sp_context.h"
+#include "sp_state.h"
 #include "sp_prim_vbuf.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_private.h"
-#include "pipe/draw/draw_vbuf.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
 
 
 #define SP_MAX_VBUF_INDEXES 1024
@@ -73,9 +74,7 @@ static const struct vertex_info *
 sp_vbuf_get_vertex_info(struct vbuf_render *vbr)
 {
    struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
-   /* XXX check for state changes? */
-   assert(!cvbr->softpipe->dirty );
-   return &cvbr->softpipe->vertex_info_vbuf;
+   return softpipe_get_vbuf_vertex_info(cvbr->softpipe);
 }
 
 
diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.h b/src/gallium/drivers/softpipe/sp_prim_vbuf.h
index 1de9cc2a894..1de9cc2a894 100644
--- a/src/mesa/pipe/softpipe/sp_prim_vbuf.h
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.h
diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c
index 6bd468a51cf..6bd468a51cf 100644
--- a/src/mesa/pipe/softpipe/sp_quad.c
+++ b/src/gallium/drivers/softpipe/sp_quad.c
diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h
index f1e0281764f..f1e0281764f 100644
--- a/src/mesa/pipe/softpipe/sp_quad.h
+++ b/src/gallium/drivers/softpipe/sp_quad.h
diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
index 4ffeac35e16..4ffeac35e16 100644
--- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c
+++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c
index 17f3ecd0b83..17f3ecd0b83 100644
--- a/src/mesa/pipe/softpipe/sp_quad_blend.c
+++ b/src/gallium/drivers/softpipe/sp_quad_blend.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c
index 2ae4e22a7de..2ae4e22a7de 100644
--- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c
+++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c
index 1f09d900ca8..1f09d900ca8 100644
--- a/src/mesa/pipe/softpipe/sp_quad_colormask.c
+++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c
index b3d3fae22fd..b3d3fae22fd 100644
--- a/src/mesa/pipe/softpipe/sp_quad_coverage.c
+++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
index a9a0754f279..a9a0754f279 100644
--- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c
+++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c
index 22ea99049fe..22ea99049fe 100644
--- a/src/mesa/pipe/softpipe/sp_quad_earlyz.c
+++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index b5d7dfca1ca..cf1b1eff75c 100644
--- a/src/mesa/pipe/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -39,12 +39,6 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
 
-#include "x86/rtasm/x86sse.h"
-
-#ifdef MESA_LLVM
-#include "pipe/llvm/gallivm.h"
-#endif
-
 #include "sp_context.h"
 #include "sp_state.h"
 #include "sp_headers.h"
@@ -60,9 +54,6 @@ struct quad_shade_stage
    struct tgsi_exec_machine machine;
    struct tgsi_exec_vector *inputs, *outputs;
    int colorOutSlot, depthOutSlot;
-#ifdef MESA_LLVM
-   struct gallivm_prog *llvm_prog;
-#endif
 };
 
 
@@ -74,46 +65,6 @@ quad_shade_stage(struct quad_stage *qs)
 }
 
 
-/**
- * Compute quad X,Y,Z,W for the four fragments in a quad.
- * Note that we only need to "compute" X and Y for the upper-left fragment.
- * We could do less work if we're not depth testing, or there's no
- * perspective-corrected attributes, but that's seldom.
- */
-static void
-setup_pos_vector(const struct tgsi_interp_coef *coef,
-                 float x, float y,
-                 struct tgsi_exec_vector *quadpos)
-{
-   uint chan;
-   /* do X */
-   quadpos->xyzw[0].f[0] = x;
-   /* do Y */
-   quadpos->xyzw[1].f[0] = y;
-   /* do Z and W for all fragments in the quad */
-   for (chan = 2; chan < 4; chan++) {
-      const float dadx = coef->dadx[chan];
-      const float dady = coef->dady[chan];
-      const float a0 = coef->a0[chan] + dadx * x + dady * y;
-      quadpos->xyzw[chan].f[0] = a0;
-      quadpos->xyzw[chan].f[1] = a0 + dadx;
-      quadpos->xyzw[chan].f[2] = a0 + dady;
-      quadpos->xyzw[chan].f[3] = a0 + dadx + dady;
-   }
-}
-
-
-typedef void (XSTDCALL *codegen_function)(
-   const struct tgsi_exec_vector *input,
-   struct tgsi_exec_vector *output,
-   float (*constant)[4],
-   struct tgsi_exec_vector *temporary,
-   const struct tgsi_interp_coef *coef
-#if 0
-   ,const struct tgsi_exec_vector *quadPos
-#endif
- );
-
 
 /**
  * Execute fragment shader for the four fragments in the quad.
@@ -132,30 +83,10 @@ shade_quad(
 
    machine->InterpCoefs = quad->coef;
 
-   /* Compute X, Y, Z, W vals for this quad */
-   setup_pos_vector(quad->posCoef, (float) quad->x0, (float) quad->y0, &machine->QuadPos);
-
    /* run shader */
-#if defined(__i386__) || defined(__386__)
-   if( softpipe->use_sse ) {
-      codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program );
-      func(
-         machine->Inputs,
-         machine->Outputs,
-         machine->Consts,
-         machine->Temps,
-         machine->InterpCoefs
-#if 0
-         ,machine->QuadPos
-#endif
-           );
-      quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]);
-   }
-   else
-#endif
-   {
-      quad->mask &= tgsi_exec_machine_run( machine );
-   }
+   quad->mask &= softpipe->fs->run( softpipe->fs, 
+				    &qss->machine,
+				    quad );
 
    /* store result color */
    if (qss->colorOutSlot >= 0) {
@@ -199,105 +130,6 @@ shade_quad(
    }
 }
 
-#ifdef MESA_LLVM
-#define DLLVM 0
-static void
-shade_quad_llvm(struct quad_stage *qs,
-                struct quad_header *quad)
-{
-   struct quad_shade_stage *qss = quad_shade_stage(qs);
-   struct softpipe_context *softpipe = qs->softpipe;
-   float dests[4][16][4] ALIGN16_ATTRIB;
-   float inputs[4][16][4] ALIGN16_ATTRIB;
-   const float fx = (float) quad->x0;
-   const float fy = (float) quad->y0;
-   struct gallivm_prog *llvm = qss->llvm_prog;
-
-   inputs[0][0][0] = fx;
-   inputs[1][0][0] = fx + 1.0f;
-   inputs[2][0][0] = fx;
-   inputs[3][0][0] = fx + 1.0f;
-
-   inputs[0][0][1] = fy;
-   inputs[1][0][1] = fy;
-   inputs[2][0][1] = fy + 1.0f;
-   inputs[3][0][1] = fy + 1.0f;
-#if DLLVM
-   debug_printf("MASK = %d\n", quad->mask);
-#endif
-   gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef);
-#if DLLVM
-   for (int i = 0; i < 4; ++i) {
-      for (int j = 0; j < 2; ++j) {
-         debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, 
-                inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]);
-      }
-   }
-#endif
-
-   quad->mask &=
-      gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs,
-                                   softpipe->mapped_constants[PIPE_SHADER_FRAGMENT],
-                                   qss->samplers);
-#if DLLVM
-   debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n",
-          dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], 
-          dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]);
-#endif
-
-   /* store result color */
-   if (qss->colorOutSlot >= 0) {
-      unsigned i;
-      /* XXX need to handle multiple color outputs someday */
-      assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot]
-             == TGSI_SEMANTIC_COLOR);
-      for (i = 0; i < QUAD_SIZE; ++i) {
-         quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0];
-         quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1];
-         quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2];
-         quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3];
-      }
-   }
-#if DLLVM
-   for (int i = 0; i < QUAD_SIZE; ++i) {
-      debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot,
-             quad->outputs.color[0][i],
-             quad->outputs.color[1][i],
-             quad->outputs.color[2][i],
-             quad->outputs.color[3][i]);
-   }
-#endif
-
-   /* store result Z */
-   if (qss->depthOutSlot >= 0) {
-      /* output[slot] is new Z */
-      uint i;
-      for (i = 0; i < 4; i++) {
-         quad->outputs.depth[i] = dests[i][0][2];
-      }
-   }
-   else {
-      /* copy input Z (which was interpolated by the executor) to output Z */
-      uint i;
-      for (i = 0; i < 4; i++) {
-         quad->outputs.depth[i] = inputs[i][0][2];
-      }
-   }
-#if DLLVM
-   debug_printf("D [%f, %f, %f, %f] mask = %d\n",
-             quad->outputs.depth[0],
-             quad->outputs.depth[1],
-             quad->outputs.depth[2],
-             quad->outputs.depth[3], quad->mask);
-#endif
-
-   /* shader may cull fragments */
-   if( quad->mask ) {
-      qs->next->run( qs->next, quad );
-   }
-}
-#endif /*MESA_LLVM*/
-
 /**
  * Per-primitive (or per-begin?) setup
  */
@@ -313,15 +145,6 @@ static void shade_begin(struct quad_stage *qs)
       qss->samplers[i].texture = &softpipe->texture[i]->base;
    }
 
-#ifdef MESA_LLVM
-   qss->llvm_prog = softpipe->fs->llvm_prog;
-#endif
-   /* XXX only do this if the fragment shader changes... */
-   tgsi_exec_machine_init(&qss->machine,
-                          softpipe->fs->shader.tokens,
-                          PIPE_MAX_SAMPLERS,
-                          qss->samplers );
-
    /* find output slots for depth, color */
    qss->colorOutSlot = -1;
    qss->depthOutSlot = -1;
@@ -335,6 +158,10 @@ static void shade_begin(struct quad_stage *qs)
          break;
       }
    }
+   
+   softpipe->fs->prepare( softpipe->fs, 
+			  &qss->machine,
+			  qss->samplers );
 
    qs->next->begin(qs->next);
 }
@@ -364,16 +191,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
 
    qss->stage.softpipe = softpipe;
    qss->stage.begin = shade_begin;
-#ifdef MESA_LLVM
-   /* disable until ported to accept
-    * x/y and soa layout
-   qss->stage.run = shade_quad_llvm;
-   */
-   softpipe->use_sse = FALSE;
-   qss->stage.run = shade_quad;
-#else
    qss->stage.run = shade_quad;
-#endif
    qss->stage.destroy = shade_destroy;
 
    /* set TGSI sampler state that's constant */
@@ -384,5 +202,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
       qss->samplers[i].cache = softpipe->tex_cache[i];
    }
 
+   tgsi_exec_machine_init( &qss->machine );
+
    return &qss->stage;
 }
diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c
index 54254df1f11..54254df1f11 100644
--- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c
+++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c
index cfe8f118081..cfe8f118081 100644
--- a/src/mesa/pipe/softpipe/sp_quad_output.c
+++ b/src/gallium/drivers/softpipe/sp_quad_output.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c
index 92a0da00833..92a0da00833 100644
--- a/src/mesa/pipe/softpipe/sp_quad_stencil.c
+++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c
diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c
index 8660432259c..8660432259c 100644
--- a/src/mesa/pipe/softpipe/sp_quad_stipple.c
+++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c
diff --git a/src/mesa/pipe/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c
index 6a8a43aedaf..adf9ccf64c6 100644
--- a/src/mesa/pipe/softpipe/sp_query.c
+++ b/src/gallium/drivers/softpipe/sp_query.c
@@ -29,7 +29,7 @@
  *    Keith Whitwell <[email protected]>
  */
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_util.h"
diff --git a/src/mesa/pipe/softpipe/sp_query.h b/src/gallium/drivers/softpipe/sp_query.h
index 05060a45759..05060a45759 100644
--- a/src/mesa/pipe/softpipe/sp_query.h
+++ b/src/gallium/drivers/softpipe/sp_query.h
diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index af955c1e17b..ef8cf67d4c3 100644
--- a/src/mesa/pipe/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -33,8 +33,6 @@
 
 #include "pipe/p_state.h"
 
-#include "x86/rtasm/x86sse.h"
-
 
 #define SP_NEW_VIEWPORT      0x1
 #define SP_NEW_RASTERIZER    0x2
@@ -53,27 +51,37 @@
 #define SP_NEW_QUERY         0x4000
 
 
+struct tgsi_sampler;
+struct tgsi_interp_coef;
+struct tgsi_exec_machine;
 
-#ifdef MESA_LLVM
-struct gallivm_prog;
-#endif
 
+/** Subclass of pipe_shader_state (though it doesn't really need to be).
+ *
+ * This is starting to look an awful lot like a quad pipeline stage...
+ */
+struct sp_fragment_shader {
+   struct pipe_shader_state   shader;
 
+   void (*prepare)( struct sp_fragment_shader *shader,
+		    struct tgsi_exec_machine *machine,
+		    struct tgsi_sampler *samplers);
 
-/** Subclass of pipe_shader_state */
-struct sp_fragment_shader_state {
-   struct pipe_shader_state   shader;
-#if defined(__i386__) || defined(__386__)
-   struct x86_function        sse2_program;
-#endif
-#ifdef MESA_LLVM
-   struct gallivm_prog *llvm_prog;
-#endif
+   /* Run the shader - this interface will get cleaned up in the
+    * future:
+    */
+   unsigned (*run)( struct sp_fragment_shader *shader,
+		    struct tgsi_exec_machine *machine,
+		    struct quad_header *quad );
+
+
+   void (*delete)( struct sp_fragment_shader * );
 };
 
+struct vertex_info;
 
 /** Subclass of pipe_shader_state */
-struct sp_vertex_shader_state {
+struct sp_vertex_shader {
    struct pipe_shader_state shader;
    struct draw_vertex_shader *draw_data;
 };
@@ -174,4 +182,12 @@ softpipe_map_texture_surfaces(struct softpipe_context *sp);
 void
 softpipe_unmap_texture_surfaces(struct softpipe_context *sp);
 
+
+struct vertex_info *
+softpipe_get_vertex_info(struct softpipe_context *softpipe);
+
+struct vertex_info *
+softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe);
+
+
 #endif
diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/gallium/drivers/softpipe/sp_state_blend.c
index 2d40d6bd8f5..2d40d6bd8f5 100644
--- a/src/mesa/pipe/softpipe/sp_state_blend.c
+++ b/src/gallium/drivers/softpipe/sp_state_blend.c
diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/gallium/drivers/softpipe/sp_state_clip.c
index 08c5f06d05d..c797c0dd3b1 100644
--- a/src/mesa/pipe/softpipe/sp_state_clip.c
+++ b/src/gallium/drivers/softpipe/sp_state_clip.c
@@ -29,7 +29,7 @@
  */
 #include "sp_context.h"
 #include "sp_state.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 void softpipe_set_clip_state( struct pipe_context *pipe,
diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c
new file mode 100644
index 00000000000..9d8fd8b750f
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_derived.c
@@ -0,0 +1,235 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_util.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_private.h"
+#include "sp_context.h"
+#include "sp_state.h"
+
+
+/**
+ * Search vertex program's outputs to find a match for the given
+ * semantic name/index.  Return the index of the output slot.
+ *
+ * Return 0 if not found.  This will cause the fragment program to use
+ * vertex attrib 0 (position) in the cases where the fragment program
+ * attempts to use a missing vertex program output.  This is an undefined
+ * condition that users shouldn't hit anyway.
+ */
+static int
+find_vs_output(const struct pipe_shader_state *vs,
+               uint semantic_name,
+               uint semantic_index)
+{
+   uint i;
+   for (i = 0; i < vs->num_outputs; i++) {
+      if (vs->output_semantic_name[i] == semantic_name &&
+          vs->output_semantic_index[i] == semantic_index)
+         return i;
+   }
+   return 0;
+}
+
+
+/**
+ * Mark the current vertex layout as "invalid".
+ * We'll validate the vertex layout later, when we start to actually
+ * render a point or line or tri.
+ */
+static void
+invalidate_vertex_layout(struct softpipe_context *softpipe)
+{
+   softpipe->vertex_info.num_attribs =  0;
+}
+
+
+/**
+ * The vertex info describes how to convert the post-transformed vertices
+ * (simple float[][4]) used by the 'draw' module into vertices for
+ * rasterization.
+ *
+ * This function validates the vertex layout and returns a pointer to a
+ * vertex_info object.
+ */
+struct vertex_info *
+softpipe_get_vertex_info(struct softpipe_context *softpipe)
+{
+   struct vertex_info *vinfo = &softpipe->vertex_info;
+
+   if (vinfo->num_attribs == 0) {
+      /* compute vertex layout now */
+      const struct pipe_shader_state *vs = &softpipe->vs->shader;
+      const struct pipe_shader_state *fs = &softpipe->fs->shader;
+      const enum interp_mode colorInterp
+         = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
+      uint i;
+
+      if (softpipe->vbuf) {
+         /* if using the post-transform vertex buffer, tell draw_vbuf to
+          * simply emit the whole post-xform vertex as-is:
+          */
+         struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf;
+         vinfo_vbuf->num_attribs = 0;
+         draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0);
+         vinfo_vbuf->size = 4 * vs->num_outputs
+                          + sizeof(struct vertex_header) / 4;
+      }
+
+      /*
+       * Loop over fragment shader inputs, searching for the matching output
+       * from the vertex shader.
+       */
+      vinfo->num_attribs = 0;
+      for (i = 0; i < fs->num_inputs; i++) {
+         int src;
+         switch (fs->input_semantic_name[i]) {
+         case TGSI_SEMANTIC_POSITION:
+            src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src);
+            break;
+
+         case TGSI_SEMANTIC_COLOR:
+            src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, 
+                                 fs->input_semantic_index[i]);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+            break;
+
+         case TGSI_SEMANTIC_FOG:
+            src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+            break;
+
+         case TGSI_SEMANTIC_GENERIC:
+            /* this includes texcoords and varying vars */
+            src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC,
+                                 fs->input_semantic_index[i]);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+            break;
+
+         default:
+            assert(0);
+         }
+      }
+
+      softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0);
+      if (softpipe->psize_slot > 0) {
+         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT,
+                               softpipe->psize_slot);
+      }
+
+      draw_compute_vertex_size(vinfo);
+   }
+
+   return vinfo;
+}
+
+
+/**
+ * Called from vbuf module.
+ *
+ * Note that there's actually two different vertex layouts in softpipe.
+ *
+ * The normal one is computed in softpipe_get_vertex_info() above and is
+ * used by the point/line/tri "setup" code.
+ *
+ * The other one (this one) is only used by the vbuf module (which is
+ * not normally used by default but used in testing).  For the vbuf module,
+ * we basically want to pass-through the draw module's vertex layout as-is.
+ * When the softpipe vbuf code begins drawing, the normal vertex layout
+ * will come into play again.
+ */
+struct vertex_info *
+softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe)
+{
+   (void) softpipe_get_vertex_info(softpipe);
+   return &softpipe->vertex_info_vbuf;
+}
+
+
+/**
+ * Recompute cliprect from scissor bounds, scissor enable and surface size.
+ */
+static void
+compute_cliprect(struct softpipe_context *sp)
+{
+   unsigned surfWidth, surfHeight;
+
+   if (sp->framebuffer.num_cbufs > 0) {
+      surfWidth = sp->framebuffer.cbufs[0]->width;
+      surfHeight = sp->framebuffer.cbufs[0]->height;
+   }
+   else {
+      /* no surface? */
+      surfWidth = sp->scissor.maxx;
+      surfHeight = sp->scissor.maxy;
+   }
+
+   if (sp->rasterizer->scissor) {
+      /* clip to scissor rect */
+      sp->cliprect.minx = MAX2(sp->scissor.minx, 0);
+      sp->cliprect.miny = MAX2(sp->scissor.miny, 0);
+      sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth);
+      sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight);
+   }
+   else {
+      /* clip to surface bounds */
+      sp->cliprect.minx = 0;
+      sp->cliprect.miny = 0;
+      sp->cliprect.maxx = surfWidth;
+      sp->cliprect.maxy = surfHeight;
+   }
+}
+
+
+/* Hopefully this will remain quite simple, otherwise need to pull in
+ * something like the state tracker mechanism.
+ */
+void softpipe_update_derived( struct softpipe_context *softpipe )
+{
+   if (softpipe->dirty & (SP_NEW_RASTERIZER |
+                          SP_NEW_FS |
+                          SP_NEW_VS))
+      invalidate_vertex_layout( softpipe );
+
+   if (softpipe->dirty & (SP_NEW_SCISSOR |
+                          SP_NEW_DEPTH_STENCIL_ALPHA |
+                          SP_NEW_FRAMEBUFFER))
+      compute_cliprect(softpipe);
+
+   if (softpipe->dirty & (SP_NEW_BLEND |
+                          SP_NEW_DEPTH_STENCIL_ALPHA |
+                          SP_NEW_FRAMEBUFFER |
+                          SP_NEW_RASTERIZER |
+                          SP_NEW_FS | 
+			  SP_NEW_QUERY))
+      sp_build_quad_pipeline(softpipe);
+
+   softpipe->dirty = 0;
+}
diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index 0b814fc2847..b0238f81737 100644
--- a/src/mesa/pipe/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -27,16 +27,15 @@
 
 #include "sp_context.h"
 #include "sp_state.h"
+#include "sp_fs.h"
 
 #include "pipe/p_defines.h"
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/draw/draw_context.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/llvm/gallivm.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
-#include "pipe/tgsi/exec/tgsi_sse2.h"
+#include "draw/draw_context.h"
+#include "tgsi/util/tgsi_dump.h"
 
 
 void *
@@ -44,40 +43,22 @@ softpipe_create_fs_state(struct pipe_context *pipe,
                          const struct pipe_shader_state *templ)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
-   struct sp_fragment_shader_state *state;
+   struct sp_fragment_shader *state;
 
-   /* Decide whether we'll be codegenerating this shader and if so do
-    * that now.
-    */
+   if (softpipe->dump_fs) 
+      tgsi_dump(templ->tokens, 0);
 
-   state = CALLOC_STRUCT(sp_fragment_shader_state);
-   if (!state)
-      return NULL;
-
-   state->shader = *templ;
+   state = softpipe_create_fs_llvm( softpipe, templ );
+   if (state)
+      return state;
+   
+   state = softpipe_create_fs_sse( softpipe, templ );
+   if (state)
+      return state;
 
-   if (softpipe->dump_fs) {
-      tgsi_dump(state->shader.tokens, 0);
-   }
-
-#ifdef MESA_LLVM
-   state->llvm_prog = 0;
-
-#if 0
-   if (!gallivm_global_cpu_engine()) {
-      gallivm_cpu_engine_create(state->llvm_prog);
-   }
-   else
-      gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog);
-#endif
-
-#elif defined(__i386__) || defined(__386__)
-   if (softpipe->use_sse) {
-      x86_init_func( &state->sse2_program );
-      tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program );
-   }
-#endif
+   state = softpipe_create_fs_exec( softpipe, templ );
 
+   assert(state);
    return state;
 }
 
@@ -87,7 +68,7 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
-   softpipe->fs = (struct sp_fragment_shader_state *) fs;
+   softpipe->fs = (struct sp_fragment_shader *) fs;
 
    softpipe->dirty |= SP_NEW_FS;
 }
@@ -96,13 +77,9 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 void
 softpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
 {
-   struct sp_fragment_shader_state *state = fs;
-
-#if defined(__i386__) || defined(__386__)
-   x86_release_func( &state->sse2_program );
-#endif
-
-   FREE( state );
+   struct sp_fragment_shader *state = fs;
+   
+   state->delete( state );
 }
 
 
@@ -111,9 +88,9 @@ softpipe_create_vs_state(struct pipe_context *pipe,
                          const struct pipe_shader_state *templ)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
-   struct sp_vertex_shader_state *state;
+   struct sp_vertex_shader *state;
 
-   state = CALLOC_STRUCT(sp_vertex_shader_state);
+   state = CALLOC_STRUCT(sp_vertex_shader);
    if (state == NULL ) {
       return NULL;
    }
@@ -136,7 +113,7 @@ softpipe_bind_vs_state(struct pipe_context *pipe, void *vs)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
-   softpipe->vs = (const struct sp_vertex_shader_state *)vs;
+   softpipe->vs = (const struct sp_vertex_shader *)vs;
 
    draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data);
 
@@ -149,8 +126,8 @@ softpipe_delete_vs_state(struct pipe_context *pipe, void *vs)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
-   struct sp_vertex_shader_state *state =
-      (struct sp_vertex_shader_state *)vs;
+   struct sp_vertex_shader *state =
+      (struct sp_vertex_shader *)vs;
 
    draw_delete_vertex_shader(softpipe->draw, state->draw_data);
    FREE( state );
diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/gallium/drivers/softpipe/sp_state_rasterizer.c
index 53755099dde..98e04352dbc 100644
--- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c
+++ b/src/gallium/drivers/softpipe/sp_state_rasterizer.c
@@ -29,7 +29,7 @@
 #include "pipe/p_util.h"
 #include "sp_context.h"
 #include "sp_state.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 
diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c
index 6a5a643c89e..460adccec4f 100644
--- a/src/mesa/pipe/softpipe/sp_state_sampler.c
+++ b/src/gallium/drivers/softpipe/sp_state_sampler.c
@@ -30,11 +30,15 @@
  */
 
 #include "pipe/p_util.h"
+
+#include "draw/draw_context.h"
+
+#include "sp_context.h"
 #include "sp_context.h"
 #include "sp_state.h"
 #include "sp_texture.h"
 #include "sp_tile_cache.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 
diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c
index e2c6893e9f5..e2c6893e9f5 100644
--- a/src/mesa/pipe/softpipe/sp_state_surface.c
+++ b/src/gallium/drivers/softpipe/sp_state_surface.c
diff --git a/src/mesa/pipe/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c
index 09ff540ccfd..f01a10de3b4 100644
--- a/src/mesa/pipe/softpipe/sp_state_vertex.c
+++ b/src/gallium/drivers/softpipe/sp_state_vertex.c
@@ -33,7 +33,7 @@
 #include "sp_state.h"
 #include "sp_surface.h"
 
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_context.h"
 
 
 void
diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c
index 5978ee48bd4..653449c4f18 100644
--- a/src/mesa/pipe/softpipe/sp_surface.c
+++ b/src/gallium/drivers/softpipe/sp_surface.c
@@ -29,7 +29,7 @@
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 #include "sp_context.h"
 #include "sp_surface.h"
 
@@ -40,6 +40,7 @@
  */
 static void
 sp_surface_copy(struct pipe_context *pipe,
+                unsigned do_flip,
 		struct pipe_surface *dst,
 		unsigned dstx, unsigned dsty,
 		struct pipe_surface *src,
@@ -53,8 +54,8 @@ sp_surface_copy(struct pipe_context *pipe,
                   dstx, dsty,
                   width, height,
                   pipe_surface_map(src),
-                  src->pitch,
-                  srcx, srcy);
+                  do_flip ? -(int) src->pitch : src->pitch,
+                  srcx, do_flip ? 1 - srcy - height : srcy);
 
    pipe_surface_unmap(src);
    pipe_surface_unmap(dst);
diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/gallium/drivers/softpipe/sp_surface.h
index 22de3ba43f6..22de3ba43f6 100644
--- a/src/mesa/pipe/softpipe/sp_surface.h
+++ b/src/gallium/drivers/softpipe/sp_surface.h
diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
index 325bdb86da5..2f82fd6abea 100644
--- a/src/mesa/pipe/softpipe/sp_tex_sample.c
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
@@ -40,7 +40,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_util.h"
-#include "pipe/tgsi/exec/tgsi_exec.h"
+#include "tgsi/exec/tgsi_exec.h"
 
 
 /*
diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/gallium/drivers/softpipe/sp_tex_sample.h
index 404bfd0c365..404bfd0c365 100644
--- a/src/mesa/pipe/softpipe/sp_tex_sample.h
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.h
diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 6de7a9b543b..6de7a9b543b 100644
--- a/src/mesa/pipe/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h
index fa646c0de9f..fa646c0de9f 100644
--- a/src/mesa/pipe/softpipe/sp_texture.h
+++ b/src/gallium/drivers/softpipe/sp_texture.h
diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 1597361b82f..dde3fabc81e 100644
--- a/src/mesa/pipe/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -34,7 +34,7 @@
 
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 #include "sp_context.h"
 #include "sp_surface.h"
 #include "sp_tile_cache.h"
diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h
index 7fd10812863..7fd10812863 100644
--- a/src/mesa/pipe/softpipe/sp_tile_cache.h
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.h
diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h
index d6b379f58c6..d6b379f58c6 100644
--- a/src/mesa/pipe/softpipe/sp_winsys.h
+++ b/src/gallium/drivers/softpipe/sp_winsys.h
diff --git a/src/mesa/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h
index 30cd729c564..30cd729c564 100644
--- a/src/mesa/pipe/p_compiler.h
+++ b/src/gallium/include/pipe/p_compiler.h
diff --git a/src/mesa/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index 92a1cd70c42..39f95695fb6 100644
--- a/src/mesa/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -178,6 +178,7 @@ struct pipe_context {
     */
 
    void (*surface_copy)(struct pipe_context *pipe,
+                        unsigned do_flip,	/*<< flip surface contents vertically */
 			struct pipe_surface *dest,
 			unsigned destx, unsigned desty,
 			struct pipe_surface *src, /* don't make this const - 
diff --git a/src/mesa/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h
index 2a11627b365..2a11627b365 100644
--- a/src/mesa/pipe/p_debug.h
+++ b/src/gallium/include/pipe/p_debug.h
diff --git a/src/mesa/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 0bf53ecb79d..0bf53ecb79d 100644
--- a/src/mesa/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
diff --git a/src/mesa/pipe/p_format.h b/src/gallium/include/pipe/p_format.h
index c9ad3243158..c9ad3243158 100644
--- a/src/mesa/pipe/p_format.h
+++ b/src/gallium/include/pipe/p_format.h
diff --git a/src/mesa/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h
index ebf6ed86bcc..ebf6ed86bcc 100644
--- a/src/mesa/pipe/p_inlines.h
+++ b/src/gallium/include/pipe/p_inlines.h
diff --git a/src/mesa/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h
index 3ce35310f6d..3ce35310f6d 100644
--- a/src/mesa/pipe/p_shader_tokens.h
+++ b/src/gallium/include/pipe/p_shader_tokens.h
diff --git a/src/mesa/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 4d3a6b2f413..1082343e2f5 100644
--- a/src/mesa/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -253,7 +253,8 @@ struct pipe_surface
    unsigned status;              /**< PIPE_SURFACE_STATUS_x */
    unsigned clear_value;         /**< may be temporary */
    unsigned cpp;                 /**< bytes per pixel */
-   unsigned width, height;
+   unsigned width;
+   unsigned height;
    unsigned pitch;               /**< in pixels */
    unsigned offset;              /**< offset from start of buffer, in bytes */
    unsigned refcount;
@@ -272,13 +273,12 @@ struct pipe_texture
    enum pipe_texture_target target; /**< PIPE_TEXTURE_x */
    enum pipe_format format;         /**< PIPE_FORMAT_x */
 
-   unsigned last_level;    /**< Index of last mipmap level present/defined */
-
    unsigned width[PIPE_MAX_TEXTURE_LEVELS];
    unsigned height[PIPE_MAX_TEXTURE_LEVELS];
    unsigned depth[PIPE_MAX_TEXTURE_LEVELS];
-   unsigned cpp;
 
+   unsigned cpp:8;
+   unsigned last_level:8;    /**< Index of last mipmap level present/defined */
    unsigned compressed:1;
 
    /* These are also refcounted:
@@ -294,7 +294,7 @@ struct pipe_texture
  */
 struct pipe_vertex_buffer
 {
-   unsigned pitch:11;    /**< stride to same attrib in next vertex, in bytes */
+   unsigned pitch;    /**< stride to same attrib in next vertex, in bytes */
    unsigned max_index;   /**< number of vertices in this buffer */
    unsigned buffer_offset;  /**< offset to start of data in buffer, in bytes */
    struct pipe_buffer *buffer;  /**< the actual buffer */
@@ -307,13 +307,13 @@ struct pipe_vertex_buffer
 struct pipe_vertex_element
 {
    /** Offset of this attribute, in bytes, from the start of the vertex */
-   unsigned src_offset:11;
+   unsigned src_offset;
 
    /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does
     * this attribute live in?
     */
-   unsigned vertex_buffer_index:5;
-   unsigned nr_components:3;
+   unsigned vertex_buffer_index:8;
+   unsigned nr_components:8;
  
    enum pipe_format src_format; 	   /**< PIPE_FORMAT_* */
 };
diff --git a/src/mesa/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h
index cd432c547ce..cd432c547ce 100644
--- a/src/mesa/pipe/p_thread.h
+++ b/src/gallium/include/pipe/p_thread.h
diff --git a/src/mesa/pipe/p_util.h b/src/gallium/include/pipe/p_util.h
index 469920efee3..d7da2801c92 100644
--- a/src/mesa/pipe/p_util.h
+++ b/src/gallium/include/pipe/p_util.h
@@ -185,12 +185,12 @@ align_free(void *ptr)
 
 
 /**
- * Duplicate of a block of memory
+ * Duplicate a block of memory.
  */
 static INLINE void *
 mem_dup(const void *src, uint size)
 {
-   void *dup = malloc(size);
+   void *dup = MALLOC(size);
    if (dup)
       memcpy(dup, src, size);
    return dup;
@@ -402,7 +402,7 @@ static INLINE int align(int value, int alignment)
 extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch,
                            unsigned dst_x, unsigned dst_y, unsigned width,
                            unsigned height, const ubyte * src,
-                           unsigned src_pitch, unsigned src_x, unsigned src_y);
+                           int src_pitch, unsigned src_x, int src_y);
 
 
 #endif
diff --git a/src/mesa/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h
index 1e81eebd782..1e81eebd782 100644
--- a/src/mesa/pipe/p_winsys.h
+++ b/src/gallium/include/pipe/p_winsys.h
diff --git a/src/gallium/state_trackers/README b/src/gallium/state_trackers/README
new file mode 100644
index 00000000000..28dd27bbd50
--- /dev/null
+++ b/src/gallium/state_trackers/README
@@ -0,0 +1,2 @@
+This directory is a placeholder for incubating state-trackers. Mesa's
+state-tracker is in src/mesa.
diff --git a/src/gallium/winsys/dri/Makefile b/src/gallium/winsys/dri/Makefile
new file mode 100644
index 00000000000..f466ce6c3cc
--- /dev/null
+++ b/src/gallium/winsys/dri/Makefile
@@ -0,0 +1,38 @@
+# src/mesa/drivers/dri/Makefile
+
+TOP = ../../../..
+
+include $(TOP)/configs/current
+
+
+
+default: $(TOP)/$(LIB_DIR) subdirs
+
+
+$(TOP)/$(LIB_DIR):
+	-mkdir $(TOP)/$(LIB_DIR)
+
+
+subdirs:
+	@for dir in $(DRI_DIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+install:
+	@for dir in $(DRI_DIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE) install) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	@for dir in $(DRI_DIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE) clean) ; \
+		fi \
+	done
+	-rm -f common/*.o
diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template
new file mode 100644
index 00000000000..2a261ed6694
--- /dev/null
+++ b/src/gallium/winsys/dri/Makefile.template
@@ -0,0 +1,113 @@
+# -*-makefile-*-
+
+MESA_MODULES = $(TOP)/src/mesa/libmesa.a
+
+COMMON_GALLIUM_SOURCES = \
+        $(TOP)/src/mesa/drivers/dri/common/utils.c \
+        $(TOP)/src/mesa/drivers/dri/common/vblank.c \
+        $(TOP)/src/mesa/drivers/dri/common/dri_util.c \
+        $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c
+
+COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
+        $(TOP)/src/mesa/drivers/common/driverfuncs.c \
+        $(TOP)/src/mesa/drivers/dri/common/texmem.c \
+        $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c
+
+COMMON_BM_SOURCES = \
+	$(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \
+	$(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c
+
+
+ifeq ($(WINDOW_SYSTEM),dri)
+WINOBJ=
+WINLIB=
+INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES)
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+else
+# miniglx
+WINOBJ=
+WINLIB=-L$(MESA)/src/glx/mini
+MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini
+INCLUDES = $(MINIGLX_INCLUDES) \
+	   $(SHARED_INCLUDES) \
+	   $(PCIACCESS_CFLAGS)
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	  $(MINIGLX_SOURCES:.c=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+endif
+
+
+### Include directories
+SHARED_INCLUDES = \
+	-I. \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-Iserver \
+	-I$(TOP)/include \
+	-I$(TOP)/include/GL/internal \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/mesa/glapi \
+	-I$(TOP)/src/mesa/math \
+	-I$(TOP)/src/mesa/transform \
+	-I$(TOP)/src/mesa/shader \
+	-I$(TOP)/src/mesa/swrast \
+	-I$(TOP)/src/mesa/swrast_setup \
+	-I$(TOP)/src/egl/main \
+	-I$(TOP)/src/egl/drivers/dri \
+	$(LIBDRM_CFLAGS)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME)
+
+
+$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
+	$(TOP)/bin/mklib -noprefix -o $@ \
+		$(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES)  $(WINOBJ) $(DRI_LIB_DEPS)
+
+
+$(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME)
+	$(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) 
+
+
+depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \
+		$(ASM_SOURCES) 2> /dev/null
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean:
+	-rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS)
+	-rm -f depend depend.bak
+
+
+install: $(LIBNAME)
+	$(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR)
+	$(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR)
+
+
+include depend
diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/gallium/winsys/dri/intel/Makefile
index 9ae0f013256..40654bb2ac3 100644
--- a/src/mesa/drivers/dri/intel_winsys/Makefile
+++ b/src/gallium/winsys/dri/intel/Makefile
@@ -7,8 +7,8 @@ LIBNAME = i915tex_dri.so
 MINIGLX_SOURCES = server/intel_dri.c
 
 PIPE_DRIVERS = \
-	$(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \
-	$(TOP)/src/mesa/pipe/i915simple/libi915simple.a
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a
 
 DRIVER_SOURCES = \
 	intel_winsys_pipe.c \
@@ -28,11 +28,11 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \
+DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm --atleast-version=2.3.1 \
 				&& echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
 
 include ../Makefile.template
 
-intel_tex_layout.o: ../intel/intel_tex_layout.c
+intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c
 
 symlinks:
diff --git a/src/mesa/drivers/dri/intel_winsys/SConscript b/src/gallium/winsys/dri/intel/SConscript
index a7cc10450e3..a7cc10450e3 100644
--- a/src/mesa/drivers/dri/intel_winsys/SConscript
+++ b/src/gallium/winsys/dri/intel/SConscript
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c
index 49e04d81ec1..49e04d81ec1 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c
+++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h
index 82feafa21f6..82feafa21f6 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h
+++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c b/src/gallium/winsys/dri/intel/intel_batchpool.c
index 33b56817f6a..33b56817f6a 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c
+++ b/src/gallium/winsys/dri/intel/intel_batchpool.c
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h b/src/gallium/winsys/dri/intel/intel_batchpool.h
index f6a95723bcf..f6a95723bcf 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h
+++ b/src/gallium/winsys/dri/intel/intel_batchpool.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c
index c033f2a592c..c033f2a592c 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_context.c
+++ b/src/gallium/winsys/dri/intel/intel_context.c
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h
index b01370c049c..b01370c049c 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_context.h
+++ b/src/gallium/winsys/dri/intel/intel_context.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/gallium/winsys/dri/intel/intel_lock.c
index 70aa7ea5f46..70aa7ea5f46 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c
+++ b/src/gallium/winsys/dri/intel/intel_lock.c
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_reg.h b/src/gallium/winsys/dri/intel/intel_reg.h
index f37c24fda9a..f37c24fda9a 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_reg.h
+++ b/src/gallium/winsys/dri/intel/intel_reg.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c
index 9e31c013a95..9e31c013a95 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c
+++ b/src/gallium/winsys/dri/intel/intel_screen.c
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h
index 3396f9e5643..3396f9e5643 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h
+++ b/src/gallium/winsys/dri/intel/intel_screen.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c
index 454cd71f6c0..56b86d6a636 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c
+++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c
@@ -220,6 +220,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv)
    if (back_surf) {
       st_notify_swapbuffers(intel_fb->stfb);
       intelDisplaySurface(dPriv, back_surf, NULL);
+      st_notify_swapbuffers_complete(intel_fb->stfb);
    }
 }
 
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/gallium/winsys/dri/intel/intel_swapbuffers.h
index 7ae5fd15a5f..7ae5fd15a5f 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h
+++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/gallium/winsys/dri/intel/intel_winsys.h
index ffc40782be7..ffc40782be7 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h
+++ b/src/gallium/winsys/dri/intel/intel_winsys.h
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c
index 1ba6a9e1b25..0ed3890e936 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c
+++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c
@@ -39,7 +39,7 @@
 #include "intel_winsys.h"
 
 #include "pipe/p_util.h"
-#include "pipe/i915simple/i915_winsys.h"
+#include "i915simple/i915_winsys.h"
 
 
 struct intel_i915_winsys {
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c
index 789a386500a..789a386500a 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c
+++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c
diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c
index cec3437831c..9e483bdc9f5 100644
--- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c
+++ b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c
@@ -34,7 +34,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_util.h"
 #include "pipe/p_format.h"
-#include "pipe/softpipe/sp_winsys.h"
+#include "softpipe/sp_winsys.h"
 
 
 struct intel_softpipe_winsys {
diff --git a/src/mesa/drivers/dri/intel_winsys/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h
index d4d58886ce0..d4d58886ce0 100644
--- a/src/mesa/drivers/dri/intel_winsys/server/i830_common.h
+++ b/src/gallium/winsys/dri/intel/server/i830_common.h
diff --git a/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h
index c2a3af8cbf7..c2a3af8cbf7 100644
--- a/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h
+++ b/src/gallium/winsys/dri/intel/server/i830_dri.h
diff --git a/src/mesa/drivers/dri/intel_winsys/server/intel.h b/src/gallium/winsys/dri/intel/server/intel.h
index 6ea72499c1c..6ea72499c1c 100644
--- a/src/mesa/drivers/dri/intel_winsys/server/intel.h
+++ b/src/gallium/winsys/dri/intel/server/intel.h
diff --git a/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c b/src/gallium/winsys/dri/intel/server/intel_dri.c
index e49c4214ad4..e49c4214ad4 100644
--- a/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c
+++ b/src/gallium/winsys/dri/intel/server/intel_dri.c
diff --git a/src/mesa/pipe/xlib/brw_aub.c b/src/gallium/winsys/xlib/brw_aub.c
index 541d50c6e40..541d50c6e40 100644
--- a/src/mesa/pipe/xlib/brw_aub.c
+++ b/src/gallium/winsys/xlib/brw_aub.c
diff --git a/src/mesa/pipe/xlib/brw_aub.h b/src/gallium/winsys/xlib/brw_aub.h
index f5c60c7be28..f5c60c7be28 100644
--- a/src/mesa/pipe/xlib/brw_aub.h
+++ b/src/gallium/winsys/xlib/brw_aub.h
diff --git a/src/mesa/pipe/xlib/fakeglx.c b/src/gallium/winsys/xlib/fakeglx.c
index 902a7550754..902a7550754 100644
--- a/src/mesa/pipe/xlib/fakeglx.c
+++ b/src/gallium/winsys/xlib/fakeglx.c
diff --git a/src/mesa/pipe/xlib/glxapi.c b/src/gallium/winsys/xlib/glxapi.c
index c2ccce6f520..c2ccce6f520 100644
--- a/src/mesa/pipe/xlib/glxapi.c
+++ b/src/gallium/winsys/xlib/glxapi.c
diff --git a/src/mesa/pipe/xlib/glxapi.h b/src/gallium/winsys/xlib/glxapi.h
index 37de81e55ac..37de81e55ac 100644
--- a/src/mesa/pipe/xlib/glxapi.h
+++ b/src/gallium/winsys/xlib/glxapi.h
diff --git a/src/mesa/pipe/xlib/glxheader.h b/src/gallium/winsys/xlib/glxheader.h
index a402191f131..a402191f131 100644
--- a/src/mesa/pipe/xlib/glxheader.h
+++ b/src/gallium/winsys/xlib/glxheader.h
diff --git a/src/mesa/pipe/xlib/realglx.c b/src/gallium/winsys/xlib/realglx.c
index 30adb7465ba..30adb7465ba 100644
--- a/src/mesa/pipe/xlib/realglx.c
+++ b/src/gallium/winsys/xlib/realglx.c
diff --git a/src/mesa/pipe/xlib/realglx.h b/src/gallium/winsys/xlib/realglx.h
index 150129db681..150129db681 100644
--- a/src/mesa/pipe/xlib/realglx.h
+++ b/src/gallium/winsys/xlib/realglx.h
diff --git a/src/mesa/pipe/xlib/xfonts.c b/src/gallium/winsys/xlib/xfonts.c
index d72c600bd1b..d72c600bd1b 100644
--- a/src/mesa/pipe/xlib/xfonts.c
+++ b/src/gallium/winsys/xlib/xfonts.c
diff --git a/src/mesa/pipe/xlib/xfonts.h b/src/gallium/winsys/xlib/xfonts.h
index e36f42f817a..e36f42f817a 100644
--- a/src/mesa/pipe/xlib/xfonts.h
+++ b/src/gallium/winsys/xlib/xfonts.h
diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c
index 03985eab5a6..e5fef1d7a81 100644
--- a/src/mesa/pipe/xlib/xm_api.c
+++ b/src/gallium/winsys/xlib/xm_api.c
@@ -1229,6 +1229,7 @@ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height )
       return;
 
    pipe->surface_copy(pipe,
+                      FALSE,
                       surf_front, x, y,  /* dest */
                       surf_back, x, y,   /* src */
                       width, height);
diff --git a/src/mesa/pipe/xlib/xm_image.c b/src/gallium/winsys/xlib/xm_image.c
index 087b4e4c3a7..087b4e4c3a7 100644
--- a/src/mesa/pipe/xlib/xm_image.c
+++ b/src/gallium/winsys/xlib/xm_image.c
diff --git a/src/mesa/pipe/xlib/xm_image.h b/src/gallium/winsys/xlib/xm_image.h
index 2a5e0f37779..2a5e0f37779 100644
--- a/src/mesa/pipe/xlib/xm_image.h
+++ b/src/gallium/winsys/xlib/xm_image.h
diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c
index c3cd22eea3f..8da596d419e 100644
--- a/src/mesa/pipe/xlib/xm_winsys.c
+++ b/src/gallium/winsys/xlib/xm_winsys.c
@@ -41,11 +41,11 @@
 #include "pipe/p_context.h"
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
-#include "pipe/softpipe/sp_winsys.h"
+#include "softpipe/sp_winsys.h"
 
 #ifdef GALLIUM_CELL
-#include "pipe/cell/ppu/cell_context.h"
-#include "pipe/cell/ppu/cell_winsys.h"
+#include "cell/ppu/cell_context.h"
+#include "cell/ppu/cell_winsys.h"
 #else
 #define TILE_SIZE 32  /* avoid compilation errors */
 #endif
diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c
index bf415702570..dbfd37bda29 100644
--- a/src/mesa/pipe/xlib/xm_winsys_aub.c
+++ b/src/gallium/winsys/xlib/xm_winsys_aub.c
@@ -39,7 +39,7 @@
 #include "pipe/p_winsys.h"
 #include "pipe/p_util.h"
 #include "pipe/p_inlines.h"
-#include "pipe/i965simple/brw_winsys.h"
+#include "i965simple/brw_winsys.h"
 #include "brw_aub.h"
 #include "xm_winsys_aub.h"
 
diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.h b/src/gallium/winsys/xlib/xm_winsys_aub.h
index 7bee1991169..7bee1991169 100644
--- a/src/mesa/pipe/xlib/xm_winsys_aub.h
+++ b/src/gallium/winsys/xlib/xm_winsys_aub.h
diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/gallium/winsys/xlib/xmesaP.h
index fa8d1f14b95..fa8d1f14b95 100644
--- a/src/mesa/pipe/xlib/xmesaP.h
+++ b/src/gallium/winsys/xlib/xmesaP.h
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index 720f1b2e026..c8cb2b592fe 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -12,16 +12,16 @@ GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY)
 
 
 PIPE_LIB = \
-	$(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \
-	$(TOP)/src/mesa/pipe/i965simple/libi965simple.a
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/i965simple/libi965simple.a
 
 ifeq ($(CONFIG_NAME), linux-cell)
-CELL_LIB = $(TOP)/src/mesa/pipe/cell/ppu/libcell.a
-CELL_LIB_SPU = $(TOP)/src/mesa/pipe/cell/spu/g3d_spu.a
+CELL_LIB = $(TOP)/src/gallium/drivers/cell/ppu/libcell.a
+CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a
 endif
 
 ifeq ($(CONFIG_NAME), linux-llvm)
-LLVM_LIB = $(TOP)/src/mesa/pipe/llvm/libgallivm.a
+LLVM_LIB = $(TOP)/src/gallium/auxiliary/llvm/libgallivm.a
 endif
 
 .SUFFIXES : .cpp
@@ -71,7 +71,7 @@ libmesa.a: $(SOLO_OBJECTS)
 	fi
 
 linux-solo: depend subdirs libmesa.a
-	cd drivers/dri ; $(MAKE)
+	cd $(TOP)/src/gallium/winsys/dri ; $(MAKE)
 
 
 #####################################################################
@@ -165,7 +165,6 @@ depend: $(ALL_SOURCES)
 subdirs:
 	@ (cd x86 ; $(MAKE))
 	@ (cd x86-64 ; $(MAKE))
-	(cd pipe ; $(MAKE))
 
 install: default
 	$(INSTALL) -d $(INSTALL_DIR)/include/GL
@@ -178,7 +177,7 @@ install: default
 		$(INSTALL) $(TOP)/$(LIB_DIR)/libOSMesa* $(INSTALL_DIR)/$(LIB_DIR); \
 	fi
 	@if [ "${DRIVER_DIRS}" = "dri" ] ; then \
-		cd drivers/dri ; $(MAKE) install ; \
+		cd $(TOP)/gallium/winsys/dri ; $(MAKE) install ; \
 	fi
 
 ## NOT INSTALLED YET:
@@ -198,7 +197,6 @@ clean:
 	(cd drivers/dri && $(MAKE) clean)
 	(cd x86 && $(MAKE) clean)
 	(cd x86-64 && $(MAKE) clean)
-	(cd pipe ; $(MAKE) clean )
 
 
 include depend
diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c
index 08c98eab486..18b033666f0 100644
--- a/src/mesa/drivers/x11/xm_api.c
+++ b/src/mesa/drivers/x11/xm_api.c
@@ -85,7 +85,7 @@
 
 #include "state_tracker/st_public.h"
 #include "state_tracker/st_context.h"
-#include "pipe/softpipe/sp_context.h"
+#include "softpipe/sp_context.h"
 #include "pipe/p_defines.h"
 
 /**
diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c
index 8ae243ae662..34287effe1c 100644
--- a/src/mesa/drivers/x11/xm_dd.c
+++ b/src/mesa/drivers/x11/xm_dd.c
@@ -53,7 +53,7 @@
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 
-#include "pipe/softpipe/sp_context.h"
+#include "softpipe/sp_context.h"
 #include "state_tracker/st_public.h"
 #include "state_tracker/st_context.h"
 #include "state_tracker/st_draw.h"
diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c
index 5533158ece0..81616b92d96 100644
--- a/src/mesa/drivers/x11/xm_surface.c
+++ b/src/mesa/drivers/x11/xm_surface.c
@@ -45,10 +45,10 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/softpipe/sp_context.h"
-#include "pipe/softpipe/sp_clear.h"
-#include "pipe/softpipe/sp_tile_cache.h"
-#include "pipe/softpipe/sp_surface.h"
+#include "softpipe/sp_context.h"
+#include "softpipe/sp_clear.h"
+#include "softpipe/sp_tile_cache.h"
+#include "softpipe/sp_surface.h"
 #include "state_tracker/st_context.h"
 
 
diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c
index a690df27727..2edc6976933 100644
--- a/src/mesa/drivers/x11/xm_winsys.c
+++ b/src/mesa/drivers/x11/xm_winsys.c
@@ -38,7 +38,7 @@
 #include "main/macros.h"
 
 #include "pipe/p_winsys.h"
-#include "pipe/softpipe/sp_winsys.h"
+#include "softpipe/sp_winsys.h"
 
 
 /**
diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h
index 4709d633942..fd2dfcd79aa 100644
--- a/src/mesa/drivers/x11/xmesaP.h
+++ b/src/mesa/drivers/x11/xmesaP.h
@@ -37,8 +37,8 @@
 #include "xm_image.h"
 #endif
 #include "state_tracker/st_cb_fbo.h"
-#include "pipe/softpipe/sp_context.h"
-#include "pipe/softpipe/sp_surface.h"
+#include "softpipe/sp_context.h"
+#include "softpipe/sp_surface.h"
 
 
 extern _glthread_Mutex _xmesa_lock;
diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c
deleted file mode 100644
index 6e86a919ce4..00000000000
--- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c
+++ /dev/null
@@ -1,393 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
- /*
-  * Authors:
-  *   Keith Whitwell <[email protected]>
-  */
-
-#include <spu_mfcio.h>
-#include <transpose_matrix4x4.h>
-
-#include "pipe/p_util.h"
-#include "pipe/p_state.h"
-#include "pipe/p_shader_tokens.h"
-#include "spu_exec.h"
-#include "spu_vertex_shader.h"
-#include "spu_main.h"
-
-
-#define DRAW_DBG 0
-
-
-static const vec_float4 defaults = { 0.0, 0.0, 0.0, 1.0 };
-
-/**
- * Fetch a float[4] vertex attribute from memory, doing format/type
- * conversion as needed.
- *
- * This is probably needed/dupliocated elsewhere, eg format
- * conversion, texture sampling etc.
- */
-#define FETCH_ATTRIB( NAME, SZ, CVT )			\
-static qword						\
-fetch_##NAME(const void *ptr)				\
-{							\
-   vec_float4 attrib = defaults;			\
-   int i;						\
-							\
-   for (i = 0; i < SZ; i++) {				\
-      attrib = spu_insert(CVT, attrib, i);		\
-   }							\
-   return (qword) attrib;				\
-}
-
-#define CVT_64_FLOAT   (float) ((double *) ptr)[i]
-#define CVT_32_FLOAT   ((float *) ptr)[i]
-
-#define CVT_8_USCALED  (float) ((unsigned char *) ptr)[i]
-#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i]
-#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i]
-
-#define CVT_8_SSCALED  (float) ((char *) ptr)[i]
-#define CVT_16_SSCALED (float) ((short *) ptr)[i]
-#define CVT_32_SSCALED (float) ((int *) ptr)[i]
-
-#define CVT_8_UNORM    (float) ((unsigned char *) ptr)[i] / 255.0f
-#define CVT_16_UNORM   (float) ((unsigned short *) ptr)[i] / 65535.0f
-#define CVT_32_UNORM   (float) ((unsigned int *) ptr)[i] / 4294967295.0f
-
-#define CVT_8_SNORM    (float) ((char *) ptr)[i] / 127.0f
-#define CVT_16_SNORM   (float) ((short *) ptr)[i] / 32767.0f
-#define CVT_32_SNORM   (float) ((int *) ptr)[i] / 2147483647.0f
-
-FETCH_ATTRIB( R64G64B64A64_FLOAT,   4, CVT_64_FLOAT )
-FETCH_ATTRIB( R64G64B64_FLOAT,      3, CVT_64_FLOAT )
-FETCH_ATTRIB( R64G64_FLOAT,         2, CVT_64_FLOAT )
-FETCH_ATTRIB( R64_FLOAT,            1, CVT_64_FLOAT )
-
-FETCH_ATTRIB( R32G32B32A32_FLOAT,   4, CVT_32_FLOAT )
-FETCH_ATTRIB( R32G32B32_FLOAT,      3, CVT_32_FLOAT )
-FETCH_ATTRIB( R32G32_FLOAT,         2, CVT_32_FLOAT )
-FETCH_ATTRIB( R32_FLOAT,            1, CVT_32_FLOAT )
-
-FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED )
-FETCH_ATTRIB( R32G32B32_USCALED,    3, CVT_32_USCALED )
-FETCH_ATTRIB( R32G32_USCALED,       2, CVT_32_USCALED )
-FETCH_ATTRIB( R32_USCALED,          1, CVT_32_USCALED )
-
-FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED )
-FETCH_ATTRIB( R32G32B32_SSCALED,    3, CVT_32_SSCALED )
-FETCH_ATTRIB( R32G32_SSCALED,       2, CVT_32_SSCALED )
-FETCH_ATTRIB( R32_SSCALED,          1, CVT_32_SSCALED )
-
-FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM )
-FETCH_ATTRIB( R32G32B32_UNORM,    3, CVT_32_UNORM )
-FETCH_ATTRIB( R32G32_UNORM,       2, CVT_32_UNORM )
-FETCH_ATTRIB( R32_UNORM,          1, CVT_32_UNORM )
-
-FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM )
-FETCH_ATTRIB( R32G32B32_SNORM,    3, CVT_32_SNORM )
-FETCH_ATTRIB( R32G32_SNORM,       2, CVT_32_SNORM )
-FETCH_ATTRIB( R32_SNORM,          1, CVT_32_SNORM )
-
-FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED )
-FETCH_ATTRIB( R16G16B16_USCALED,    3, CVT_16_USCALED )
-FETCH_ATTRIB( R16G16_USCALED,       2, CVT_16_USCALED )
-FETCH_ATTRIB( R16_USCALED,          1, CVT_16_USCALED )
-
-FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED )
-FETCH_ATTRIB( R16G16B16_SSCALED,    3, CVT_16_SSCALED )
-FETCH_ATTRIB( R16G16_SSCALED,       2, CVT_16_SSCALED )
-FETCH_ATTRIB( R16_SSCALED,          1, CVT_16_SSCALED )
-
-FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM )
-FETCH_ATTRIB( R16G16B16_UNORM,    3, CVT_16_UNORM )
-FETCH_ATTRIB( R16G16_UNORM,       2, CVT_16_UNORM )
-FETCH_ATTRIB( R16_UNORM,          1, CVT_16_UNORM )
-
-FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM )
-FETCH_ATTRIB( R16G16B16_SNORM,    3, CVT_16_SNORM )
-FETCH_ATTRIB( R16G16_SNORM,       2, CVT_16_SNORM )
-FETCH_ATTRIB( R16_SNORM,          1, CVT_16_SNORM )
-
-FETCH_ATTRIB( R8G8B8A8_USCALED,   4, CVT_8_USCALED )
-FETCH_ATTRIB( R8G8B8_USCALED,     3, CVT_8_USCALED )
-FETCH_ATTRIB( R8G8_USCALED,       2, CVT_8_USCALED )
-FETCH_ATTRIB( R8_USCALED,         1, CVT_8_USCALED )
-
-FETCH_ATTRIB( R8G8B8A8_SSCALED,  4, CVT_8_SSCALED )
-FETCH_ATTRIB( R8G8B8_SSCALED,    3, CVT_8_SSCALED )
-FETCH_ATTRIB( R8G8_SSCALED,      2, CVT_8_SSCALED )
-FETCH_ATTRIB( R8_SSCALED,        1, CVT_8_SSCALED )
-
-FETCH_ATTRIB( R8G8B8A8_UNORM,  4, CVT_8_UNORM )
-FETCH_ATTRIB( R8G8B8_UNORM,    3, CVT_8_UNORM )
-FETCH_ATTRIB( R8G8_UNORM,      2, CVT_8_UNORM )
-FETCH_ATTRIB( R8_UNORM,        1, CVT_8_UNORM )
-
-FETCH_ATTRIB( R8G8B8A8_SNORM,  4, CVT_8_SNORM )
-FETCH_ATTRIB( R8G8B8_SNORM,    3, CVT_8_SNORM )
-FETCH_ATTRIB( R8G8_SNORM,      2, CVT_8_SNORM )
-FETCH_ATTRIB( R8_SNORM,        1, CVT_8_SNORM )
-
-FETCH_ATTRIB( A8R8G8B8_UNORM,       4, CVT_8_UNORM )
-//FETCH_ATTRIB( R8G8B8A8_UNORM,       4, CVT_8_UNORM )
-
-
-
-static spu_fetch_func get_fetch_func( enum pipe_format format )
-{
-#if 0
-   {
-      char tmp[80];
-      pf_sprint_name(tmp, format);
-      _mesa_printf("%s: %s\n", __FUNCTION__, tmp);
-   }
-#endif
-
-   switch (format) {
-   case PIPE_FORMAT_R64_FLOAT:
-      return fetch_R64_FLOAT;
-   case PIPE_FORMAT_R64G64_FLOAT:
-      return fetch_R64G64_FLOAT;
-   case PIPE_FORMAT_R64G64B64_FLOAT:
-      return fetch_R64G64B64_FLOAT;
-   case PIPE_FORMAT_R64G64B64A64_FLOAT:
-      return fetch_R64G64B64A64_FLOAT;
-
-   case PIPE_FORMAT_R32_FLOAT:
-      return fetch_R32_FLOAT;
-   case PIPE_FORMAT_R32G32_FLOAT:
-      return fetch_R32G32_FLOAT;
-   case PIPE_FORMAT_R32G32B32_FLOAT:
-      return fetch_R32G32B32_FLOAT;
-   case PIPE_FORMAT_R32G32B32A32_FLOAT:
-      return fetch_R32G32B32A32_FLOAT;
-
-   case PIPE_FORMAT_R32_UNORM:
-      return fetch_R32_UNORM;
-   case PIPE_FORMAT_R32G32_UNORM:
-      return fetch_R32G32_UNORM;
-   case PIPE_FORMAT_R32G32B32_UNORM:
-      return fetch_R32G32B32_UNORM;
-   case PIPE_FORMAT_R32G32B32A32_UNORM:
-      return fetch_R32G32B32A32_UNORM;
-
-   case PIPE_FORMAT_R32_USCALED:
-      return fetch_R32_USCALED;
-   case PIPE_FORMAT_R32G32_USCALED:
-      return fetch_R32G32_USCALED;
-   case PIPE_FORMAT_R32G32B32_USCALED:
-      return fetch_R32G32B32_USCALED;
-   case PIPE_FORMAT_R32G32B32A32_USCALED:
-      return fetch_R32G32B32A32_USCALED;
-
-   case PIPE_FORMAT_R32_SNORM:
-      return fetch_R32_SNORM;
-   case PIPE_FORMAT_R32G32_SNORM:
-      return fetch_R32G32_SNORM;
-   case PIPE_FORMAT_R32G32B32_SNORM:
-      return fetch_R32G32B32_SNORM;
-   case PIPE_FORMAT_R32G32B32A32_SNORM:
-      return fetch_R32G32B32A32_SNORM;
-
-   case PIPE_FORMAT_R32_SSCALED:
-      return fetch_R32_SSCALED;
-   case PIPE_FORMAT_R32G32_SSCALED:
-      return fetch_R32G32_SSCALED;
-   case PIPE_FORMAT_R32G32B32_SSCALED:
-      return fetch_R32G32B32_SSCALED;
-   case PIPE_FORMAT_R32G32B32A32_SSCALED:
-      return fetch_R32G32B32A32_SSCALED;
-
-   case PIPE_FORMAT_R16_UNORM:
-      return fetch_R16_UNORM;
-   case PIPE_FORMAT_R16G16_UNORM:
-      return fetch_R16G16_UNORM;
-   case PIPE_FORMAT_R16G16B16_UNORM:
-      return fetch_R16G16B16_UNORM;
-   case PIPE_FORMAT_R16G16B16A16_UNORM:
-      return fetch_R16G16B16A16_UNORM;
-
-   case PIPE_FORMAT_R16_USCALED:
-      return fetch_R16_USCALED;
-   case PIPE_FORMAT_R16G16_USCALED:
-      return fetch_R16G16_USCALED;
-   case PIPE_FORMAT_R16G16B16_USCALED:
-      return fetch_R16G16B16_USCALED;
-   case PIPE_FORMAT_R16G16B16A16_USCALED:
-      return fetch_R16G16B16A16_USCALED;
-
-   case PIPE_FORMAT_R16_SNORM:
-      return fetch_R16_SNORM;
-   case PIPE_FORMAT_R16G16_SNORM:
-      return fetch_R16G16_SNORM;
-   case PIPE_FORMAT_R16G16B16_SNORM:
-      return fetch_R16G16B16_SNORM;
-   case PIPE_FORMAT_R16G16B16A16_SNORM:
-      return fetch_R16G16B16A16_SNORM;
-
-   case PIPE_FORMAT_R16_SSCALED:
-      return fetch_R16_SSCALED;
-   case PIPE_FORMAT_R16G16_SSCALED:
-      return fetch_R16G16_SSCALED;
-   case PIPE_FORMAT_R16G16B16_SSCALED:
-      return fetch_R16G16B16_SSCALED;
-   case PIPE_FORMAT_R16G16B16A16_SSCALED:
-      return fetch_R16G16B16A16_SSCALED;
-
-   case PIPE_FORMAT_R8_UNORM:
-      return fetch_R8_UNORM;
-   case PIPE_FORMAT_R8G8_UNORM:
-      return fetch_R8G8_UNORM;
-   case PIPE_FORMAT_R8G8B8_UNORM:
-      return fetch_R8G8B8_UNORM;
-   case PIPE_FORMAT_R8G8B8A8_UNORM:
-      return fetch_R8G8B8A8_UNORM;
-
-   case PIPE_FORMAT_R8_USCALED:
-      return fetch_R8_USCALED;
-   case PIPE_FORMAT_R8G8_USCALED:
-      return fetch_R8G8_USCALED;
-   case PIPE_FORMAT_R8G8B8_USCALED:
-      return fetch_R8G8B8_USCALED;
-   case PIPE_FORMAT_R8G8B8A8_USCALED:
-      return fetch_R8G8B8A8_USCALED;
-
-   case PIPE_FORMAT_R8_SNORM:
-      return fetch_R8_SNORM;
-   case PIPE_FORMAT_R8G8_SNORM:
-      return fetch_R8G8_SNORM;
-   case PIPE_FORMAT_R8G8B8_SNORM:
-      return fetch_R8G8B8_SNORM;
-   case PIPE_FORMAT_R8G8B8A8_SNORM:
-      return fetch_R8G8B8A8_SNORM;
-
-   case PIPE_FORMAT_R8_SSCALED:
-      return fetch_R8_SSCALED;
-   case PIPE_FORMAT_R8G8_SSCALED:
-      return fetch_R8G8_SSCALED;
-   case PIPE_FORMAT_R8G8B8_SSCALED:
-      return fetch_R8G8B8_SSCALED;
-   case PIPE_FORMAT_R8G8B8A8_SSCALED:
-      return fetch_R8G8B8A8_SSCALED;
-
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
-      return fetch_A8R8G8B8_UNORM;
-
-   case 0:
-      return NULL;		/* not sure why this is needed */
-
-   default:
-      assert(0);
-      return NULL;
-   }
-}
-
-
-/**
- * Fetch vertex attributes for 'count' vertices.
- */
-static void generic_vertex_fetch(struct spu_vs_context *draw,
-                                 struct spu_exec_machine *machine,
-                                 const unsigned *elts,
-                                 unsigned count)
-{
-   unsigned nr_attrs = draw->vertex_fetch.nr_attrs;
-   unsigned attr;
-
-   assert(count <= 4);
-
-   wait_on_mask(1 << TAG_VERTEX_BUFFER);
-
-#if DRAW_DBG
-   printf("SPU: %s count = %u, nr_attrs = %u\n", 
-          __FUNCTION__, count, nr_attrs);
-#endif
-
-   /* loop over vertex attributes (vertex shader inputs)
-    */
-   for (attr = 0; attr < nr_attrs; attr++) {
-      const unsigned pitch = draw->vertex_fetch.pitch[attr];
-      const uint64_t src = draw->vertex_fetch.src_ptr[attr];
-      const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr];
-      unsigned i;
-      qword p[4];
-
-
-      /* Fetch four attributes for four vertices.  
-       * 
-       * Could fetch directly into AOS format, but this is meant to be
-       * a prototype for an sse implementation, which would have
-       * difficulties doing that.
-       */
-      for (i = 0; i < count; i++) {
-         uint8_t buffer[32] ALIGN16_ATTRIB;
-         const uint64_t addr = src + (elts[i] * pitch);
-         const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32;
-
-#if DRAW_DBG
-         printf("SPU: fetching = 0x%llx\n", addr);
-#endif
-         mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0);
-         wait_on_mask(1 << TAG_VERTEX_BUFFER);
-
-         p[i] = (*fetch)(buffer + (addr & 0x0f));
-      }
-
-      /* Be nice and zero out any missing vertices: 
-       */
-      for (/* empty */; i < 4; i++) 
-          p[i] = si_xor(p[i], p[i]);
-
-      /* Transpose/swizzle into vector-friendly format.  Currently
-       * assuming that all vertex shader inputs are float[4], but this
-       * isn't true -- if the vertex shader only wants tex0.xy, we
-       * could optimize for that.
-       *
-       * To do so fully without codegen would probably require an
-       * excessive number of fetch functions, but we could at least
-       * minimize the transpose step:
-       */
-      _transpose_matrix4x4(&machine->Inputs[attr].xyzw[0].q, p);
-   }
-}
-
-
-void spu_update_vertex_fetch( struct spu_vs_context *draw )
-{
-   unsigned i;
-
-   
-   for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) {
-      draw->vertex_fetch.fetch[i] =
-          get_fetch_func(draw->vertex_fetch.format[i]);
-   }
-
-   draw->vertex_fetch.fetch_func = generic_vertex_fetch;
-}
diff --git a/src/mesa/pipe/draw/Makefile b/src/mesa/pipe/draw/Makefile
deleted file mode 100644
index 451911a3545..00000000000
--- a/src/mesa/pipe/draw/Makefile
+++ /dev/null
@@ -1,2 +0,0 @@
-default:
-	cd .. ; make
diff --git a/src/mesa/pipe/llvm/gallivm_p.h b/src/mesa/pipe/llvm/gallivm_p.h
deleted file mode 100644
index 2c6e5e8f5f2..00000000000
--- a/src/mesa/pipe/llvm/gallivm_p.h
+++ /dev/null
@@ -1,56 +0,0 @@
-#ifndef GALLIVM_P_H
-#define GALLIVM_P_H
-
-#ifdef MESA_LLVM
-
-namespace llvm {
-   class Module;
-}
-
-#if defined __cplusplus
-extern "C" {
-#endif
-
-enum gallivm_shader_type;
-enum gallivm_vector_layout;
-
-struct gallivm_interpolate {
-   int attrib;
-   int chan;
-   int type;
-};
-
-struct gallivm_ir {
-   llvm::Module *module;
-   int id;
-   enum gallivm_shader_type type;
-   enum gallivm_vector_layout layout;
-   int num_components;
-   int   num_consts;
-
-   //FIXME: this might not be enough for some shaders
-   struct gallivm_interpolate interpolators[32*4];
-   int   num_interp;
-};
-
-struct gallivm_prog {
-   llvm::Module *module;
-   void *function;
-
-   int   id;
-   enum gallivm_shader_type type;
-
-   int   num_consts;
-
-   //FIXME: this might not be enough for some shaders
-   struct gallivm_interpolate interpolators[32*4];
-   int   num_interp;
-};
-
-#endif /* MESA_LLVM */
-
-#if defined __cplusplus
-} // extern "C"
-#endif
-
-#endif
diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp
deleted file mode 100644
index 9ac4d8fbc73..00000000000
--- a/src/mesa/pipe/llvm/instructionssoa.cpp
+++ /dev/null
@@ -1,26 +0,0 @@
-#include "instructionssoa.h"
-
-InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func,
-                                 llvm::BasicBlock *block, StorageSoa *storage)
-{
-}
-
-std::vector<llvm::Value*> InstructionsSoa::add(const std::vector<llvm::Value*> in1,
-                                               const std::vector<llvm::Value*> in2)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-std::vector<llvm::Value*> InstructionsSoa::mul(const std::vector<llvm::Value*> in1,
-                                               const std::vector<llvm::Value*> in2)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-void InstructionsSoa::end()
-{
-}
diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp
deleted file mode 100644
index 90a25a440ab..00000000000
--- a/src/mesa/pipe/llvm/llvm_base_shader.cpp
+++ /dev/null
@@ -1,833 +0,0 @@
-// Generated by llvm2cpp - DO NOT MODIFY!
-
-
-Module* createBaseShader() {
-  // Module Construction
-  Module* mod = new Module("Shader");
-  mod->setDataLayout("e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128");
-  mod->setTargetTriple("i686-pc-linux-gnu");
-  
-  // Type Definitions
-  std::vector<const Type*>StructTy_struct_ShaderInput_fields;
-  VectorType* VectorTy_1 = VectorType::get(Type::FloatTy, 4);
-  
-  PointerType* PointerTy_0 = PointerType::get(VectorTy_1, 0);
-  
-  StructTy_struct_ShaderInput_fields.push_back(PointerTy_0);
-  StructTy_struct_ShaderInput_fields.push_back(PointerTy_0);
-  StructTy_struct_ShaderInput_fields.push_back(PointerTy_0);
-  StructTy_struct_ShaderInput_fields.push_back(PointerTy_0);
-  StructTy_struct_ShaderInput_fields.push_back(IntegerType::get(32));
-  StructType* StructTy_struct_ShaderInput = StructType::get(StructTy_struct_ShaderInput_fields, /*isPacked=*/false);
-  mod->addTypeName("struct.ShaderInput", StructTy_struct_ShaderInput);
-  
-  OpaqueType* OpaqueTy_struct_pipe_sampler_state = OpaqueType::get();
-  mod->addTypeName("struct.pipe_sampler_state", OpaqueTy_struct_pipe_sampler_state);
-  
-  OpaqueType* OpaqueTy_struct_softpipe_tile_cache = OpaqueType::get();
-  mod->addTypeName("struct.softpipe_tile_cache", OpaqueTy_struct_softpipe_tile_cache);
-  
-  std::vector<const Type*>StructTy_struct_tgsi_sampler_fields;
-  PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_sampler_state, 0);
-  
-  StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2);
-  std::vector<const Type*>FuncTy_4_args;
-  PATypeHolder StructTy_struct_tgsi_sampler_fwd = OpaqueType::get();
-  PointerType* PointerTy_5 = PointerType::get(StructTy_struct_tgsi_sampler_fwd, 0);
-  
-  FuncTy_4_args.push_back(PointerTy_5);
-  PointerType* PointerTy_6 = PointerType::get(Type::FloatTy, 0);
-  
-  FuncTy_4_args.push_back(PointerTy_6);
-  FuncTy_4_args.push_back(PointerTy_6);
-  FuncTy_4_args.push_back(PointerTy_6);
-  FuncTy_4_args.push_back(Type::FloatTy);
-  ArrayType* ArrayTy_8 = ArrayType::get(Type::FloatTy, 4);
-  
-  PointerType* PointerTy_7 = PointerType::get(ArrayTy_8, 0);
-  
-  FuncTy_4_args.push_back(PointerTy_7);
-  FunctionType* FuncTy_4 = FunctionType::get(
-    /*Result=*/Type::VoidTy,
-    /*Params=*/FuncTy_4_args,
-    /*isVarArg=*/false);
-  
-  PointerType* PointerTy_3 = PointerType::get(FuncTy_4, 0);
-  
-  StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3);
-  PointerType* PointerTy_9 = PointerType::get(IntegerType::get(8), 0);
-  
-  StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_9);
-  PointerType* PointerTy_10 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache, 0);
-  
-  StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_10);
-  StructType* StructTy_struct_tgsi_sampler = StructType::get(StructTy_struct_tgsi_sampler_fields, /*isPacked=*/false);
-  mod->addTypeName("struct.tgsi_sampler", StructTy_struct_tgsi_sampler);
-  cast<OpaqueType>(StructTy_struct_tgsi_sampler_fwd.get())->refineAbstractTypeTo(StructTy_struct_tgsi_sampler);
-  StructTy_struct_tgsi_sampler = cast<StructType>(StructTy_struct_tgsi_sampler_fwd.get());
-  
-  
-  std::vector<const Type*>FuncTy_11_args;
-  ArrayType* ArrayTy_13 = ArrayType::get(VectorTy_1, 16);
-  
-  PointerType* PointerTy_12 = PointerType::get(ArrayTy_13, 0);
-  
-  FuncTy_11_args.push_back(PointerTy_12);
-  ArrayType* ArrayTy_15 = ArrayType::get(ArrayTy_8, 16);
-  
-  PointerType* PointerTy_14 = PointerType::get(ArrayTy_15, 0);
-  
-  FuncTy_11_args.push_back(PointerTy_14);
-  FuncTy_11_args.push_back(IntegerType::get(32));
-  FuncTy_11_args.push_back(IntegerType::get(32));
-  FunctionType* FuncTy_11 = FunctionType::get(
-    /*Result=*/Type::VoidTy,
-    /*Params=*/FuncTy_11_args,
-    /*isVarArg=*/false);
-  
-  std::vector<const Type*>FuncTy_16_args;
-  FuncTy_16_args.push_back(PointerTy_0);
-  FuncTy_16_args.push_back(PointerTy_7);
-  FuncTy_16_args.push_back(IntegerType::get(32));
-  FunctionType* FuncTy_16 = FunctionType::get(
-    /*Result=*/Type::VoidTy,
-    /*Params=*/FuncTy_16_args,
-    /*isVarArg=*/false);
-  
-  std::vector<const Type*>FuncTy_17_args;
-  FuncTy_17_args.push_back(PointerTy_7);
-  FuncTy_17_args.push_back(PointerTy_0);
-  FuncTy_17_args.push_back(IntegerType::get(32));
-  FunctionType* FuncTy_17 = FunctionType::get(
-    /*Result=*/Type::VoidTy,
-    /*Params=*/FuncTy_17_args,
-    /*isVarArg=*/false);
-  
-  std::vector<const Type*>FuncTy_18_args;
-  FuncTy_18_args.push_back(PointerTy_9);
-  FuncTy_18_args.push_back(PointerTy_9);
-  FuncTy_18_args.push_back(PointerTy_7);
-  FuncTy_18_args.push_back(IntegerType::get(32));
-  FuncTy_18_args.push_back(IntegerType::get(32));
-  FuncTy_18_args.push_back(IntegerType::get(32));
-  FuncTy_18_args.push_back(IntegerType::get(32));
-  FunctionType* FuncTy_18 = FunctionType::get(
-    /*Result=*/Type::VoidTy,
-    /*Params=*/FuncTy_18_args,
-    /*isVarArg=*/false);
-  
-  ArrayType* ArrayTy_20 = ArrayType::get(VectorTy_1, 32);
-  
-  PointerType* PointerTy_19 = PointerType::get(ArrayTy_20, 0);
-  
-  ArrayType* ArrayTy_22 = ArrayType::get(VectorTy_1, 128);
-  
-  PointerType* PointerTy_21 = PointerType::get(ArrayTy_22, 0);
-  
-  PointerType* PointerTy_23 = PointerType::get(StructTy_struct_ShaderInput, 0);
-  
-  PointerType* PointerTy_24 = PointerType::get(PointerTy_0, 0);
-  
-  std::vector<const Type*>FuncTy_26_args;
-  FuncTy_26_args.push_back(PointerTy_23);
-  FunctionType* FuncTy_26 = FunctionType::get(
-    /*Result=*/Type::VoidTy,
-    /*Params=*/FuncTy_26_args,
-    /*isVarArg=*/false);
-  
-  PointerType* PointerTy_25 = PointerType::get(FuncTy_26, 0);
-  
-  std::vector<const Type*>FuncTy_27_args;
-  FuncTy_27_args.push_back(Type::FloatTy);
-  FuncTy_27_args.push_back(Type::FloatTy);
-  FuncTy_27_args.push_back(PointerTy_12);
-  FuncTy_27_args.push_back(PointerTy_12);
-  FuncTy_27_args.push_back(IntegerType::get(32));
-  FuncTy_27_args.push_back(PointerTy_7);
-  FuncTy_27_args.push_back(IntegerType::get(32));
-  FuncTy_27_args.push_back(PointerTy_5);
-  FunctionType* FuncTy_27 = FunctionType::get(
-    /*Result=*/IntegerType::get(32),
-    /*Params=*/FuncTy_27_args,
-    /*isVarArg=*/false);
-  
-  PointerType* PointerTy_28 = PointerType::get(IntegerType::get(32), 0);
-  
-  
-  // Function Declarations
-  
-  Function* func_from_array = new Function(
-    /*Type=*/FuncTy_11,
-    /*Linkage=*/GlobalValue::ExternalLinkage,
-    /*Name=*/"from_array", mod); 
-  func_from_array->setCallingConv(CallingConv::C);
-  const ParamAttrsList *func_from_array_PAL = 0;
-  {
-    ParamAttrsVector Attrs;
-    ParamAttrsWithIndex PAWI;
-    PAWI.index = 0; PAWI.attrs = 0  | ParamAttr::NoUnwind;
-    Attrs.push_back(PAWI);
-    func_from_array_PAL = ParamAttrsList::get(Attrs);
-    
-  }
-  func_from_array->setParamAttrs(func_from_array_PAL);
-  
-  Function* func_from_consts = new Function(
-    /*Type=*/FuncTy_16,
-    /*Linkage=*/GlobalValue::ExternalLinkage,
-    /*Name=*/"from_consts", mod); 
-  func_from_consts->setCallingConv(CallingConv::C);
-  const ParamAttrsList *func_from_consts_PAL = 0;
-  {
-    ParamAttrsVector Attrs;
-    ParamAttrsWithIndex PAWI;
-    PAWI.index = 0; PAWI.attrs = 0  | ParamAttr::NoUnwind;
-    Attrs.push_back(PAWI);
-    func_from_consts_PAL = ParamAttrsList::get(Attrs);
-    
-  }
-  func_from_consts->setParamAttrs(func_from_consts_PAL);
-  
-  Function* func_to_array = new Function(
-    /*Type=*/FuncTy_17,
-    /*Linkage=*/GlobalValue::ExternalLinkage,
-    /*Name=*/"to_array", mod); 
-  func_to_array->setCallingConv(CallingConv::C);
-  const ParamAttrsList *func_to_array_PAL = 0;
-  {
-    ParamAttrsVector Attrs;
-    ParamAttrsWithIndex PAWI;
-    PAWI.index = 0; PAWI.attrs = 0  | ParamAttr::NoUnwind;
-    Attrs.push_back(PAWI);
-    func_to_array_PAL = ParamAttrsList::get(Attrs);
-    
-  }
-  func_to_array->setParamAttrs(func_to_array_PAL);
-  
-  Function* func_run_vertex_shader = new Function(
-    /*Type=*/FuncTy_18,
-    /*Linkage=*/GlobalValue::ExternalLinkage,
-    /*Name=*/"run_vertex_shader", mod); 
-  func_run_vertex_shader->setCallingConv(CallingConv::C);
-  const ParamAttrsList *func_run_vertex_shader_PAL = 0;
-  func_run_vertex_shader->setParamAttrs(func_run_vertex_shader_PAL);
-  
-  Function* func_execute_shader = new Function(
-    /*Type=*/FuncTy_26,
-    /*Linkage=*/GlobalValue::ExternalLinkage,
-    /*Name=*/"execute_shader", mod); // (external, no body)
-  func_execute_shader->setCallingConv(CallingConv::C);
-  const ParamAttrsList *func_execute_shader_PAL = 0;
-  func_execute_shader->setParamAttrs(func_execute_shader_PAL);
-  
-  Function* func_run_fragment_shader = new Function(
-    /*Type=*/FuncTy_27,
-    /*Linkage=*/GlobalValue::ExternalLinkage,
-    /*Name=*/"run_fragment_shader", mod); 
-  func_run_fragment_shader->setCallingConv(CallingConv::C);
-  const ParamAttrsList *func_run_fragment_shader_PAL = 0;
-  func_run_fragment_shader->setParamAttrs(func_run_fragment_shader_PAL);
-  
-  // Global Variable Declarations
-
-  
-  // Constant Definitions
-  Constant* const_int32_29 = Constant::getNullValue(IntegerType::get(32));
-  ConstantInt* const_int32_30 = ConstantInt::get(APInt(32,  "-1", 10));
-  ConstantInt* const_int32_31 = ConstantInt::get(APInt(32,  "1", 10));
-  UndefValue* const_packed_32 = UndefValue::get(VectorTy_1);
-  ConstantInt* const_int32_33 = ConstantInt::get(APInt(32,  "2", 10));
-  ConstantInt* const_int32_34 = ConstantInt::get(APInt(32,  "3", 10));
-  ConstantInt* const_int32_35 = ConstantInt::get(APInt(32,  "4", 10));
-  
-  // Global Variable Definitions
-  
-  // Function Definitions
-  
-  // Function: from_array (func_from_array)
-  {
-    Function::arg_iterator args = func_from_array->arg_begin();
-    Value* ptr_res = args++;
-    ptr_res->setName("res");
-    Value* ptr_ainputs = args++;
-    ptr_ainputs->setName("ainputs");
-    Value* int32_count = args++;
-    int32_count->setName("count");
-    Value* int32_num_attribs = args++;
-    int32_num_attribs->setName("num_attribs");
-    
-    BasicBlock* label_entry = new BasicBlock("entry",func_from_array,0);
-    BasicBlock* label_forcond2_preheader_split = new BasicBlock("forcond2.preheader.split",func_from_array,0);
-    BasicBlock* label_forcond2 = new BasicBlock("forcond2",func_from_array,0);
-    BasicBlock* label_forbody6 = new BasicBlock("forbody6",func_from_array,0);
-    BasicBlock* label_forinc57 = new BasicBlock("forinc57",func_from_array,0);
-    BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0);
-    
-    // Block entry (label_entry)
-    ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_29, "cmp", label_entry);
-    ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_29, "cmp5", label_entry);
-    BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry);
-    new BranchInst(label_forcond2_preheader_split, label_afterfor60, int1_bothcond, label_entry);
-    
-    // Block forcond2.preheader.split (label_forcond2_preheader_split)
-    BinaryOperator* int32_tmp21 = BinaryOperator::create(Instruction::Add, int32_count, const_int32_30, "tmp21", label_forcond2_preheader_split);
-    ICmpInst* int1_tmp22 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp21, const_int32_29, "tmp22", label_forcond2_preheader_split);
-    SelectInst* int32_tmp25 = new SelectInst(int1_tmp22, const_int32_31, int32_count, "tmp25", label_forcond2_preheader_split);
-    new BranchInst(label_forcond2, label_forcond2_preheader_split);
-    
-    // Block forcond2 (label_forcond2)
-    Argument* fwdref_38 = new Argument(IntegerType::get(32));
-    PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forcond2);
-    int32_i_0_reg2mem_0->reserveOperandSpace(2);
-    int32_i_0_reg2mem_0->addIncoming(const_int32_29, label_forcond2_preheader_split);
-    int32_i_0_reg2mem_0->addIncoming(fwdref_38, label_forinc57);
-    
-    Argument* fwdref_39 = new Argument(VectorTy_1);
-    PHINode* packed_vec_1_reg2mem_0 = new PHINode(VectorTy_1, "vec.1.reg2mem.0", label_forcond2);
-    packed_vec_1_reg2mem_0->reserveOperandSpace(2);
-    packed_vec_1_reg2mem_0->addIncoming(const_packed_32, label_forcond2_preheader_split);
-    packed_vec_1_reg2mem_0->addIncoming(fwdref_39, label_forinc57);
-    
-    BinaryOperator* int32_tmp = BinaryOperator::create(Instruction::Add, int32_num_attribs, const_int32_30, "tmp", label_forcond2);
-    ICmpInst* int1_tmp18 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp, const_int32_29, "tmp18", label_forcond2);
-    SelectInst* int32_tmp19 = new SelectInst(int1_tmp18, const_int32_31, int32_num_attribs, "tmp19", label_forcond2);
-    new BranchInst(label_forbody6, label_forcond2);
-    
-    // Block forbody6 (label_forbody6)
-    Argument* fwdref_41 = new Argument(IntegerType::get(32));
-    PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6);
-    int32_j_0_reg2mem_0->reserveOperandSpace(2);
-    int32_j_0_reg2mem_0->addIncoming(const_int32_29, label_forcond2);
-    int32_j_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6);
-    
-    PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody6);
-    packed_vec_0_reg2mem_0->reserveOperandSpace(2);
-    packed_vec_0_reg2mem_0->addIncoming(packed_vec_1_reg2mem_0, label_forcond2);
-    packed_vec_0_reg2mem_0->addIncoming(fwdref_39, label_forbody6);
-    
-    std::vector<Value*> ptr_arraydecay11_indices;
-    ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0);
-    ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0);
-    ptr_arraydecay11_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6);
-    LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6);
-    InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_29, "tmp15", label_forbody6);
-    std::vector<Value*> ptr_arrayidx23_indices;
-    ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0);
-    ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0);
-    ptr_arrayidx23_indices.push_back(const_int32_31);
-    Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6);
-    LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6);
-    InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_31, "tmp26", label_forbody6);
-    std::vector<Value*> ptr_arrayidx34_indices;
-    ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0);
-    ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0);
-    ptr_arrayidx34_indices.push_back(const_int32_33);
-    Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6);
-    LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6);
-    InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_33, "tmp37", label_forbody6);
-    std::vector<Value*> ptr_arrayidx45_indices;
-    ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0);
-    ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0);
-    ptr_arrayidx45_indices.push_back(const_int32_34);
-    Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6);
-    LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6);
-    InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_34, "tmp48", label_forbody6);
-    std::vector<Value*> ptr_arrayidx54_indices;
-    ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0);
-    ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0);
-    Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6);
-    StoreInst* void_42 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6);
-    BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_31, "indvar.next", label_forbody6);
-    ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_tmp19, "exitcond", label_forbody6);
-    new BranchInst(label_forinc57, label_forbody6, int1_exitcond, label_forbody6);
-    
-    // Block forinc57 (label_forinc57)
-    BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_31, "indvar.next20", label_forinc57);
-    ICmpInst* int1_exitcond26 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, int32_tmp25, "exitcond26", label_forinc57);
-    new BranchInst(label_afterfor60, label_forcond2, int1_exitcond26, label_forinc57);
-    
-    // Block afterfor60 (label_afterfor60)
-    new ReturnInst(label_afterfor60);
-    
-    // Resolve Forward References
-    fwdref_39->replaceAllUsesWith(packed_tmp48); delete fwdref_39;
-    fwdref_41->replaceAllUsesWith(int32_indvar_next); delete fwdref_41;
-    fwdref_38->replaceAllUsesWith(int32_indvar_next20); delete fwdref_38;
-    
-  }
-  
-  // Function: from_consts (func_from_consts)
-  {
-    Function::arg_iterator args = func_from_consts->arg_begin();
-    Value* ptr_res_46 = args++;
-    ptr_res_46->setName("res");
-    Value* ptr_ainputs_47 = args++;
-    ptr_ainputs_47->setName("ainputs");
-    Value* int32_count_48 = args++;
-    int32_count_48->setName("count");
-    
-    BasicBlock* label_entry_49 = new BasicBlock("entry",func_from_consts,0);
-    BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_from_consts,0);
-    BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0);
-    BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0);
-    
-    // Block entry (label_entry_49)
-    ICmpInst* int1_cmp_50 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_48, const_int32_29, "cmp", label_entry_49);
-    new BranchInst(label_forbody_preheader, label_afterfor, int1_cmp_50, label_entry_49);
-    
-    // Block forbody.preheader (label_forbody_preheader)
-    BinaryOperator* int32_tmp_52 = BinaryOperator::create(Instruction::Add, int32_count_48, const_int32_30, "tmp", label_forbody_preheader);
-    ICmpInst* int1_tmp9 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_52, const_int32_29, "tmp9", label_forbody_preheader);
-    SelectInst* int32_tmp10 = new SelectInst(int1_tmp9, const_int32_31, int32_count_48, "tmp10", label_forbody_preheader);
-    new BranchInst(label_forbody, label_forbody_preheader);
-    
-    // Block forbody (label_forbody)
-    Argument* fwdref_55 = new Argument(IntegerType::get(32));
-    PHINode* int32_i_0_reg2mem_0_54 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody);
-    int32_i_0_reg2mem_0_54->reserveOperandSpace(2);
-    int32_i_0_reg2mem_0_54->addIncoming(const_int32_29, label_forbody_preheader);
-    int32_i_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody);
-    
-    Argument* fwdref_57 = new Argument(VectorTy_1);
-    PHINode* packed_vec_0_reg2mem_0_56 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody);
-    packed_vec_0_reg2mem_0_56->reserveOperandSpace(2);
-    packed_vec_0_reg2mem_0_56->addIncoming(const_packed_32, label_forbody_preheader);
-    packed_vec_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody);
-    
-    std::vector<Value*> ptr_arraydecay_indices;
-    ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_54);
-    ptr_arraydecay_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_47, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody);
-    LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody);
-    InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_56, float_tmp5, const_int32_29, "tmp7", label_forbody);
-    std::vector<Value*> ptr_arrayidx12_indices;
-    ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_54);
-    ptr_arrayidx12_indices.push_back(const_int32_31);
-    Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody);
-    LoadInst* float_tmp13_58 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody);
-    InsertElementInst* packed_tmp15_59 = new InsertElementInst(packed_tmp7, float_tmp13_58, const_int32_31, "tmp15", label_forbody);
-    std::vector<Value*> ptr_arrayidx20_indices;
-    ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_54);
-    ptr_arrayidx20_indices.push_back(const_int32_33);
-    Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody);
-    LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody);
-    InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_59, float_tmp21, const_int32_33, "tmp23", label_forbody);
-    std::vector<Value*> ptr_arrayidx28_indices;
-    ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_54);
-    ptr_arrayidx28_indices.push_back(const_int32_34);
-    Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody);
-    LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody);
-    InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_34, "tmp31", label_forbody);
-    GetElementPtrInst* ptr_arrayidx34_60 = new GetElementPtrInst(ptr_res_46, int32_i_0_reg2mem_0_54, "arrayidx34", label_forbody);
-    StoreInst* void_61 = new StoreInst(packed_tmp31, ptr_arrayidx34_60, false, label_forbody);
-    BinaryOperator* int32_indvar_next_62 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_54, const_int32_31, "indvar.next", label_forbody);
-    ICmpInst* int1_exitcond_63 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_62, int32_tmp10, "exitcond", label_forbody);
-    new BranchInst(label_afterfor, label_forbody, int1_exitcond_63, label_forbody);
-    
-    // Block afterfor (label_afterfor)
-    new ReturnInst(label_afterfor);
-    
-    // Resolve Forward References
-    fwdref_57->replaceAllUsesWith(packed_tmp31); delete fwdref_57;
-    fwdref_55->replaceAllUsesWith(int32_indvar_next_62); delete fwdref_55;
-    
-  }
-  
-  // Function: to_array (func_to_array)
-  {
-    Function::arg_iterator args = func_to_array->arg_begin();
-    Value* ptr_dests = args++;
-    ptr_dests->setName("dests");
-    Value* ptr_in = args++;
-    ptr_in->setName("in");
-    Value* int32_num_attribs_66 = args++;
-    int32_num_attribs_66->setName("num_attribs");
-    
-    BasicBlock* label_entry_67 = new BasicBlock("entry",func_to_array,0);
-    BasicBlock* label_forbody_preheader_68 = new BasicBlock("forbody.preheader",func_to_array,0);
-    BasicBlock* label_forbody_69 = new BasicBlock("forbody",func_to_array,0);
-    BasicBlock* label_afterfor_70 = new BasicBlock("afterfor",func_to_array,0);
-    
-    // Block entry (label_entry_67)
-    ICmpInst* int1_cmp_71 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_66, const_int32_29, "cmp", label_entry_67);
-    new BranchInst(label_forbody_preheader_68, label_afterfor_70, int1_cmp_71, label_entry_67);
-    
-    // Block forbody.preheader (label_forbody_preheader_68)
-    BinaryOperator* int32_tmp_73 = BinaryOperator::create(Instruction::Add, int32_num_attribs_66, const_int32_30, "tmp", label_forbody_preheader_68);
-    ICmpInst* int1_tmp9_74 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_73, const_int32_29, "tmp9", label_forbody_preheader_68);
-    SelectInst* int32_tmp10_75 = new SelectInst(int1_tmp9_74, const_int32_31, int32_num_attribs_66, "tmp10", label_forbody_preheader_68);
-    new BranchInst(label_forbody_69, label_forbody_preheader_68);
-    
-    // Block forbody (label_forbody_69)
-    Argument* fwdref_78 = new Argument(IntegerType::get(32));
-    PHINode* int32_i_0_reg2mem_0_77 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_69);
-    int32_i_0_reg2mem_0_77->reserveOperandSpace(2);
-    int32_i_0_reg2mem_0_77->addIncoming(const_int32_29, label_forbody_preheader_68);
-    int32_i_0_reg2mem_0_77->addIncoming(fwdref_78, label_forbody_69);
-    
-    std::vector<Value*> ptr_arraydecay_79_indices;
-    ptr_arraydecay_79_indices.push_back(int32_i_0_reg2mem_0_77);
-    ptr_arraydecay_79_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay_79 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_79_indices.begin(), ptr_arraydecay_79_indices.end(), "arraydecay", label_forbody_69);
-    GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_77, "arrayidx6", label_forbody_69);
-    LoadInst* packed_tmp7_80 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_69);
-    ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_80, const_int32_29, "tmp11", label_forbody_69);
-    StoreInst* void_81 = new StoreInst(float_tmp11, ptr_arraydecay_79, false, label_forbody_69);
-    std::vector<Value*> ptr_arrayidx13_indices;
-    ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_77);
-    ptr_arrayidx13_indices.push_back(const_int32_31);
-    Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_69);
-    ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_80, const_int32_31, "tmp15", label_forbody_69);
-    StoreInst* void_82 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_69);
-    std::vector<Value*> ptr_arrayidx17_indices;
-    ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_77);
-    ptr_arrayidx17_indices.push_back(const_int32_33);
-    Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_69);
-    ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_80, const_int32_33, "tmp19", label_forbody_69);
-    StoreInst* void_83 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_69);
-    std::vector<Value*> ptr_arrayidx21_indices;
-    ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_77);
-    ptr_arrayidx21_indices.push_back(const_int32_34);
-    Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_69);
-    ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_80, const_int32_34, "tmp23", label_forbody_69);
-    StoreInst* void_84 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_69);
-    BinaryOperator* int32_indvar_next_85 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_77, const_int32_31, "indvar.next", label_forbody_69);
-    ICmpInst* int1_exitcond_86 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_85, int32_tmp10_75, "exitcond", label_forbody_69);
-    new BranchInst(label_afterfor_70, label_forbody_69, int1_exitcond_86, label_forbody_69);
-    
-    // Block afterfor (label_afterfor_70)
-    new ReturnInst(label_afterfor_70);
-    
-    // Resolve Forward References
-    fwdref_78->replaceAllUsesWith(int32_indvar_next_85); delete fwdref_78;
-    
-  }
-  
-  // Function: run_vertex_shader (func_run_vertex_shader)
-  {
-    Function::arg_iterator args = func_run_vertex_shader->arg_begin();
-    Value* ptr_inputs = args++;
-    ptr_inputs->setName("inputs");
-    Value* ptr_results = args++;
-    ptr_results->setName("results");
-    Value* ptr_aconsts = args++;
-    ptr_aconsts->setName("aconsts");
-    Value* int32_num_vertices = args++;
-    int32_num_vertices->setName("num_vertices");
-    Value* int32_num_inputs = args++;
-    int32_num_inputs->setName("num_inputs");
-    Value* int32_num_attribs_89 = args++;
-    int32_num_attribs_89->setName("num_attribs");
-    Value* int32_num_consts = args++;
-    int32_num_consts->setName("num_consts");
-    
-    BasicBlock* label_entry_90 = new BasicBlock("entry",func_run_vertex_shader,0);
-    BasicBlock* label_forbody_preheader_i = new BasicBlock("forbody.preheader.i",func_run_vertex_shader,0);
-    BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_vertex_shader,0);
-    BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0);
-    
-    // Block entry (label_entry_90)
-    AllocaInst* ptr_consts = new AllocaInst(ArrayTy_20, "consts", label_entry_90);
-    AllocaInst* ptr_temps = new AllocaInst(ArrayTy_22, "temps", label_entry_90);
-    AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_90);
-    std::vector<Value*> ptr_tmp_indices;
-    ptr_tmp_indices.push_back(const_int32_29);
-    ptr_tmp_indices.push_back(const_int32_29);
-    Instruction* ptr_tmp = new GetElementPtrInst(ptr_args, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_90);
-    CastInst* ptr_conv = new BitCastInst(ptr_results, PointerTy_0, "conv", label_entry_90);
-    StoreInst* void_91 = new StoreInst(ptr_conv, ptr_tmp, false, label_entry_90);
-    std::vector<Value*> ptr_tmp2_indices;
-    ptr_tmp2_indices.push_back(const_int32_29);
-    ptr_tmp2_indices.push_back(const_int32_31);
-    Instruction* ptr_tmp2 = new GetElementPtrInst(ptr_args, ptr_tmp2_indices.begin(), ptr_tmp2_indices.end(), "tmp2", label_entry_90);
-    CastInst* ptr_conv4 = new BitCastInst(ptr_inputs, PointerTy_0, "conv4", label_entry_90);
-    StoreInst* void_92 = new StoreInst(ptr_conv4, ptr_tmp2, false, label_entry_90);
-    ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_29, "cmp.i", label_entry_90);
-    new BranchInst(label_forbody_preheader_i, label_from_consts_exit, int1_cmp_i, label_entry_90);
-    
-    // Block forbody.preheader.i (label_forbody_preheader_i)
-    BinaryOperator* int32_tmp_i = BinaryOperator::create(Instruction::Add, int32_num_consts, const_int32_30, "tmp.i", label_forbody_preheader_i);
-    ICmpInst* int1_tmp9_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i, const_int32_29, "tmp9.i", label_forbody_preheader_i);
-    SelectInst* int32_tmp10_i = new SelectInst(int1_tmp9_i, const_int32_31, int32_num_consts, "tmp10.i", label_forbody_preheader_i);
-    new BranchInst(label_forbody_i, label_forbody_preheader_i);
-    
-    // Block forbody.i (label_forbody_i)
-    Argument* fwdref_95 = new Argument(IntegerType::get(32));
-    PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i);
-    int32_i_0_reg2mem_0_i->reserveOperandSpace(2);
-    int32_i_0_reg2mem_0_i->addIncoming(const_int32_29, label_forbody_preheader_i);
-    int32_i_0_reg2mem_0_i->addIncoming(fwdref_95, label_forbody_i);
-    
-    Argument* fwdref_96 = new Argument(VectorTy_1);
-    PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i);
-    packed_vec_0_reg2mem_0_i->reserveOperandSpace(2);
-    packed_vec_0_reg2mem_0_i->addIncoming(const_packed_32, label_forbody_preheader_i);
-    packed_vec_0_reg2mem_0_i->addIncoming(fwdref_96, label_forbody_i);
-    
-    std::vector<Value*> ptr_arraydecay_i_indices;
-    ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i);
-    ptr_arraydecay_i_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_i);
-    LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i, "tmp5.i", false, label_forbody_i);
-    InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp5_i, const_int32_29, "tmp7.i", label_forbody_i);
-    std::vector<Value*> ptr_arrayidx12_i_indices;
-    ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i);
-    ptr_arrayidx12_i_indices.push_back(const_int32_31);
-    Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i);
-    LoadInst* float_tmp13_i = new LoadInst(ptr_arrayidx12_i, "tmp13.i", false, label_forbody_i);
-    InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_tmp7_i, float_tmp13_i, const_int32_31, "tmp15.i", label_forbody_i);
-    std::vector<Value*> ptr_arrayidx20_i_indices;
-    ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i);
-    ptr_arrayidx20_i_indices.push_back(const_int32_33);
-    Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i);
-    LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i);
-    InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i, float_tmp21_i, const_int32_33, "tmp23.i", label_forbody_i);
-    std::vector<Value*> ptr_arrayidx28_i_indices;
-    ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i);
-    ptr_arrayidx28_i_indices.push_back(const_int32_34);
-    Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i);
-    LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i);
-    InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_34, "tmp31.i", label_forbody_i);
-    std::vector<Value*> ptr_arrayidx34_i_indices;
-    ptr_arrayidx34_i_indices.push_back(const_int32_29);
-    ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i);
-    Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody_i);
-    StoreInst* void_97 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i);
-    BinaryOperator* int32_indvar_next_98 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_31, "indvar.next", label_forbody_i);
-    ICmpInst* int1_exitcond_99 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_98, int32_tmp10_i, "exitcond", label_forbody_i);
-    new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond_99, label_forbody_i);
-    
-    // Block from_consts.exit (label_from_consts_exit)
-    std::vector<Value*> ptr_tmp7_indices;
-    ptr_tmp7_indices.push_back(const_int32_29);
-    ptr_tmp7_indices.push_back(const_int32_34);
-    Instruction* ptr_tmp7 = new GetElementPtrInst(ptr_args, ptr_tmp7_indices.begin(), ptr_tmp7_indices.end(), "tmp7", label_from_consts_exit);
-    std::vector<Value*> ptr_arraydecay8_indices;
-    ptr_arraydecay8_indices.push_back(const_int32_29);
-    ptr_arraydecay8_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_consts, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_from_consts_exit);
-    StoreInst* void_101 = new StoreInst(ptr_arraydecay8, ptr_tmp7, false, label_from_consts_exit);
-    std::vector<Value*> ptr_tmp9_indices;
-    ptr_tmp9_indices.push_back(const_int32_29);
-    ptr_tmp9_indices.push_back(const_int32_33);
-    Instruction* ptr_tmp9 = new GetElementPtrInst(ptr_args, ptr_tmp9_indices.begin(), ptr_tmp9_indices.end(), "tmp9", label_from_consts_exit);
-    std::vector<Value*> ptr_arraydecay10_indices;
-    ptr_arraydecay10_indices.push_back(const_int32_29);
-    ptr_arraydecay10_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_temps, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_from_consts_exit);
-    StoreInst* void_102 = new StoreInst(ptr_arraydecay10, ptr_tmp9, false, label_from_consts_exit);
-    CallInst* void_103 = new CallInst(func_execute_shader, ptr_args, "", label_from_consts_exit);
-    void_103->setCallingConv(CallingConv::C);
-    void_103->setTailCall(false);const ParamAttrsList *void_103_PAL = 0;
-    void_103->setParamAttrs(void_103_PAL);
-    
-    new ReturnInst(label_from_consts_exit);
-    
-    // Resolve Forward References
-    fwdref_96->replaceAllUsesWith(packed_tmp31_i); delete fwdref_96;
-    fwdref_95->replaceAllUsesWith(int32_indvar_next_98); delete fwdref_95;
-    
-  }
-  
-  // Function: run_fragment_shader (func_run_fragment_shader)
-  {
-    Function::arg_iterator args = func_run_fragment_shader->arg_begin();
-    Value* float_x = args++;
-    float_x->setName("x");
-    Value* float_y = args++;
-    float_y->setName("y");
-    Value* ptr_results_105 = args++;
-    ptr_results_105->setName("results");
-    Value* ptr_inputs_106 = args++;
-    ptr_inputs_106->setName("inputs");
-    Value* int32_num_inputs_107 = args++;
-    int32_num_inputs_107->setName("num_inputs");
-    Value* ptr_aconsts_108 = args++;
-    ptr_aconsts_108->setName("aconsts");
-    Value* int32_num_consts_109 = args++;
-    int32_num_consts_109->setName("num_consts");
-    Value* ptr_samplers = args++;
-    ptr_samplers->setName("samplers");
-    
-    BasicBlock* label_entry_110 = new BasicBlock("entry",func_run_fragment_shader,0);
-    BasicBlock* label_forbody_preheader_i_111 = new BasicBlock("forbody.preheader.i",func_run_fragment_shader,0);
-    BasicBlock* label_forbody_i_112 = new BasicBlock("forbody.i",func_run_fragment_shader,0);
-    BasicBlock* label_from_consts_exit_113 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0);
-    
-    // Block entry (label_entry_110)
-    AllocaInst* ptr_consts_114 = new AllocaInst(ArrayTy_20, "consts", label_entry_110);
-    AllocaInst* ptr_temps_115 = new AllocaInst(ArrayTy_22, "temps", label_entry_110);
-    AllocaInst* ptr_args_116 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_110);
-    std::vector<Value*> ptr_tmp_117_indices;
-    ptr_tmp_117_indices.push_back(const_int32_29);
-    ptr_tmp_117_indices.push_back(const_int32_35);
-    Instruction* ptr_tmp_117 = new GetElementPtrInst(ptr_args_116, ptr_tmp_117_indices.begin(), ptr_tmp_117_indices.end(), "tmp", label_entry_110);
-    StoreInst* void_118 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_entry_110);
-    ICmpInst* int1_cmp_i_119 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_109, const_int32_29, "cmp.i", label_entry_110);
-    new BranchInst(label_forbody_preheader_i_111, label_from_consts_exit_113, int1_cmp_i_119, label_entry_110);
-    
-    // Block forbody.preheader.i (label_forbody_preheader_i_111)
-    BinaryOperator* int32_tmp_i_121 = BinaryOperator::create(Instruction::Add, int32_num_consts_109, const_int32_30, "tmp.i", label_forbody_preheader_i_111);
-    ICmpInst* int1_tmp9_i_122 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i_121, const_int32_29, "tmp9.i", label_forbody_preheader_i_111);
-    SelectInst* int32_tmp10_i_123 = new SelectInst(int1_tmp9_i_122, const_int32_31, int32_num_consts_109, "tmp10.i", label_forbody_preheader_i_111);
-    new BranchInst(label_forbody_i_112, label_forbody_preheader_i_111);
-    
-    // Block forbody.i (label_forbody_i_112)
-    Argument* fwdref_126 = new Argument(IntegerType::get(32));
-    PHINode* int32_i_0_reg2mem_0_i_125 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_112);
-    int32_i_0_reg2mem_0_i_125->reserveOperandSpace(2);
-    int32_i_0_reg2mem_0_i_125->addIncoming(const_int32_29, label_forbody_preheader_i_111);
-    int32_i_0_reg2mem_0_i_125->addIncoming(fwdref_126, label_forbody_i_112);
-    
-    Argument* fwdref_128 = new Argument(VectorTy_1);
-    PHINode* packed_vec_0_reg2mem_0_i_127 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_112);
-    packed_vec_0_reg2mem_0_i_127->reserveOperandSpace(2);
-    packed_vec_0_reg2mem_0_i_127->addIncoming(const_packed_32, label_forbody_preheader_i_111);
-    packed_vec_0_reg2mem_0_i_127->addIncoming(fwdref_128, label_forbody_i_112);
-    
-    std::vector<Value*> ptr_arraydecay_i_129_indices;
-    ptr_arraydecay_i_129_indices.push_back(int32_i_0_reg2mem_0_i_125);
-    ptr_arraydecay_i_129_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay_i_129 = new GetElementPtrInst(ptr_aconsts_108, ptr_arraydecay_i_129_indices.begin(), ptr_arraydecay_i_129_indices.end(), "arraydecay.i", label_forbody_i_112);
-    LoadInst* float_tmp5_i_130 = new LoadInst(ptr_arraydecay_i_129, "tmp5.i", false, label_forbody_i_112);
-    InsertElementInst* packed_tmp7_i_131 = new InsertElementInst(packed_vec_0_reg2mem_0_i_127, float_tmp5_i_130, const_int32_29, "tmp7.i", label_forbody_i_112);
-    std::vector<Value*> ptr_arrayidx12_i_132_indices;
-    ptr_arrayidx12_i_132_indices.push_back(int32_i_0_reg2mem_0_i_125);
-    ptr_arrayidx12_i_132_indices.push_back(const_int32_31);
-    Instruction* ptr_arrayidx12_i_132 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx12_i_132_indices.begin(), ptr_arrayidx12_i_132_indices.end(), "arrayidx12.i", label_forbody_i_112);
-    LoadInst* float_tmp13_i_133 = new LoadInst(ptr_arrayidx12_i_132, "tmp13.i", false, label_forbody_i_112);
-    InsertElementInst* packed_tmp15_i_134 = new InsertElementInst(packed_tmp7_i_131, float_tmp13_i_133, const_int32_31, "tmp15.i", label_forbody_i_112);
-    std::vector<Value*> ptr_arrayidx20_i_135_indices;
-    ptr_arrayidx20_i_135_indices.push_back(int32_i_0_reg2mem_0_i_125);
-    ptr_arrayidx20_i_135_indices.push_back(const_int32_33);
-    Instruction* ptr_arrayidx20_i_135 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx20_i_135_indices.begin(), ptr_arrayidx20_i_135_indices.end(), "arrayidx20.i", label_forbody_i_112);
-    LoadInst* float_tmp21_i_136 = new LoadInst(ptr_arrayidx20_i_135, "tmp21.i", false, label_forbody_i_112);
-    InsertElementInst* packed_tmp23_i_137 = new InsertElementInst(packed_tmp15_i_134, float_tmp21_i_136, const_int32_33, "tmp23.i", label_forbody_i_112);
-    std::vector<Value*> ptr_arrayidx28_i_138_indices;
-    ptr_arrayidx28_i_138_indices.push_back(int32_i_0_reg2mem_0_i_125);
-    ptr_arrayidx28_i_138_indices.push_back(const_int32_34);
-    Instruction* ptr_arrayidx28_i_138 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx28_i_138_indices.begin(), ptr_arrayidx28_i_138_indices.end(), "arrayidx28.i", label_forbody_i_112);
-    LoadInst* float_tmp29_i_139 = new LoadInst(ptr_arrayidx28_i_138, "tmp29.i", false, label_forbody_i_112);
-    InsertElementInst* packed_tmp31_i_140 = new InsertElementInst(packed_tmp23_i_137, float_tmp29_i_139, const_int32_34, "tmp31.i", label_forbody_i_112);
-    std::vector<Value*> ptr_arrayidx34_i_141_indices;
-    ptr_arrayidx34_i_141_indices.push_back(const_int32_29);
-    ptr_arrayidx34_i_141_indices.push_back(int32_i_0_reg2mem_0_i_125);
-    Instruction* ptr_arrayidx34_i_141 = new GetElementPtrInst(ptr_consts_114, ptr_arrayidx34_i_141_indices.begin(), ptr_arrayidx34_i_141_indices.end(), "arrayidx34.i", label_forbody_i_112);
-    StoreInst* void_142 = new StoreInst(packed_tmp31_i_140, ptr_arrayidx34_i_141, false, label_forbody_i_112);
-    BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_125, const_int32_31, "indvar.next7", label_forbody_i_112);
-    ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_tmp10_i_123, "exitcond8", label_forbody_i_112);
-    new BranchInst(label_from_consts_exit_113, label_forbody_i_112, int1_exitcond8, label_forbody_i_112);
-    
-    // Block from_consts.exit (label_from_consts_exit_113)
-    std::vector<Value*> ptr_tmp3_indices;
-    ptr_tmp3_indices.push_back(const_int32_29);
-    ptr_tmp3_indices.push_back(const_int32_34);
-    Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_116, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay4_indices;
-    ptr_arraydecay4_indices.push_back(const_int32_29);
-    ptr_arraydecay4_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_114, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_113);
-    StoreInst* void_144 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_tmp5_indices;
-    ptr_tmp5_indices.push_back(const_int32_29);
-    ptr_tmp5_indices.push_back(const_int32_33);
-    Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_116, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay6_indices;
-    ptr_arraydecay6_indices.push_back(const_int32_29);
-    ptr_arraydecay6_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_115, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_113);
-    StoreInst* void_145 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_tmp8_indices;
-    ptr_tmp8_indices.push_back(const_int32_29);
-    ptr_tmp8_indices.push_back(const_int32_31);
-    Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args_116, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_from_consts_exit_113);
-    std::vector<Value*> ptr_tmp12_indices;
-    ptr_tmp12_indices.push_back(const_int32_29);
-    ptr_tmp12_indices.push_back(const_int32_29);
-    Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args_116, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay11_146_indices;
-    ptr_arraydecay11_146_indices.push_back(const_int32_29);
-    ptr_arraydecay11_146_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay11_146 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_146_indices.begin(), ptr_arraydecay11_146_indices.end(), "arraydecay11", label_from_consts_exit_113);
-    StoreInst* void_147 = new StoreInst(ptr_arraydecay11_146, ptr_tmp8, false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay16_indices;
-    ptr_arraydecay16_indices.push_back(const_int32_29);
-    ptr_arraydecay16_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_from_consts_exit_113);
-    StoreInst* void_148 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_from_consts_exit_113);
-    StoreInst* void_149 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113);
-    CallInst* void_150 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113);
-    void_150->setCallingConv(CallingConv::C);
-    void_150->setTailCall(false);const ParamAttrsList *void_150_PAL = 0;
-    void_150->setParamAttrs(void_150_PAL);
-    
-    LoadInst* int32_tmp23 = new LoadInst(ptr_tmp_117, "tmp23", false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay11_1_indices;
-    ptr_arraydecay11_1_indices.push_back(const_int32_31);
-    ptr_arraydecay11_1_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_113);
-    StoreInst* void_151 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8, false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay16_1_indices;
-    ptr_arraydecay16_1_indices.push_back(const_int32_31);
-    ptr_arraydecay16_1_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_113);
-    StoreInst* void_152 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12, false, label_from_consts_exit_113);
-    StoreInst* void_153 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113);
-    CallInst* void_154 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113);
-    void_154->setCallingConv(CallingConv::C);
-    void_154->setTailCall(false);const ParamAttrsList *void_154_PAL = 0;
-    void_154->setParamAttrs(void_154_PAL);
-    
-    LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp_117, "tmp23.1", false, label_from_consts_exit_113);
-    BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_31, "shl.1", label_from_consts_exit_113);
-    BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay11_2_indices;
-    ptr_arraydecay11_2_indices.push_back(const_int32_33);
-    ptr_arraydecay11_2_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_113);
-    StoreInst* void_155 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8, false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay16_2_indices;
-    ptr_arraydecay16_2_indices.push_back(const_int32_33);
-    ptr_arraydecay16_2_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_113);
-    StoreInst* void_156 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12, false, label_from_consts_exit_113);
-    StoreInst* void_157 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113);
-    CallInst* void_158 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113);
-    void_158->setCallingConv(CallingConv::C);
-    void_158->setTailCall(false);const ParamAttrsList *void_158_PAL = 0;
-    void_158->setParamAttrs(void_158_PAL);
-    
-    LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp_117, "tmp23.2", false, label_from_consts_exit_113);
-    BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_113);
-    BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay11_3_indices;
-    ptr_arraydecay11_3_indices.push_back(const_int32_34);
-    ptr_arraydecay11_3_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_113);
-    StoreInst* void_159 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8, false, label_from_consts_exit_113);
-    std::vector<Value*> ptr_arraydecay16_3_indices;
-    ptr_arraydecay16_3_indices.push_back(const_int32_34);
-    ptr_arraydecay16_3_indices.push_back(const_int32_29);
-    Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_113);
-    StoreInst* void_160 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12, false, label_from_consts_exit_113);
-    StoreInst* void_161 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113);
-    CallInst* void_162 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113);
-    void_162->setCallingConv(CallingConv::C);
-    void_162->setTailCall(false);const ParamAttrsList *void_162_PAL = 0;
-    void_162->setParamAttrs(void_162_PAL);
-    
-    LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp_117, "tmp23.3", false, label_from_consts_exit_113);
-    BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_113);
-    BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_113);
-    BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_30, "neg", label_from_consts_exit_113);
-    new ReturnInst(int32_neg, label_from_consts_exit_113);
-    
-    // Resolve Forward References
-    fwdref_128->replaceAllUsesWith(packed_tmp31_i_140); delete fwdref_128;
-    fwdref_126->replaceAllUsesWith(int32_indvar_next7); delete fwdref_126;
-    
-  }
-  
-  return mod;
-}
diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c
deleted file mode 100644
index fa50b60e663..00000000000
--- a/src/mesa/pipe/llvm/llvm_entry.c
+++ /dev/null
@@ -1,163 +0,0 @@
-/* clang --emit-llvm llvm_entry.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */
-/**************************************************************************
- *
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-
- /*
-  * Authors:
-  *   Zack Rusin [email protected]
-  */
-
-/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */
-typedef __attribute__(( ocu_vector_type(4) )) float float4;
-
-void from_array(float4 (*res)[16], float (*ainputs)[16][4],
-                int count, int num_attribs)
-{
-   for (int i = 0; i < count; ++i) {
-      for (int j = 0; j < num_attribs; ++j) {
-         float4 vec;
-         vec.x = ainputs[i][j][0];
-         vec.y = ainputs[i][j][1];
-         vec.z = ainputs[i][j][2];
-         vec.w = ainputs[i][j][3];
-         res[i][j] = vec;
-      }
-   }
-}
-
-void from_consts(float4 *res, float (*ainputs)[4],
-                 int count)
-{
-   for (int i = 0; i < count; ++i) {
-      float4 vec;
-      vec.x = ainputs[i][0];
-      vec.y = ainputs[i][1];
-      vec.z = ainputs[i][2];
-      vec.w = ainputs[i][3];
-      res[i] = vec;
-   }
-}
-
-void to_array(float (*dests)[4], float4 *in, int num_attribs)
-{
-   for (int i = 0; i < num_attribs; ++i) {
-      float  *rd = dests[i];
-      float4  ri = in[i];
-      rd[0] = ri.x;
-      rd[1] = ri.y;
-      rd[2] = ri.z;
-      rd[3] = ri.w;
-   }
-}
-
-
-struct ShaderInput
-{
-   float4  *dests;
-   float4  *inputs;
-   float4  *temps;
-   float4  *consts;
-   int     kilmask;
-};
-
-extern void execute_shader(struct ShaderInput *input);
-
-void run_vertex_shader(void *inputs,
-                       void *results,
-                       float (*aconsts)[4],
-                       int num_vertices,
-                       int num_inputs,
-                       int num_attribs,
-                       int num_consts)
-{
-   float4  consts[32];
-   float4  temps[128];//MAX_PROGRAM_TEMPS
-
-   struct ShaderInput args;
-   args.dests  = results;
-   args.inputs = inputs;
-
-   /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n",
-     num_vertices, num_inputs, num_attribs, num_consts);*/
-   from_consts(consts, aconsts, num_consts);
-   args.consts = consts;
-   args.temps = temps;
-
-   execute_shader(&args);
-}
-
-
-struct pipe_sampler_state;
-struct softpipe_tile_cache;
-
-#define NUM_CHANNELS 4  /* R,G,B,A */
-#define QUAD_SIZE    4  /* 4 pixel/quad */
-
-struct tgsi_sampler
-{
-   const struct pipe_sampler_state *state;
-   /** Get samples for four fragments in a quad */
-   void (*get_samples)(struct tgsi_sampler *sampler,
-                       const float s[QUAD_SIZE],
-                       const float t[QUAD_SIZE],
-                       const float p[QUAD_SIZE],
-                       float lodbias,
-                       float rgba[NUM_CHANNELS][QUAD_SIZE]);
-   void *pipe; /*XXX temporary*/
-   struct softpipe_tile_cache *cache;
-};
-
-
-int run_fragment_shader(float x, float y,
-                        float4 (*results)[16],
-                        float4 (*inputs)[16],
-                        int num_inputs,
-                        float (*aconsts)[4],
-                        int num_consts,
-                        struct tgsi_sampler *samplers)
-{
-   float4  consts[32];
-   float4  temps[128];//MAX_PROGRAM_TEMPS
-   struct ShaderInput args;
-   int mask = 0;
-   args.kilmask = 0;
-
-   from_consts(consts, aconsts, num_consts);
-   args.consts = consts;
-   args.temps = temps;
-   //printf("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\n", x, y);
-   for (int i = 0; i < 4; ++i) {
-      args.inputs  = inputs[i];
-      args.dests   = results[i];
-      mask = args.kilmask;
-      args.kilmask = 0;
-      execute_shader(&args);
-      args.kilmask = mask | (args.kilmask << i);
-   }
-   return ~args.kilmask;
-}
-
diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp
deleted file mode 100644
index b2aca3557a9..00000000000
--- a/src/mesa/pipe/llvm/storagesoa.cpp
+++ /dev/null
@@ -1,117 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-
-#include "storagesoa.h"
-
-
-#include "pipe/p_shader_tokens.h"
-#include <llvm/BasicBlock.h>
-#include <llvm/Module.h>
-#include <llvm/Value.h>
-
-#include <llvm/CallingConv.h>
-#include <llvm/Constants.h>
-#include <llvm/DerivedTypes.h>
-#include <llvm/InstrTypes.h>
-#include <llvm/Instructions.h>
-
-using namespace llvm;
-
-StorageSoa::StorageSoa(llvm::BasicBlock *block,
-                       llvm::Value *input,
-                       llvm::Value *output,
-                       llvm::Value *consts)
-{
-}
-
-void StorageSoa::addImmediate(float *vec)
-{
-}
-
-llvm::Value *StorageSoa::addrElement(int idx) const
-{
-   return 0;
-}
-
-std::vector<llvm::Value*> StorageSoa::inputElement(int idx, int swizzle,
-                                                   llvm::Value *indIdx)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-std::vector<llvm::Value*> StorageSoa::constElement(int idx, int swizzle,
-                                                   llvm::Value *indIdx)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-std::vector<llvm::Value*> StorageSoa::outputElement(int idx, int swizzle,
-                                                    llvm::Value *indIdx)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-std::vector<llvm::Value*> StorageSoa::tempElement(int idx, int swizzle,
-                                                  llvm::Value *indIdx)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-std::vector<llvm::Value*> StorageSoa::immediateElement(int idx, int swizzle)
-{
-   std::vector<llvm::Value*> res(4);
-
-   return res;
-}
-
-llvm::Value * StorageSoa::extractIndex(llvm::Value *vec)
-{
-   return 0;
-}
-
-void StorageSoa::storeOutput(int dstIdx, const std::vector<llvm::Value*> &val,
-                             int mask)
-{
-}
-
-void StorageSoa::storeTemp(int idx, const std::vector<llvm::Value*> &val,
-                           int mask)
-{
-}
-
-void StorageSoa::storeAddress(int idx, const std::vector<llvm::Value*> &val,
-                              int mask)
-{
-}
diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c
deleted file mode 100644
index 39c3e1afe10..00000000000
--- a/src/mesa/pipe/softpipe/sp_state_derived.c
+++ /dev/null
@@ -1,190 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include "pipe/p_util.h"
-#include "pipe/p_shader_tokens.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_vertex.h"
-#include "pipe/draw/draw_private.h"
-#include "sp_context.h"
-#include "sp_state.h"
-
-
-static int
-find_vs_output(const struct pipe_shader_state *vs,
-               uint semantic_name,
-               uint semantic_index)
-{
-   uint i;
-   for (i = 0; i < vs->num_outputs; i++) {
-      if (vs->output_semantic_name[i] == semantic_name &&
-          vs->output_semantic_index[i] == semantic_index)
-         return i;
-   }
-   return -1;
-}
-
-
-
-/**
- * Determine how to map vertex program outputs to fragment program inputs.
- * Basically, this will be used when computing the triangle interpolation
- * coefficients from the post-transform vertex attributes.
- */
-static void calculate_vertex_layout( struct softpipe_context *softpipe )
-{
-   const struct pipe_shader_state *vs = &softpipe->vs->shader;
-   const struct pipe_shader_state *fs = &softpipe->fs->shader;
-   const enum interp_mode colorInterp
-      = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
-   struct vertex_info *vinfo = &softpipe->vertex_info;
-   uint i;
-
-   if (softpipe->vbuf) {
-      /* if using the post-transform vertex buffer, tell draw_vbuf to
-       * simply emit the whole post-xform vertex as-is:
-       */
-      struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf;
-      vinfo_vbuf->num_attribs = 0;
-      draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0);
-      vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4;
-   }
-
-   /*
-    * Loop over fragment shader inputs, searching for the matching output
-    * from the vertex shader.
-    */
-   vinfo->num_attribs = 0;
-   for (i = 0; i < fs->num_inputs; i++) {
-      int src;
-      switch (fs->input_semantic_name[i]) {
-      case TGSI_SEMANTIC_POSITION:
-         src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0);
-         assert(src >= 0);
-         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src);
-         break;
-
-      case TGSI_SEMANTIC_COLOR:
-         src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, 
-                              fs->input_semantic_index[i]);
-         assert(src >= 0);
-         draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
-         break;
-
-      case TGSI_SEMANTIC_FOG:
-         src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0);
-#if 1
-         if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */
-            src = 0;
-#endif
-         assert(src >= 0);
-         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
-         break;
-
-      case TGSI_SEMANTIC_GENERIC:
-         /* this includes texcoords and varying vars */
-         src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC,
-                              fs->input_semantic_index[i]);
-         assert(src >= 0);
-         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
-         break;
-
-      default:
-         assert(0);
-      }
-   }
-
-   softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0);
-   if (softpipe->psize_slot >= 0) {
-      draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT,
-                            softpipe->psize_slot);
-   }
-
-   draw_compute_vertex_size(vinfo);
-}
-
-
-/**
- * Recompute cliprect from scissor bounds, scissor enable and surface size.
- */
-static void
-compute_cliprect(struct softpipe_context *sp)
-{
-   unsigned surfWidth, surfHeight;
-
-   if (sp->framebuffer.num_cbufs > 0) {
-      surfWidth = sp->framebuffer.cbufs[0]->width;
-      surfHeight = sp->framebuffer.cbufs[0]->height;
-   }
-   else {
-      /* no surface? */
-      surfWidth = sp->scissor.maxx;
-      surfHeight = sp->scissor.maxy;
-   }
-
-   if (sp->rasterizer->scissor) {
-      /* clip to scissor rect */
-      sp->cliprect.minx = MAX2(sp->scissor.minx, 0);
-      sp->cliprect.miny = MAX2(sp->scissor.miny, 0);
-      sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth);
-      sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight);
-   }
-   else {
-      /* clip to surface bounds */
-      sp->cliprect.minx = 0;
-      sp->cliprect.miny = 0;
-      sp->cliprect.maxx = surfWidth;
-      sp->cliprect.maxy = surfHeight;
-   }
-}
-
-
-/* Hopefully this will remain quite simple, otherwise need to pull in
- * something like the state tracker mechanism.
- */
-void softpipe_update_derived( struct softpipe_context *softpipe )
-{
-   if (softpipe->dirty & (SP_NEW_RASTERIZER |
-                          SP_NEW_FS |
-                          SP_NEW_VS))
-      calculate_vertex_layout( softpipe );
-
-   if (softpipe->dirty & (SP_NEW_SCISSOR |
-                          SP_NEW_DEPTH_STENCIL_ALPHA |
-                          SP_NEW_FRAMEBUFFER))
-      compute_cliprect(softpipe);
-
-   if (softpipe->dirty & (SP_NEW_BLEND |
-                          SP_NEW_DEPTH_STENCIL_ALPHA |
-                          SP_NEW_FRAMEBUFFER |
-                          SP_NEW_RASTERIZER |
-                          SP_NEW_FS | 
-			  SP_NEW_QUERY))
-      sp_build_quad_pipeline(softpipe);
-
-   softpipe->dirty = 0;
-}
diff --git a/src/mesa/ppc/rtasm/spe_asm.c b/src/mesa/ppc/rtasm/spe_asm.c
new file mode 100644
index 00000000000..10376372502
--- /dev/null
+++ b/src/mesa/ppc/rtasm/spe_asm.c
@@ -0,0 +1,385 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file spe_asm.c
+ * Real-time assembly generation interface for Cell B.E. SPEs.
+ *
+ * \author Ian Romanick <[email protected]>
+ */
+#ifdef GALLIUM_CELL
+#include <inttypes.h>
+#include <imports.h>
+#include "spe_asm.h"
+
+/**
+ * SPE instruction types
+ *
+ * There are 6 primary instruction encodings used on the Cell's SPEs.  Each of
+ * the following unions encodes one type.
+ *
+ * \bug
+ * If, at some point, we start generating SPE code from a little-endian host
+ * these unions will not work.
+ */
+/*@{*/
+/**
+ * Encode one output register with two input registers
+ */
+union spe_inst_RR {
+    uint32_t bits;
+    struct {
+	unsigned op:11;
+	unsigned rB:7;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with three input registers
+ */
+union spe_inst_RRR {
+    uint32_t bits;
+    struct {
+	unsigned op:4;
+	unsigned rT:7;
+	unsigned rB:7;
+	unsigned rA:7;
+	unsigned rC:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with one input reg. and a 7-bit signed immed
+ */
+union spe_inst_RI7 {
+    uint32_t bits;
+    struct {
+	unsigned op:11;
+	unsigned i7:7;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with one input reg. and an 8-bit signed immed
+ */
+union spe_inst_RI8 {
+    uint32_t bits;
+    struct {
+	unsigned op:10;
+	unsigned i8:8;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with one input reg. and a 10-bit signed immed
+ */
+union spe_inst_RI10 {
+    uint32_t bits;
+    struct {
+	unsigned op:8;
+	unsigned i10:10;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with a 16-bit signed immediate
+ */
+union spe_inst_RI16 {
+    uint32_t bits;
+    struct {
+	unsigned op:9;
+	unsigned i16:16;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with a 18-bit signed immediate
+ */
+union spe_inst_RI18 {
+    uint32_t bits;
+    struct {
+	unsigned op:7;
+	unsigned i18:18;
+	unsigned rT:7;
+    } inst;
+};
+/*@}*/
+
+
+static void emit_RR(struct spe_function *p, unsigned op, unsigned rT,
+		    unsigned rA, unsigned rB)
+{
+    union spe_inst_RR inst;
+    inst.inst.op = op;
+    inst.inst.rB = rB;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+static void emit_RRR(struct spe_function *p, unsigned op, unsigned rT,
+		    unsigned rA, unsigned rB, unsigned rC)
+{
+    union spe_inst_RRR inst;
+    inst.inst.op = op;
+    inst.inst.rT = rT;
+    inst.inst.rB = rB;
+    inst.inst.rA = rA;
+    inst.inst.rC = rC;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+static void emit_RI7(struct spe_function *p, unsigned op, unsigned rT,
+		     unsigned rA, int imm)
+{
+    union spe_inst_RI7 inst;
+    inst.inst.op = op;
+    inst.inst.i7 = imm;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+
+static void emit_RI8(struct spe_function *p, unsigned op, unsigned rT,
+		     unsigned rA, int imm)
+{
+    union spe_inst_RI8 inst;
+    inst.inst.op = op;
+    inst.inst.i8 = imm;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+
+static void emit_RI10(struct spe_function *p, unsigned op, unsigned rT,
+		      unsigned rA, int imm)
+{
+    union spe_inst_RI10 inst;
+    inst.inst.op = op;
+    inst.inst.i10 = imm;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+static void emit_RI16(struct spe_function *p, unsigned op, unsigned rT,
+		      int imm)
+{
+    union spe_inst_RI16 inst;
+    inst.inst.op = op;
+    inst.inst.i16 = imm;
+    inst.inst.rT = rT;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+static void emit_RI18(struct spe_function *p, unsigned op, unsigned rT,
+		      int imm)
+{
+    union spe_inst_RI18 inst;
+    inst.inst.op = op;
+    inst.inst.i18 = imm;
+    inst.inst.rT = rT;
+    *p->csr = inst.bits;
+    p->csr++;
+}
+
+
+
+
+#define EMIT_(_name, _op) \
+void _name (struct spe_function *p, unsigned rT) \
+{ \
+    emit_RR(p, _op, rT, 0, 0); \
+}
+
+#define EMIT_R(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, unsigned rA) \
+{ \
+    emit_RR(p, _op, rT, rA, 0); \
+}
+
+#define EMIT_RR(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB) \
+{ \
+    emit_RR(p, _op, rT, rA, rB); \
+}
+
+#define EMIT_RRR(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB, unsigned rC) \
+{ \
+    emit_RRR(p, _op, rT, rA, rB, rC); \
+}
+
+#define EMIT_RI7(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \
+{ \
+    emit_RI7(p, _op, rT, rA, imm); \
+}
+
+#define EMIT_RI8(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \
+{ \
+    emit_RI8(p, _op, rT, rA, 155 - imm); \
+}
+
+#define EMIT_RI10(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \
+{ \
+    emit_RI10(p, _op, rT, rA, imm); \
+}
+
+#define EMIT_RI16(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, int imm) \
+{ \
+    emit_RI16(p, _op, rT, imm); \
+}
+
+#define EMIT_RI18(_name, _op) \
+void _name (struct spe_function *p, unsigned rT, int imm) \
+{ \
+    emit_RI18(p, _op, rT, imm); \
+}
+
+#define EMIT_I16(_name, _op) \
+void _name (struct spe_function *p, int imm) \
+{ \
+    emit_RI16(p, _op, 0, imm); \
+}
+
+#include "spe_asm.h"
+
+
+/*
+ */
+void spe_init_func(struct spe_function *p, unsigned code_size)
+{
+    p->store = _mesa_align_malloc(code_size, 16);
+    p->csr = p->store;
+}
+
+
+void spe_release_func(struct spe_function *p)
+{
+    _mesa_align_free(p->store);
+    p->store = NULL;
+    p->csr = NULL;
+}
+
+
+void spe_bi(struct spe_function *p, unsigned rA, int d, int e)
+{
+    emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4));
+}
+
+void spe_iret(struct spe_function *p, unsigned rA, int d, int e)
+{
+    emit_RI7(p, 0x1aa, 0, rA, (d << 5) | (e << 4));
+}
+
+void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d,
+		int e)
+{
+    emit_RI7(p, 0x1ab, rT, rA, (d << 5) | (e << 4));
+}
+
+void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d,
+		int e)
+{
+    emit_RI7(p, 0x1a9, rT, rA, (d << 5) | (e << 4));
+}
+
+void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, int d,
+		int e)
+{
+    emit_RI7(p, 0x128, rT, rA, (d << 5) | (e << 4));
+}
+
+void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e)
+{
+    emit_RI7(p, 0x129, rT, rA, (d << 5) | (e << 4));
+}
+
+void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e)
+{
+    emit_RI7(p, 0x12a, rT, rA, (d << 5) | (e << 4));
+}
+
+void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e)
+{
+    emit_RI7(p, 0x12b, rT, rA, (d << 5) | (e << 4));
+}
+
+
+/* Hint-for-branch instructions
+ */
+#if 0
+hbr;
+hbra;
+hbrr;
+#endif
+
+
+/* Control instructions
+ */
+#if 0
+stop;
+EMIT_RR  (spe_stopd, 0x140);
+EMIT_    (spe_lnop,  0x001);
+EMIT_    (spe_nop,   0x201);
+sync;
+EMIT_    (spe_dsync, 0x003);
+EMIT_R   (spe_mfspr, 0x00c);
+EMIT_R   (spe_mtspr, 0x10c);
+#endif
+
+#endif /* GALLIUM_CELL */
diff --git a/src/mesa/ppc/rtasm/spe_asm.h b/src/mesa/ppc/rtasm/spe_asm.h
new file mode 100644
index 00000000000..6d69ae655d1
--- /dev/null
+++ b/src/mesa/ppc/rtasm/spe_asm.h
@@ -0,0 +1,314 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file spe_asm.h
+ * Real-time assembly generation interface for Cell B.E. SPEs.
+ *
+ * \author Ian Romanick <[email protected]>
+ */
+
+#ifndef SPE_ASM_H
+#define SPE_ASM_H
+
+struct spe_function {
+    /**
+     *
+     */
+    uint32_t *store;
+    uint32_t *csr;
+    const char *fn;
+};
+
+extern void spe_init_func(struct spe_function *p, unsigned code_size);
+extern void spe_release_func(struct spe_function *p);
+
+#endif /* SPE_ASM_H */
+
+#ifndef EMIT_
+#define EMIT_(name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT)
+#define EMIT_R(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, unsigned rA)
+#define EMIT_RR(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \
+			   unsigned rB)
+#define EMIT_RRR(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \
+			   unsigned rB, unsigned rC)
+#define EMIT_RI7(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \
+			   int imm)
+#define EMIT_RI8(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \
+			   int imm)
+#define EMIT_RI10(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \
+			   int imm)
+#define EMIT_RI16(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, int imm)
+#define EMIT_RI18(_name, _op) \
+    extern void _name (struct spe_function *p, unsigned rT, int imm)
+#define EMIT_I16(_name, _op) \
+    extern void _name (struct spe_function *p, int imm)
+#define UNDEF_EMIT_MACROS
+#endif /* EMIT_ */
+
+
+/* Memory load / store instructions
+ */
+EMIT_RI10(spe_lqd,  0x034);
+EMIT_RR  (spe_lqx,  0x1c4);
+EMIT_RI16(spe_lqa,  0x061);
+EMIT_RI16(spe_lqr,  0x067);
+EMIT_RI10(spe_stqd, 0x024);
+EMIT_RR  (spe_stqx, 0x144);
+EMIT_RI16(spe_stqa, 0x041);
+EMIT_RI16(spe_stqr, 0x047);
+EMIT_RI7 (spe_cbd,  0x1f4);
+EMIT_RR  (spe_cbx,  0x1d4);
+EMIT_RI7 (spe_chd,  0x1f5);
+EMIT_RI7 (spe_chx,  0x1d5);
+EMIT_RI7 (spe_cwd,  0x1f6);
+EMIT_RI7 (spe_cwx,  0x1d6);
+EMIT_RI7 (spe_cdd,  0x1f7);
+EMIT_RI7 (spe_cdx,  0x1d7);
+
+
+/* Constant formation instructions
+ */
+EMIT_RI16(spe_ilh,   0x083);
+EMIT_RI16(spe_ilhu,  0x082);
+EMIT_RI16(spe_il,    0x081);
+EMIT_RI18(spe_ila,   0x021);
+EMIT_RI16(spe_iohl,  0x0c1);
+EMIT_RI16(spe_fsmbi, 0x0c5);
+
+
+/* Integer and logical instructions
+ */
+EMIT_RR  (spe_ah,      0x0c8);
+EMIT_RI10(spe_ahi,     0x01d);
+EMIT_RR  (spe_a,       0x0c0);
+EMIT_RI10(spe_ai,      0x01c);
+EMIT_RR  (spe_sfh,     0x048);
+EMIT_RI10(spe_sfhi,    0x00d);
+EMIT_RR  (spe_sf,      0x040);
+EMIT_RI10(spe_sfi,     0x00c);
+EMIT_RR  (spe_addx,    0x340);
+EMIT_RR  (spe_cg,      0x0c2);
+EMIT_RR  (spe_cgx,     0x342);
+EMIT_RR  (spe_sfx,     0x341);
+EMIT_RR  (spe_bg,      0x042);
+EMIT_RR  (spe_bgx,     0x343);
+EMIT_RR  (spe_mpy,     0x3c4);
+EMIT_RR  (spe_mpyu,    0x3cc);
+EMIT_RI10(spe_mpyi,    0x074);
+EMIT_RI10(spe_mpyui,   0x075);
+EMIT_RRR (spe_mpya,    0x00c);
+EMIT_RR  (spe_mpyh,    0x3c5);
+EMIT_RR  (spe_mpys,    0x3c7);
+EMIT_RR  (spe_mpyhh,   0x3c6);
+EMIT_RR  (spe_mpyhha,  0x346);
+EMIT_RR  (spe_mpyhhu,  0x3ce);
+EMIT_RR  (spe_mpyhhau, 0x34e);
+EMIT_R   (spe_clz,     0x2a5);
+EMIT_R   (spe_cntb,    0x2b4);
+EMIT_R   (spe_fsmb,    0x1b6);
+EMIT_R   (spe_fsmh,    0x1b5);
+EMIT_R   (spe_fsm,     0x1b4);
+EMIT_R   (spe_gbb,     0x1b2);
+EMIT_R   (spe_gbh,     0x1b1);
+EMIT_R   (spe_gb,      0x1b0);
+EMIT_RR  (spe_avgb,    0x0d3);
+EMIT_RR  (spe_absdb,   0x053);
+EMIT_RR  (spe_sumb,    0x253);
+EMIT_R   (spe_xsbh,    0x2b6);
+EMIT_R   (spe_xshw,    0x2ae);
+EMIT_R   (spe_xswd,    0x2a6);
+EMIT_RR  (spe_and,     0x0c1);
+EMIT_RR  (spe_andc,    0x2c1);
+EMIT_RI10(spe_andbi,   0x016);
+EMIT_RI10(spe_andhi,   0x015);
+EMIT_RI10(spe_andi,    0x014);
+EMIT_RR  (spe_or,      0x041);
+EMIT_RR  (spe_orc,     0x2c9);
+EMIT_RI10(spe_orbi,    0x006);
+EMIT_RI10(spe_orhi,    0x005);
+EMIT_RI10(spe_ori,     0x004);
+EMIT_R   (spe_orx,     0x1f0);
+EMIT_RR  (spe_xor,     0x241);
+EMIT_RI10(spe_xorbi,   0x026);
+EMIT_RI10(spe_xorhi,   0x025);
+EMIT_RI10(spe_xori,    0x024);
+EMIT_RR  (spe_nand,    0x0c9);
+EMIT_RR  (spe_nor,     0x049);
+EMIT_RR  (spe_eqv,     0x249);
+EMIT_RRR (spe_selb,    0x008);
+EMIT_RRR (spe_shufb,   0x00b);
+
+
+/* Shift and rotate instructions
+ */
+EMIT_RR  (spe_shlh,      0x05f);
+EMIT_RI7 (spe_shlhi,     0x07f);
+EMIT_RR  (spe_shl,       0x05b);
+EMIT_RI7 (spe_shli,      0x07b);
+EMIT_RR  (spe_shlqbi,    0x1db);
+EMIT_RI7 (spe_shlqbii,   0x1fb);
+EMIT_RR  (spe_shlqby,    0x1df);
+EMIT_RI7 (spe_shlqbyi,   0x1ff);
+EMIT_RR  (spe_shlqbybi,  0x1cf);
+EMIT_RR  (spe_roth,      0x05c);
+EMIT_RI7 (spe_rothi,     0x07c);
+EMIT_RR  (spe_rot,       0x058);
+EMIT_RI7 (spe_roti,      0x078);
+EMIT_RR  (spe_rotqby,    0x1dc);
+EMIT_RI7 (spe_rotqbyi,   0x1fc);
+EMIT_RR  (spe_rotqbybi,  0x1cc);
+EMIT_RR  (spe_rotqbi,    0x1d8);
+EMIT_RI7 (spe_rotqbii,   0x1f8);
+EMIT_RR  (spe_rothm,     0x05d);
+EMIT_RI7 (spe_rothmi,    0x07d);
+EMIT_RR  (spe_rotm,      0x059);
+EMIT_RI7 (spe_rotmi,     0x079);
+EMIT_RR  (spe_rotqmby,   0x1dd);
+EMIT_RI7 (spe_rotqmbyi,  0x1fd);
+EMIT_RR  (spe_rotqmbybi, 0x1cd);
+EMIT_RR  (spe_rotqmbi,   0x1c9);
+EMIT_RI7 (spe_rotqmbii,  0x1f9);
+EMIT_RR  (spe_rotmah,    0x05e);
+EMIT_RI7 (spe_rotmahi,   0x07e);
+EMIT_RR  (spe_rotma,     0x05a);
+EMIT_RI7 (spe_rotmai,    0x07a);
+
+
+/* Compare, branch, and halt instructions
+ */
+EMIT_RR  (spe_heq,       0x3d8);
+EMIT_RI10(spe_heqi,      0x07f);
+EMIT_RR  (spe_hgt,       0x258);
+EMIT_RI10(spe_hgti,      0x04f);
+EMIT_RR  (spe_hlgt,      0x2d8);
+EMIT_RI10(spe_hlgti,     0x05f);
+EMIT_RR  (spe_ceqb,      0x3d0);
+EMIT_RI10(spe_ceqbi,     0x07e);
+EMIT_RR  (spe_ceqh,      0x3c8);
+EMIT_RI10(spe_ceqhi,     0x07d);
+EMIT_RR  (spe_ceq,       0x3c0);
+EMIT_RI10(spe_ceqi,      0x07c);
+EMIT_RR  (spe_cgtb,      0x250);
+EMIT_RI10(spe_cgtbi,     0x04e);
+EMIT_RR  (spe_cgth,      0x248);
+EMIT_RI10(spe_cgthi,     0x04d);
+EMIT_RR  (spe_cgt,       0x240);
+EMIT_RI10(spe_cgti,      0x04c);
+EMIT_RR  (spe_clgtb,     0x2d0);
+EMIT_RI10(spe_clgtbi,    0x05e);
+EMIT_RR  (spe_clgth,     0x2c8);
+EMIT_RI10(spe_clgthi,    0x05d);
+EMIT_RR  (spe_clgt,      0x2c0);
+EMIT_RI10(spe_clgti,     0x05c);
+EMIT_I16 (spe_br,        0x064);
+EMIT_I16 (spe_bra,       0x060);
+EMIT_RI16(spe_brsl,      0x066);
+EMIT_RI16(spe_brasl,     0x062);
+EMIT_RI16(spe_brnz,      0x042);
+EMIT_RI16(spe_brz,       0x040);
+EMIT_RI16(spe_brhnz,     0x046);
+EMIT_RI16(spe_brhz,      0x044);
+
+extern void spe_bi(struct spe_function *p, unsigned rA, int d, int e);
+extern void spe_iret(struct spe_function *p, unsigned rA, int d, int e);
+extern void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA,
+    int d, int e);
+extern void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA,
+    int d, int e);
+extern void spe_biz(struct spe_function *p, unsigned rT, unsigned rA,
+    int d, int e);
+extern void spe_binz(struct spe_function *p, unsigned rT, unsigned rA,
+    int d, int e);
+extern void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA,
+    int d, int e);
+extern void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA,
+    int d, int e);
+
+
+/* Floating-point instructions
+ */
+EMIT_RR  (spe_fa,         0x2c4);
+EMIT_RR  (spe_dfa,        0x2cc);
+EMIT_RR  (spe_fs,         0x2c5);
+EMIT_RR  (spe_dfs,        0x2cd);
+EMIT_RR  (spe_fm,         0x2c6);
+EMIT_RR  (spe_dfm,        0x2ce);
+EMIT_RRR (spe_fma,        0x00e);
+EMIT_RR  (spe_dfma,       0x35c);
+EMIT_RRR (spe_fnms,       0x00d);
+EMIT_RR  (spe_dfnms,      0x35e);
+EMIT_RRR (spe_fms,        0x00f);
+EMIT_RR  (spe_dfms,       0x35d);
+EMIT_RR  (spe_dfnma,      0x35f);
+EMIT_R   (spe_frest,      0x1b8);
+EMIT_R   (spe_frsqest,    0x1b9);
+EMIT_RR  (spe_fi,         0x3d4);
+EMIT_RI8 (spe_csflt,      0x1da);
+EMIT_RI8 (spe_cflts,      0x1d8);
+EMIT_RI8 (spe_cuflt,      0x1db);
+EMIT_RI8 (spe_cfltu,      0x1d9);
+EMIT_R   (spe_frds,       0x3b9);
+EMIT_R   (spe_fesd,       0x3b8);
+EMIT_RR  (spe_dfceq,      0x3c3);
+EMIT_RR  (spe_dfcmeq,     0x3cb);
+EMIT_RR  (spe_dfcgt,      0x2c3);
+EMIT_RR  (spe_dfcmgt,     0x2cb);
+EMIT_RI7 (spe_dftsv,      0x3bf);
+EMIT_RR  (spe_fceq,       0x3c2);
+EMIT_RR  (spe_fcmeq,      0x3ca);
+EMIT_RR  (spe_fcgt,       0x2c2);
+EMIT_RR  (spe_fcmgt,      0x2ca);
+EMIT_R   (spe_fscrwr,     0x3ba);
+EMIT_    (spe_fscrrd,     0x398);
+
+
+/* Channel instructions
+ */
+EMIT_R   (spe_rdch,       0x00d);
+EMIT_R   (spe_rdchcnt,    0x00f);
+EMIT_R   (spe_wrch,       0x10d);
+
+
+#ifdef UNDEF_EMIT_MACROS
+#undef EMIT_
+#undef EMIT_R
+#undef EMIT_RR
+#undef EMIT_RRR
+#undef EMIT_RI7
+#undef EMIT_RI8
+#undef EMIT_RI10
+#undef EMIT_RI16
+#undef EMIT_RI18
+#undef EMIT_I16
+#undef UNDEF_EMIT_MACROS
+#endif /* EMIT_ */
diff --git a/src/mesa/sources b/src/mesa/sources
index 96ae3dbca00..f83d247a1e2 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources
@@ -158,45 +158,48 @@ VF_SOURCES = \
 
 
 DRAW_SOURCES = \
-	pipe/draw/draw_clip.c \
-	pipe/draw/draw_context.c\
-	pipe/draw/draw_cull.c \
-	pipe/draw/draw_debug.c \
-	pipe/draw/draw_flatshade.c \
-	pipe/draw/draw_offset.c \
-	pipe/draw/draw_prim.c \
-	pipe/draw/draw_stipple.c \
-	pipe/draw/draw_twoside.c \
-	pipe/draw/draw_unfilled.c \
-	pipe/draw/draw_validate.c \
-	pipe/draw/draw_vbuf.c \
-	pipe/draw/draw_vertex.c \
-	pipe/draw/draw_vertex_cache.c \
-	pipe/draw/draw_vertex_fetch.c \
-	pipe/draw/draw_vertex_shader.c \
-	pipe/draw/draw_vf.c \
-	pipe/draw/draw_vf_generic.c \
-	pipe/draw/draw_vf_sse.c \
-	pipe/draw/draw_wide_prims.c
+	$(TOP)/src/gallium/auxiliary/draw/draw_clip.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vs_exec.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vs_sse.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vs_llvm.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_context.c\
+	$(TOP)/src/gallium/auxiliary/draw/draw_cull.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_debug.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_flatshade.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_offset.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_prim.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_stipple.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_twoside.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_unfilled.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_validate.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vbuf.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vertex.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vertex_cache.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vertex_fetch.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vertex_shader.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vf.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vf_generic.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_vf_sse.c \
+	$(TOP)/src/gallium/auxiliary/draw/draw_wide_prims.c
 
 TGSIEXEC_SOURCES = \
-	pipe/tgsi/exec/tgsi_exec.c \
-	pipe/tgsi/exec/tgsi_sse2.c
+	$(TOP)/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c \
+	$(TOP)/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
 
 TGSIUTIL_SOURCES = \
-	pipe/tgsi/util/tgsi_build.c \
-	pipe/tgsi/util/tgsi_dump.c \
-	pipe/tgsi/util/tgsi_parse.c \
-	pipe/tgsi/util/tgsi_util.c
+	$(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_build.c \
+	$(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_dump.c \
+	$(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_parse.c \
+	$(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_util.c
 
 STATECACHE_SOURCES = \
-	pipe/cso_cache/cso_hash.c \
-	pipe/cso_cache/cso_cache.c
+	$(TOP)/src/gallium/auxiliary/cso_cache/cso_hash.c \
+	$(TOP)/src/gallium/auxiliary/cso_cache/cso_cache.c
 
 PIPEUTIL_SOURCES = \
-	pipe/util/p_debug.c \
-	pipe/util/p_tile.c \
-	pipe/util/p_util.c
+	$(TOP)/src/gallium/auxiliary/util/p_debug.c \
+	$(TOP)/src/gallium/auxiliary/util/p_tile.c \
+	$(TOP)/src/gallium/auxiliary/util/p_util.c
 
 STATETRACKER_SOURCES = \
 	state_tracker/st_atom.c \
@@ -288,6 +291,7 @@ ASM_C_SOURCES =	\
 	x86/rtasm/x86sse.c \
 	sparc/sparc.c \
 	ppc/common_ppc.c \
+	ppc/rtasm/spe_asm.c \
 	x86-64/x86-64.c
 
 X86_SOURCES =			\
@@ -330,13 +334,13 @@ __COMMON_DRIVER_SOURCES =			\
 	drivers/common/driverfuncs.c
 
 X11_DRIVER_SOURCES =		\
-	pipe/xlib/glxapi.c	\
-	pipe/xlib/fakeglx.c	\
-	pipe/xlib/xfonts.c	\
-	pipe/xlib/xm_api.c	\
-	pipe/xlib/xm_winsys.c	\
-	pipe/xlib/xm_winsys_aub.c	\
-	pipe/xlib/brw_aub.c
+	$(TOP)/src/gallium/winsys/xlib/glxapi.c	\
+	$(TOP)/src/gallium/winsys/xlib/fakeglx.c	\
+	$(TOP)/src/gallium/winsys/xlib/xfonts.c	\
+	$(TOP)/src/gallium/winsys/xlib/xm_api.c	\
+	$(TOP)/src/gallium/winsys/xlib/xm_winsys.c	\
+	$(TOP)/src/gallium/winsys/xlib/xm_winsys_aub.c	\
+	$(TOP)/src/gallium/winsys/xlib/brw_aub.c
 
 OSMESA_DRIVER_SOURCES = \
 	drivers/osmesa/osmesa.c
@@ -424,7 +428,10 @@ FBDEV_DRIVER_OBJECTS = $(FBDEV_DRIVER_SOURCES:.c=.o)
 INCLUDE_DIRS = \
 	-I$(TOP)/include \
 	-I$(TOP)/src/mesa \
-	-I$(TOP)/src/mesa/main
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary
 
 OLD_INCLUDE_DIRS = \
 	-I$(TOP)/src/mesa/tnl \
@@ -434,4 +441,4 @@ OLD_INCLUDE_DIRS = \
 	-I$(TOP)/src/mesa/shader \
 	-I$(TOP)/src/mesa/shader/grammar \
 	-I$(TOP)/src/mesa/shader/slang \
-	-I$(TOP)/src/mesa/pipe/tgsi
+	-I$(TOP)/s$(TOP)/src/gallium/auxiliary/tgsi
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 2c6ec8421b0..b67b620eaa5 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -43,7 +43,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_shader_tokens.h"
 
-#include "pipe/cso_cache/cso_cache.h"
+#include "cso_cache/cso_cache.h"
 
 #include "st_context.h"
 #include "st_cache.h"
diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c
index e0965b217ab..2979e7fae54 100644
--- a/src/mesa/state_tracker/st_cache.c
+++ b/src/mesa/state_tracker/st_cache.c
@@ -36,8 +36,8 @@
 
 #include "pipe/p_state.h"
 
-#include "pipe/cso_cache/cso_cache.h"
-#include "pipe/cso_cache/cso_hash.h"
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_hash.h"
 
 
 /* Those function will either find the state of the given template
diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h
index e0c176b0ffc..b81de316ec9 100644
--- a/src/mesa/state_tracker/st_cache.h
+++ b/src/mesa/state_tracker/st_cache.h
@@ -33,7 +33,7 @@
 #ifndef ST_CACHE_H
 #define ST_CACHE_H
 
-#include "pipe/cso_cache/cso_cache.h"
+#include "cso_cache/cso_cache.h"
 
 struct pipe_blend_state;
 struct pipe_sampler_state;
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
index 3a3bf9016dd..663c4f205d8 100644
--- a/src/mesa/state_tracker/st_cb_accum.c
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -43,7 +43,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 
 
 #define UNCLAMPED_FLOAT_TO_SHORT(us, f)  \
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index 410062e1e80..78baf772f44 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -272,6 +272,14 @@ clear_with_quad(GLcontext *ctx,
    const GLfloat x1 = ctx->DrawBuffer->_Xmax;
    const GLfloat y1 = ctx->DrawBuffer->_Ymax;
 
+   /*
+   printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__, 
+	  color ? "color, " : "",
+	  depth ? "depth, " : "",
+	  stencil ? "stencil" : "",
+	  x0, y0,
+	  x1, y1);
+   */
 
    /* blend state: RGBA masking */
    {
@@ -389,13 +397,44 @@ clear_with_quad(GLcontext *ctx,
  * Determine if we need to clear the depth buffer by drawing a quad.
  */
 static INLINE GLboolean
-check_clear_color_with_quad(GLcontext *ctx)
+check_clear_color_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (!ctx->Color.ColorMask[0] ||
+       !ctx->Color.ColorMask[1] ||
+       !ctx->Color.ColorMask[2] ||
+       !ctx->Color.ColorMask[3])
+      return TRUE;
+
+   return FALSE;
+}
+
+
+static INLINE GLboolean
+check_clear_depth_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
-   return !(ctx->Color.ColorMask[0] &&
-            ctx->Color.ColorMask[1] &&
-            ctx->Color.ColorMask[2] &&
-            ctx->Color.ColorMask[3] &&
-            !ctx->Scissor.Enabled);
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
+   GLboolean maskStencil
+      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (maskStencil)
+      return TRUE;
+
+   return FALSE;
 }
 
 
@@ -407,10 +446,19 @@ check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    const struct st_renderbuffer *strb = st_renderbuffer(rb);
    const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
-   return  ctx->Scissor.Enabled
-      || (isDS && 
-	  strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
-	  ctx->DrawBuffer->Visual.stencilBits > 0);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (isDS && 
+       strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
+       ctx->DrawBuffer->Visual.stencilBits > 0)
+      return TRUE;
+
+   return FALSE;
 }
 
 
@@ -425,9 +473,27 @@ check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
    const GLuint stencilMax = (1 << rb->StencilBits) - 1;
    const GLboolean maskStencil
       = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
-   return maskStencil
-      || ctx->Scissor.Enabled
-      || (isDS && ctx->DrawBuffer->Visual.depthBits > 0);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (maskStencil) 
+      return TRUE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   /* This is correct, but it is necessary to look at the depth clear
+    * value held in the surface when it comes time to issue the clear,
+    * rather than taking depth and stencil clear values from the
+    * current state.
+    */
+   if (isDS && 
+       strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
+       ctx->DrawBuffer->Visual.depthBits > 0)
+      return TRUE;
+
+   return FALSE;
 }
 
 
@@ -436,14 +502,10 @@ check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 static void
 clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
-   struct st_renderbuffer *strb = st_renderbuffer(rb);
-
-   if (ctx->Color.ColorMask[0] &&
-       ctx->Color.ColorMask[1] &&
-       ctx->Color.ColorMask[2] &&
-       ctx->Color.ColorMask[3] &&
-       !ctx->Scissor.Enabled)
+   if (!check_clear_color_with_quad( ctx, rb ))
    {
+      struct st_renderbuffer *strb = st_renderbuffer(rb);
+
       /* clear whole buffer w/out masking */
       GLuint clearValue
          = color_value(strb->surface->format, ctx->Color.ClearColor);
@@ -482,14 +544,8 @@ static void
 clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    struct st_renderbuffer *strb = st_renderbuffer(rb);
-   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
-   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
-   GLboolean maskStencil
-      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
 
-   if (maskStencil ||
-       ctx->Scissor.Enabled ||
-       (isDS && ctx->DrawBuffer->Visual.depthBits > 0)) {
+   if (check_clear_stencil_with_quad(ctx, rb)) {
       /* masking or scissoring or combined depth/stencil buffer */
       clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE);
    }
@@ -505,13 +561,10 @@ static void
 clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    struct st_renderbuffer *strb = st_renderbuffer(rb);
-   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
-   GLboolean maskStencil
-      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
 
    assert(is_depth_stencil_format(strb->surface->format));
 
-   if (!maskStencil && !ctx->Scissor.Enabled) {
+   if (check_clear_depth_stencil_with_quad(ctx, rb)) {
       /* clear whole buffer w/out masking */
       GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear);
 
@@ -520,7 +573,7 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
          clearValue |= ctx->Stencil.Clear << 24;
          break;
       case PIPE_FORMAT_Z24S8_UNORM:
-         clearValue |= clearValue | ctx->Stencil.Clear;
+         clearValue |= ctx->Stencil.Clear;
          break;
       default:
          assert(0);
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 3245a7488bc..e2d4e06da19 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -56,7 +56,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 #include "shader/prog_instruction.h"
 
 
@@ -1259,6 +1259,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    if (st->haveFramebufferSurfaces) {
       /* copy source framebuffer surface into mipmap/texture */
       pipe->surface_copy(pipe,
+                         FALSE,
 			 psTex, /* dest */
 			 0, 0, /* destx/y */
 			 psRead,
diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c
index 31744151f1d..5315294c07b 100644
--- a/src/mesa/state_tracker/st_cb_feedback.c
+++ b/src/mesa/state_tracker/st_cb_feedback.c
@@ -53,10 +53,10 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/cso_cache/cso_cache.h"
+#include "cso_cache/cso_cache.h"
 
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_private.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
 
 
 /**
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
index af3ee655048..61d4f4c41c6 100644
--- a/src/mesa/state_tracker/st_cb_program.c
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -39,8 +39,8 @@
 #include "shader/programopt.h"
 #include "shader/shader_api.h"
 
-#include "pipe/cso_cache/cso_cache.h"
-#include "pipe/draw/draw_context.h"
+#include "cso_cache/cso_cache.h"
+#include "draw/draw_context.h"
 
 #include "st_context.h"
 #include "st_program.h"
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
index 7e347c48938..5b0eb6022be 100644
--- a/src/mesa/state_tracker/st_cb_rasterpos.c
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -44,8 +44,8 @@
 #include "st_draw.h"
 #include "st_cb_rasterpos.h"
 #include "st_draw.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/draw/draw_private.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
 #include "shader/prog_instruction.h"
 #include "vbo/vbo.h"
 
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index 868c5f3c5f9..c89c74229e5 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -40,7 +40,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 #include "st_context.h"
 #include "st_cb_readpixels.h"
 #include "st_cb_fbo.h"
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 0ea367549b8..03dbb30b0fe 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -47,7 +47,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
-#include "pipe/util/p_tile.h"
+#include "util/p_tile.h"
 
 
 #define DBG if (0) printf
@@ -1132,6 +1132,7 @@ do_copy_texsubimage(GLcontext *ctx,
    struct pipe_context *pipe = ctx->st->pipe;
    struct pipe_surface *dest_surface;
    uint dest_format, src_format;
+   uint do_flip = FALSE;
 
    (void) texImage;
 
@@ -1153,6 +1154,7 @@ do_copy_texsubimage(GLcontext *ctx,
 
    if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
       srcY = strb->Base.Height - srcY - height;
+      do_flip = TRUE;
    }
 
    src_format = strb->surface->format;
@@ -1190,6 +1192,7 @@ do_copy_texsubimage(GLcontext *ctx,
 #else
 
       pipe->surface_copy(pipe,
+                         do_flip,
 			 /* dest */
 			 dest_surface,
 			 destX, destY,
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index bf4618bed8e..09e389f9dc7 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -54,8 +54,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_winsys.h"
 #include "pipe/p_inlines.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/cso_cache/cso_cache.h"
+#include "draw/draw_context.h"
+#include "cso_cache/cso_cache.h"
 
 
 /**
diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c
index 57450e52bf4..5888bcb98a3 100644
--- a/src/mesa/state_tracker/st_debug.c
+++ b/src/mesa/state_tracker/st_debug.c
@@ -31,9 +31,9 @@
 
 #include "pipe/p_state.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
+#include "tgsi/util/tgsi_dump.h"
 
-#include "pipe/cso_cache/cso_cache.h"
+#include "cso_cache/cso_cache.h"
 
 #include "st_context.h"
 #include "st_debug.h"
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index ae9f5c8b117..1c0fa8c6aad 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -47,8 +47,8 @@
 #include "pipe/p_winsys.h"
 #include "pipe/p_inlines.h"
 
-#include "pipe/draw/draw_private.h"
-#include "pipe/draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_context.h"
 
 
 static GLuint double_types[4] = {
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
index 459941cca87..6c09b86033b 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.c
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -37,7 +37,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "pipe/cso_cache/cso_cache.h"
+#include "cso_cache/cso_cache.h"
 
 #include "st_context.h"
 #include "st_draw.h"
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index 325aa201734..97206752af3 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -32,9 +32,9 @@
 
 #include "pipe/p_compiler.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/tgsi/util/tgsi_parse.h"
-#include "pipe/tgsi/util/tgsi_build.h"
-#include "pipe/tgsi/util/tgsi_util.h"
+#include "tgsi/util/tgsi_parse.h"
+#include "tgsi/util/tgsi_build.h"
+#include "tgsi/util/tgsi_util.h"
 #include "st_mesa_to_tgsi.h"
 #include "shader/prog_instruction.h"
 #include "shader/prog_parameter.h"
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index c8297badedd..dc992ee9c24 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -38,8 +38,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
-#include "pipe/draw/draw_context.h"
-#include "pipe/tgsi/util/tgsi_dump.h"
+#include "draw/draw_context.h"
+#include "tgsi/util/tgsi_dump.h"
 
 #include "st_context.h"
 #include "st_cache.h"
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
index 2622d009530..b86f416c9b4 100644
--- a/src/mesa/state_tracker/st_texture.c
+++ b/src/mesa/state_tracker/st_texture.c
@@ -300,6 +300,7 @@ st_texture_image_copy(struct pipe_context *pipe,
       src_surface = pipe->get_tex_surface(pipe, src, face, srcLevel, i);
 
       pipe->surface_copy(pipe,
+                         FALSE,
 			 dst_surface,
 			 0, 0, /* destX, Y */
 			 src_surface,
diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c
index f8da6e405f4..39c0e9b851a 100644
--- a/src/mesa/x86/rtasm/x86sse.c
+++ b/src/mesa/x86/rtasm/x86sse.c
@@ -12,55 +12,73 @@ static unsigned char *cptr( void (*label)() )
 }
 
 
+static void do_realloc( struct x86_function *p )
+{
+   if (p->size == 0) {
+      p->size = 1024;
+      p->store = _mesa_exec_malloc(p->size);
+      p->csr = p->store;
+   }
+   else {
+      unsigned used = p->csr - p->store;
+      unsigned char *tmp = p->store;
+      p->size *= 2;
+      p->store = _mesa_exec_malloc(p->size);
+      memcpy(p->store, tmp, used);
+      p->csr = p->store + used;
+      _mesa_exec_free(tmp);
+   }
+}
+
 /* Emit bytes to the instruction stream:
  */
-static void emit_1b( struct x86_function *p, char b0 )
+static unsigned char *reserve( struct x86_function *p, int bytes )
 {
-   *(char *)(p->csr++) = b0;
+   if (p->csr + bytes - p->store > p->size)
+      do_realloc(p);
+
+   {
+      unsigned char *csr = p->csr;
+      p->csr += bytes;
+      return csr;
+   }
 }
 
-static void emit_1i( struct x86_function *p, int i0 )
+
+
+static void emit_1b( struct x86_function *p, char b0 )
 {
-   *(int *)(p->csr) = i0;
-   p->csr += 4;
+   char *csr = (char *)reserve(p, 1);
+   *csr = b0;
 }
 
-static void disassem( struct x86_function *p, const char *fn )
+static void emit_1i( struct x86_function *p, int i0 )
 {
-#if DISASSEM && 0
-   if (fn && fn != p->fn) {
-      _mesa_printf("0x%x: %s\n", p->csr, fn);
-      p->fn = fn;
-   }
-#endif
+   int *icsr = (int *)reserve(p, sizeof(i0));
+   *icsr = i0;
 }
 
-static void emit_1ub_fn( struct x86_function *p, unsigned char b0, const char *fn )
+static void emit_1ub( struct x86_function *p, unsigned char b0 )
 {
-   disassem(p, fn);
-   *(p->csr++) = b0;
+   unsigned char *csr = reserve(p, 1);
+   *csr++ = b0;
 }
 
-static void emit_2ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, const char *fn )
+static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 )
 {
-   disassem(p, fn);
-   *(p->csr++) = b0;
-   *(p->csr++) = b1;
+   unsigned char *csr = reserve(p, 2);
+   *csr++ = b0;
+   *csr++ = b1;
 }
 
-static void emit_3ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2, const char *fn )
+static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 )
 {
-   disassem(p, fn);
-   *(p->csr++) = b0;
-   *(p->csr++) = b1;
-   *(p->csr++) = b2;
+   unsigned char *csr = reserve(p, 3);
+   *csr++ = b0;
+   *csr++ = b1;
+   *csr++ = b2;
 }
 
-#define emit_1ub(p, b0)         emit_1ub_fn(p, b0, __FUNCTION__)
-#define emit_2ub(p, b0, b1)     emit_2ub_fn(p, b0, b1, __FUNCTION__)
-#define emit_3ub(p, b0, b1, b2) emit_3ub_fn(p, b0, b1, b2, __FUNCTION__)
-
-
 
 /* Build a modRM byte + possible displacement.  No treatment of SIB
  * indexing.  BZZT - no way to encode an absolute address.
@@ -77,13 +95,13 @@ static void emit_modrm( struct x86_function *p,
    val |= reg.idx << 3;		/* reg field */
    val |= regmem.idx;		/* r/m field */
    
-   emit_1ub_fn(p, val, 0);
+   emit_1ub(p, val);
 
    /* Oh-oh we've stumbled into the SIB thing.
     */
    if (regmem.file == file_REG32 &&
        regmem.idx == reg_SP) {
-      emit_1ub_fn(p, 0x24, 0);		/* simplistic! */
+      emit_1ub(p, 0x24);		/* simplistic! */
    }
 
    switch (regmem.mod) {
@@ -124,14 +142,14 @@ static void emit_op_modrm( struct x86_function *p,
 {  
    switch (dst.mod) {
    case mod_REG:
-      emit_1ub_fn(p, op_dst_is_reg, 0);
+      emit_1ub(p, op_dst_is_reg);
       emit_modrm(p, dst, src);
       break;
    case mod_INDIRECT:
    case mod_DISP32:
    case mod_DISP8:
       assert(src.mod == mod_REG);
-      emit_1ub_fn(p, op_dst_is_mem, 0);
+      emit_1ub(p, op_dst_is_mem);
       emit_modrm(p, src, dst);
       break;
    default:
@@ -258,11 +276,24 @@ void x86_jmp( struct x86_function *p, unsigned char *label)
    emit_1i(p, label - x86_get_label(p) - 4);
 }
 
+#if 0
+/* This doesn't work once we start reallocating & copying the
+ * generated code on buffer fills, because the call is relative to the
+ * current pc.
+ */
 void x86_call( struct x86_function *p, void (*label)())
 {
    emit_1ub(p, 0xe8);
    emit_1i(p, cptr(label) - x86_get_label(p) - 4);
 }
+#else
+void x86_call( struct x86_function *p, struct x86_reg reg)
+{
+   emit_1ub(p, 0xff);
+   emit_modrm(p, reg, reg);
+}
+#endif
+
 
 /* michal:
  * Temporary. As I need immediate operands, and dont want to mess with the codegen,
@@ -789,7 +820,7 @@ static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86
 	 assert(0);
    }
    else if (dst.idx == 0) {
-      assert(arg.file = file_REG32);
+      assert(arg.file == file_REG32);
       emit_1ub(p, 0xd8);
       emit_modrm_noreg(p, argmem_noreg, arg);
    }
@@ -1125,11 +1156,14 @@ struct x86_reg x86_fn_arg( struct x86_function *p,
 
 void x86_init_func( struct x86_function *p )
 {
-   x86_init_func_size(p, 2048);
+   p->size = 0;
+   p->store = NULL;
+   p->csr = p->store;
 }
 
 void x86_init_func_size( struct x86_function *p, unsigned code_size )
 {
+   p->size = code_size;
    p->store = _mesa_exec_malloc(code_size);
    p->csr = p->store;
 }
@@ -1138,12 +1172,14 @@ void x86_release_func( struct x86_function *p )
 {
    _mesa_exec_free(p->store);
    p->store = NULL;
+   p->csr = NULL;
+   p->size = 0;
 }
 
 
 void (*x86_get_func( struct x86_function *p ))(void)
 {
-   if (DISASSEM)
+   if (DISASSEM && p->store)
       _mesa_printf("disassemble %p %p\n", p->store, p->csr);
    return (void (*)(void)) (unsigned long) p->store;
 }
diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h
index 63b9a36392a..c2aa416492e 100644
--- a/src/mesa/x86/rtasm/x86sse.h
+++ b/src/mesa/x86/rtasm/x86sse.h
@@ -16,6 +16,7 @@ struct x86_reg {
 };
 
 struct x86_function {
+   unsigned size;
    unsigned char *store;
    unsigned char *csr;
    unsigned stack_offset;
@@ -118,7 +119,8 @@ void x86_fixup_fwd_jump( struct x86_function *p,
 
 void x86_jmp( struct x86_function *p, unsigned char *label );
 
-void x86_call( struct x86_function *p, void (*label)() );
+/* void x86_call( struct x86_function *p, void (*label)() ); */
+void x86_call( struct x86_function *p, struct x86_reg reg);
 
 /* michal:
  * Temporary. As I need immediate operands, and dont want to mess with the codegen,