Merge branch 'master' of ssh://git.freedesktop.org/git/mesa/mesa into pipe-video

Conflicts:
	src/gallium/auxiliary/Makefile
	src/gallium/auxiliary/SConscript
	src/gallium/auxiliary/util/u_format.csv
	src/gallium/auxiliary/vl/vl_compositor.c
	src/gallium/auxiliary/vl/vl_compositor.h
	src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
	src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h
	src/gallium/drivers/identity/id_objects.c
	src/gallium/drivers/identity/id_objects.h
	src/gallium/drivers/identity/id_screen.c
	src/gallium/drivers/nv40/Makefile
	src/gallium/drivers/nv40/nv40_screen.c
	src/gallium/drivers/softpipe/sp_texture.c
	src/gallium/drivers/softpipe/sp_texture.h
	src/gallium/drivers/softpipe/sp_video_context.c
	src/gallium/drivers/softpipe/sp_video_context.h
	src/gallium/include/pipe/p_format.h
	src/gallium/include/pipe/p_screen.h
	src/gallium/include/pipe/p_video_context.h
	src/gallium/include/pipe/p_video_state.h
	src/gallium/include/state_tracker/dri1_api.h
	src/gallium/include/state_tracker/drm_api.h
	src/gallium/state_trackers/dri/common/dri_context.c
	src/gallium/state_trackers/xorg/xvmc/attributes.c
	src/gallium/state_trackers/xorg/xvmc/block.c
	src/gallium/state_trackers/xorg/xvmc/context.c
	src/gallium/state_trackers/xorg/xvmc/subpicture.c
	src/gallium/state_trackers/xorg/xvmc/surface.c
	src/gallium/state_trackers/xorg/xvmc/tests/.gitignore
	src/gallium/state_trackers/xorg/xvmc/tests/Makefile
	src/gallium/state_trackers/xorg/xvmc/xvmc_private.h
	src/gallium/winsys/drm/radeon/core/radeon_drm.c
	src/gallium/winsys/g3dvl/vl_winsys.h
	src/gallium/winsys/g3dvl/xlib/xsp_winsys.c
	src/gallium/winsys/sw/Makefile

37 files changed, 2581 insertions, 579 deletions

diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index bb0988543f5..02abddf1491 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -31,19 +31,49 @@
   */
 
 
+#include "pipe/p_context.h"
 #include "util/u_memory.h"
 #include "util/u_math.h"
 #include "draw_context.h"
 #include "draw_vs.h"
 #include "draw_gs.h"
 
+#if HAVE_LLVM
+#include "gallivm/lp_bld_init.h"
+#endif
 
-struct draw_context *draw_create( void )
+struct draw_context *draw_create( struct pipe_context *pipe )
 {
    struct draw_context *draw = CALLOC_STRUCT( draw_context );
    if (draw == NULL)
       goto fail;
 
+#if HAVE_LLVM
+   lp_build_init();
+   assert(lp_build_engine);
+   draw->engine = lp_build_engine;
+#endif
+
+   if (!draw_init(draw))
+      goto fail;
+
+   draw->pipe = pipe;
+
+   return draw;
+
+fail:
+   draw_destroy( draw );
+   return NULL;
+}
+
+boolean draw_init(struct draw_context *draw)
+{
+   /*
+    * Note that several functions compute the clipmask of the predefined
+    * formats with hardcoded formulas instead of using these. So modifications
+    * here must be reflected there too.
+    */
+
    ASSIGN_4V( draw->plane[0], -1,  0,  0, 1 );
    ASSIGN_4V( draw->plane[1],  1,  0,  0, 1 );
    ASSIGN_4V( draw->plane[2],  0, -1,  0, 1 );
@@ -57,31 +87,40 @@ struct draw_context *draw_create( void )
 
 
    if (!draw_pipeline_init( draw ))
-      goto fail;
+      return FALSE;
 
    if (!draw_pt_init( draw ))
-      goto fail;
+      return FALSE;
 
    if (!draw_vs_init( draw ))
-      goto fail;
+      return FALSE;
 
    if (!draw_gs_init( draw ))
-      goto fail;
+      return FALSE;
 
-   return draw;
-
-fail:
-   draw_destroy( draw );   
-   return NULL;
+   return TRUE;
 }
 
 
 void draw_destroy( struct draw_context *draw )
 {
+   struct pipe_context *pipe;
+   int i, j;
+
    if (!draw)
       return;
 
+   pipe = draw->pipe;
 
+   /* free any rasterizer CSOs that we may have created.
+    */
+   for (i = 0; i < 2; i++) {
+      for (j = 0; j < 2; j++) {
+         if (draw->rasterizer_no_cull[i][j]) {
+            pipe->delete_rasterizer_state(pipe, draw->rasterizer_no_cull[i][j]);
+         }
+      }
+   }
 
    /* Not so fast -- we're just borrowing this at the moment.
     * 
@@ -123,12 +162,17 @@ void draw_set_mrd(struct draw_context *draw, double mrd)
  * This causes the drawing pipeline to be rebuilt.
  */
 void draw_set_rasterizer_state( struct draw_context *draw,
-                                const struct pipe_rasterizer_state *raster )
+                                const struct pipe_rasterizer_state *raster,
+                                void *rast_handle )
 {
-   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   if (!draw->suspend_flushing) {
+      draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
 
-   draw->rasterizer = raster;
-   draw->bypass_clipping = draw->driver.bypass_clipping;
+      draw->rasterizer = raster;
+      draw->rast_handle = rast_handle;
+
+      draw->bypass_clipping = draw->driver.bypass_clipping;
+   }
 }
 
 
@@ -261,6 +305,17 @@ draw_wide_point_threshold(struct draw_context *draw, float threshold)
 
 
 /**
+ * Should the draw module handle point->quad conversion for drawing sprites?
+ */
+void
+draw_wide_point_sprites(struct draw_context *draw, boolean draw_sprite)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->pipeline.wide_point_sprites = draw_sprite;
+}
+
+
+/**
  * Tells the draw module to draw lines with triangles if their width
  * is greater than this threshold.
  */
@@ -413,12 +468,14 @@ void draw_set_render( struct draw_context *draw,
 void
 draw_set_mapped_element_buffer_range( struct draw_context *draw,
                                       unsigned eltSize,
+                                      int eltBias,
                                       unsigned min_index,
                                       unsigned max_index,
                                       const void *elements )
 {
    draw->pt.user.elts = elements;
    draw->pt.user.eltSize = eltSize;
+   draw->pt.user.eltBias = eltBias;
    draw->pt.user.min_index = min_index;
    draw->pt.user.max_index = max_index;
 }
@@ -427,10 +484,12 @@ draw_set_mapped_element_buffer_range( struct draw_context *draw,
 void
 draw_set_mapped_element_buffer( struct draw_context *draw,
                                 unsigned eltSize,
+                                int eltBias,
                                 const void *elements )
 {
    draw->pt.user.elts = elements;
    draw->pt.user.eltSize = eltSize;
+   draw->pt.user.eltBias = eltBias;
    draw->pt.user.min_index = 0;
    draw->pt.user.max_index = 0xffffffff;
 }
@@ -481,3 +540,37 @@ draw_current_shader_position_output(const struct draw_context *draw)
       return draw->gs.position_output;
    return draw->vs.position_output;
 }
+
+
+/**
+ * Return a pointer/handle for a driver/CSO rasterizer object which
+ * disabled culling, stippling, unfilled tris, etc.
+ * This is used by some pipeline stages (such as wide_point, aa_line
+ * and aa_point) which convert points/lines into triangles.  In those
+ * cases we don't want to accidentally cull the triangles.
+ *
+ * \param scissor  should the rasterizer state enable scissoring?
+ * \param flatshade  should the rasterizer state use flat shading?
+ * \return  rasterizer CSO handle
+ */
+void *
+draw_get_rasterizer_no_cull( struct draw_context *draw,
+                             boolean scissor,
+                             boolean flatshade )
+{
+   if (!draw->rasterizer_no_cull[scissor][flatshade]) {
+      /* create now */
+      struct pipe_context *pipe = draw->pipe;
+      struct pipe_rasterizer_state rast;
+
+      memset(&rast, 0, sizeof(rast));
+      rast.scissor = scissor;
+      rast.flatshade = flatshade;
+      rast.front_winding = PIPE_WINDING_CCW;
+      rast.gl_rasterization_rules = draw->rasterizer->gl_rasterization_rules;
+
+      draw->rasterizer_no_cull[scissor][flatshade] =
+         pipe->create_rasterizer_state(pipe, &rast);
+   }
+   return draw->rasterizer_no_cull[scissor][flatshade];
+}
diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index acd81b9712d..b905c2f2da6 100644
--- a/src/gallium/auxiliary/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -40,7 +40,6 @@
 
 #include "pipe/p_state.h"
 
-
 struct pipe_context;
 struct draw_context;
 struct draw_stage;
@@ -49,7 +48,7 @@ struct draw_geometry_shader;
 struct tgsi_sampler;
 
 
-struct draw_context *draw_create( void );
+struct draw_context *draw_create( struct pipe_context *pipe );
 
 void draw_destroy( struct draw_context *draw );
 
@@ -60,13 +59,16 @@ void draw_set_clip_state( struct draw_context *pipe,
                           const struct pipe_clip_state *clip );
 
 void draw_set_rasterizer_state( struct draw_context *draw,
-                                const struct pipe_rasterizer_state *raster );
+                                const struct pipe_rasterizer_state *raster,
+                                void *rast_handle );
 
 void draw_set_rasterize_stage( struct draw_context *draw,
                                struct draw_stage *stage );
 
 void draw_wide_point_threshold(struct draw_context *draw, float threshold);
 
+void draw_wide_point_sprites(struct draw_context *draw, boolean draw_sprite);
+
 void draw_wide_line_threshold(struct draw_context *draw, float threshold);
 
 void draw_enable_line_stipple(struct draw_context *draw, boolean enable);
@@ -140,12 +142,14 @@ void draw_set_vertex_elements(struct draw_context *draw,
 void
 draw_set_mapped_element_buffer_range( struct draw_context *draw,
                                       unsigned eltSize,
+                                      int eltBias,
                                       unsigned min_index,
                                       unsigned max_index,
                                       const void *elements );
 
 void draw_set_mapped_element_buffer( struct draw_context *draw,
                                      unsigned eltSize, 
+                                     int eltBias,
                                      const void *elements );
 
 void draw_set_mapped_vertex_buffer(struct draw_context *draw,
@@ -197,6 +201,4 @@ boolean draw_need_pipeline(const struct draw_context *draw,
                            const struct pipe_rasterizer_state *rasterizer,
                            unsigned prim );
 
-
-
 #endif /* DRAW_CONTEXT_H */
diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
index 7069aa6b181..131deed43e4 100644
--- a/src/gallium/auxiliary/draw/draw_gs.c
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -342,10 +342,10 @@ void draw_geometry_shader_delete(struct draw_geometry_shader *shader)
 void draw_geometry_shader_prepare(struct draw_geometry_shader *shader,
                                   struct draw_context *draw)
 {
-    if (shader->machine->Tokens != shader->state.tokens) {
-       tgsi_exec_machine_bind_shader(shader->machine,
-                                     shader->state.tokens,
-                                     draw->gs.num_samplers,
-                                     draw->gs.samplers);
-    }
+   if (shader && shader->machine->Tokens != shader->state.tokens) {
+      tgsi_exec_machine_bind_shader(shader->machine,
+                                    shader->state.tokens,
+                                    draw->gs.num_samplers,
+                                    draw->gs.samplers);
+   }
 }
diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c
new file mode 100644
index 00000000000..27383221b9b
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_llvm.c
@@ -0,0 +1,855 @@
+#include "draw_llvm.h"
+
+#include "draw_context.h"
+#include "draw_vs.h"
+
+#include "gallivm/lp_bld_arit.h"
+#include "gallivm/lp_bld_struct.h"
+#include "gallivm/lp_bld_type.h"
+#include "gallivm/lp_bld_flow.h"
+#include "gallivm/lp_bld_debug.h"
+#include "gallivm/lp_bld_tgsi.h"
+#include "gallivm/lp_bld_printf.h"
+
+#include "tgsi/tgsi_exec.h"
+
+#include "util/u_cpu_detect.h"
+#include "util/u_string.h"
+
+#include <llvm-c/Transforms/Scalar.h>
+
+#define DEBUG_STORE 0
+
+
+/* generates the draw jit function */
+static void
+draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *var);
+static void
+draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *var);
+
+static void
+init_globals(struct draw_llvm *llvm)
+{
+   LLVMTypeRef texture_type;
+
+   /* struct draw_jit_texture */
+   {
+      LLVMTypeRef elem_types[4];
+
+      elem_types[DRAW_JIT_TEXTURE_WIDTH]  = LLVMInt32Type();
+      elem_types[DRAW_JIT_TEXTURE_HEIGHT] = LLVMInt32Type();
+      elem_types[DRAW_JIT_TEXTURE_STRIDE] = LLVMInt32Type();
+      elem_types[DRAW_JIT_TEXTURE_DATA]   = LLVMPointerType(LLVMInt8Type(), 0);
+
+      texture_type = LLVMStructType(elem_types, Elements(elem_types), 0);
+
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, width,
+                             llvm->target, texture_type,
+                             DRAW_JIT_TEXTURE_WIDTH);
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, height,
+                             llvm->target, texture_type,
+                             DRAW_JIT_TEXTURE_HEIGHT);
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, stride,
+                             llvm->target, texture_type,
+                             DRAW_JIT_TEXTURE_STRIDE);
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, data,
+                             llvm->target, texture_type,
+                             DRAW_JIT_TEXTURE_DATA);
+      LP_CHECK_STRUCT_SIZE(struct draw_jit_texture,
+                           llvm->target, texture_type);
+
+      LLVMAddTypeName(llvm->module, "texture", texture_type);
+   }
+
+
+   /* struct draw_jit_context */
+   {
+      LLVMTypeRef elem_types[3];
+      LLVMTypeRef context_type;
+
+      elem_types[0] = LLVMPointerType(LLVMFloatType(), 0); /* vs_constants */
+      elem_types[1] = LLVMPointerType(LLVMFloatType(), 0); /* vs_constants */
+      elem_types[2] = LLVMArrayType(texture_type, PIPE_MAX_SAMPLERS); /* textures */
+
+      context_type = LLVMStructType(elem_types, Elements(elem_types), 0);
+
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_context, vs_constants,
+                             llvm->target, context_type, 0);
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_context, gs_constants,
+                             llvm->target, context_type, 1);
+      LP_CHECK_MEMBER_OFFSET(struct draw_jit_context, textures,
+                             llvm->target, context_type,
+                             DRAW_JIT_CONTEXT_TEXTURES_INDEX);
+      LP_CHECK_STRUCT_SIZE(struct draw_jit_context,
+                           llvm->target, context_type);
+
+      LLVMAddTypeName(llvm->module, "draw_jit_context", context_type);
+
+      llvm->context_ptr_type = LLVMPointerType(context_type, 0);
+   }
+   {
+      LLVMTypeRef buffer_ptr = LLVMPointerType(LLVMIntType(8), 0);
+      llvm->buffer_ptr_type = LLVMPointerType(buffer_ptr, 0);
+   }
+   /* struct pipe_vertex_buffer */
+   {
+      LLVMTypeRef elem_types[4];
+      LLVMTypeRef vb_type;
+
+      elem_types[0] = LLVMInt32Type();
+      elem_types[1] = LLVMInt32Type();
+      elem_types[2] = LLVMInt32Type();
+      elem_types[3] = LLVMPointerType(LLVMOpaqueType(), 0); /* vs_constants */
+
+      vb_type = LLVMStructType(elem_types, Elements(elem_types), 0);
+
+      LP_CHECK_MEMBER_OFFSET(struct pipe_vertex_buffer, stride,
+                             llvm->target, vb_type, 0);
+      LP_CHECK_MEMBER_OFFSET(struct pipe_vertex_buffer, buffer_offset,
+                             llvm->target, vb_type, 2);
+      LP_CHECK_STRUCT_SIZE(struct pipe_vertex_buffer,
+                           llvm->target, vb_type);
+
+      LLVMAddTypeName(llvm->module, "pipe_vertex_buffer", vb_type);
+
+      llvm->vb_ptr_type = LLVMPointerType(vb_type, 0);
+   }
+}
+
+static LLVMTypeRef
+create_vertex_header(struct draw_llvm *llvm, int data_elems)
+{
+   /* struct vertex_header */
+   LLVMTypeRef elem_types[3];
+   LLVMTypeRef vertex_header;
+   char struct_name[24];
+
+   util_snprintf(struct_name, 23, "vertex_header%d", data_elems);
+
+   elem_types[0]  = LLVMIntType(32);
+   elem_types[1]  = LLVMArrayType(LLVMFloatType(), 4);
+   elem_types[2]  = LLVMArrayType(elem_types[1], data_elems);
+
+   vertex_header = LLVMStructType(elem_types, Elements(elem_types), 0);
+
+   /* these are bit-fields and we can't take address of them
+      LP_CHECK_MEMBER_OFFSET(struct vertex_header, clipmask,
+      llvm->target, vertex_header,
+      DRAW_JIT_VERTEX_CLIPMASK);
+      LP_CHECK_MEMBER_OFFSET(struct vertex_header, edgeflag,
+      llvm->target, vertex_header,
+      DRAW_JIT_VERTEX_EDGEFLAG);
+      LP_CHECK_MEMBER_OFFSET(struct vertex_header, pad,
+      llvm->target, vertex_header,
+      DRAW_JIT_VERTEX_PAD);
+      LP_CHECK_MEMBER_OFFSET(struct vertex_header, vertex_id,
+      llvm->target, vertex_header,
+      DRAW_JIT_VERTEX_VERTEX_ID);
+   */
+   LP_CHECK_MEMBER_OFFSET(struct vertex_header, clip,
+                          llvm->target, vertex_header,
+                          DRAW_JIT_VERTEX_CLIP);
+   LP_CHECK_MEMBER_OFFSET(struct vertex_header, data,
+                          llvm->target, vertex_header,
+                          DRAW_JIT_VERTEX_DATA);
+
+   LLVMAddTypeName(llvm->module, struct_name, vertex_header);
+
+   return LLVMPointerType(vertex_header, 0);
+}
+
+struct draw_llvm *
+draw_llvm_create(struct draw_context *draw)
+{
+   struct draw_llvm *llvm = CALLOC_STRUCT( draw_llvm );
+
+   util_cpu_detect();
+
+   llvm->draw = draw;
+   llvm->engine = draw->engine;
+
+   debug_assert(llvm->engine);
+
+   llvm->module = LLVMModuleCreateWithName("draw_llvm");
+   llvm->provider = LLVMCreateModuleProviderForExistingModule(llvm->module);
+
+   LLVMAddModuleProvider(llvm->engine, llvm->provider);
+
+   llvm->target = LLVMGetExecutionEngineTargetData(llvm->engine);
+
+   llvm->pass = LLVMCreateFunctionPassManager(llvm->provider);
+   LLVMAddTargetData(llvm->target, llvm->pass);
+   /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
+    * but there are more on SVN. */
+   /* TODO: Add more passes */
+   LLVMAddConstantPropagationPass(llvm->pass);
+   if(util_cpu_caps.has_sse4_1) {
+      /* FIXME: There is a bug in this pass, whereby the combination of fptosi
+       * and sitofp (necessary for trunc/floor/ceil/round implementation)
+       * somehow becomes invalid code.
+       */
+      LLVMAddInstructionCombiningPass(llvm->pass);
+   }
+   LLVMAddPromoteMemoryToRegisterPass(llvm->pass);
+   LLVMAddGVNPass(llvm->pass);
+   LLVMAddCFGSimplificationPass(llvm->pass);
+
+   init_globals(llvm);
+
+
+#if 0
+   LLVMDumpModule(llvm->module);
+#endif
+
+   return llvm;
+}
+
+void
+draw_llvm_destroy(struct draw_llvm *llvm)
+{
+   FREE(llvm);
+}
+
+struct draw_llvm_variant *
+draw_llvm_prepare(struct draw_llvm *llvm, int num_inputs)
+{
+   struct draw_llvm_variant *variant = MALLOC(sizeof(struct draw_llvm_variant));
+
+   draw_llvm_make_variant_key(llvm, &variant->key);
+
+   llvm->vertex_header_ptr_type = create_vertex_header(llvm, num_inputs);
+
+   draw_llvm_generate(llvm, variant);
+   draw_llvm_generate_elts(llvm, variant);
+
+   return variant;
+}
+
+static void
+generate_vs(struct draw_llvm *llvm,
+            LLVMBuilderRef builder,
+            LLVMValueRef (*outputs)[NUM_CHANNELS],
+            const LLVMValueRef (*inputs)[NUM_CHANNELS],
+            LLVMValueRef context_ptr)
+{
+   const struct tgsi_token *tokens = llvm->draw->vs.vertex_shader->state.tokens;
+   struct lp_type vs_type;
+   LLVMValueRef consts_ptr = draw_jit_context_vs_constants(builder, context_ptr);
+
+   memset(&vs_type, 0, sizeof vs_type);
+   vs_type.floating = TRUE; /* floating point values */
+   vs_type.sign = TRUE;     /* values are signed */
+   vs_type.norm = FALSE;    /* values are not limited to [0,1] or [-1,1] */
+   vs_type.width = 32;      /* 32-bit float */
+   vs_type.length = 4;      /* 4 elements per vector */
+#if 0
+   num_vs = 4;              /* number of vertices per block */
+#endif
+
+   /*tgsi_dump(tokens, 0);*/
+   lp_build_tgsi_soa(builder,
+                     tokens,
+                     vs_type,
+                     NULL /*struct lp_build_mask_context *mask*/,
+                     consts_ptr,
+                     NULL /*pos*/,
+                     inputs,
+                     outputs,
+                     NULL/*sampler*/,
+                     &llvm->draw->vs.vertex_shader->info);
+}
+
+#if DEBUG_STORE
+static void print_vectorf(LLVMBuilderRef builder,
+                         LLVMValueRef vec)
+{
+   LLVMValueRef val[4];
+   val[0] = LLVMBuildExtractElement(builder, vec,
+                                    LLVMConstInt(LLVMInt32Type(), 0, 0), "");
+   val[1] = LLVMBuildExtractElement(builder, vec,
+                                    LLVMConstInt(LLVMInt32Type(), 1, 0), "");
+   val[2] = LLVMBuildExtractElement(builder, vec,
+                                    LLVMConstInt(LLVMInt32Type(), 2, 0), "");
+   val[3] = LLVMBuildExtractElement(builder, vec,
+                                    LLVMConstInt(LLVMInt32Type(), 3, 0), "");
+   lp_build_printf(builder, "vector = [%f, %f, %f, %f]\n",
+                   val[0], val[1], val[2], val[3]);
+}
+#endif
+
+static void
+generate_fetch(LLVMBuilderRef builder,
+               LLVMValueRef vbuffers_ptr,
+               LLVMValueRef *res,
+               struct pipe_vertex_element *velem,
+               LLVMValueRef vbuf,
+               LLVMValueRef index)
+{
+   LLVMValueRef indices = LLVMConstInt(LLVMInt64Type(), velem->vertex_buffer_index, 0);
+   LLVMValueRef vbuffer_ptr = LLVMBuildGEP(builder, vbuffers_ptr,
+                                           &indices, 1, "");
+   LLVMValueRef vb_stride = draw_jit_vbuffer_stride(builder, vbuf);
+   LLVMValueRef vb_max_index = draw_jit_vbuffer_max_index(builder, vbuf);
+   LLVMValueRef vb_buffer_offset = draw_jit_vbuffer_offset(builder, vbuf);
+   LLVMValueRef cond;
+   LLVMValueRef stride;
+
+   cond = LLVMBuildICmp(builder, LLVMIntULE, index, vb_max_index, "");
+
+   index = LLVMBuildSelect(builder, cond, index, vb_max_index, "");
+
+   stride = LLVMBuildMul(builder, vb_stride, index, "");
+
+   vbuffer_ptr = LLVMBuildLoad(builder, vbuffer_ptr, "vbuffer");
+
+   stride = LLVMBuildAdd(builder, stride,
+                         vb_buffer_offset,
+                         "");
+   stride = LLVMBuildAdd(builder, stride,
+                         LLVMConstInt(LLVMInt32Type(), velem->src_offset, 0),
+                         "");
+
+   /*lp_build_printf(builder, "vbuf index = %d, stride is %d\n", indices, stride);*/
+   vbuffer_ptr = LLVMBuildGEP(builder, vbuffer_ptr, &stride, 1, "");
+
+   *res = draw_llvm_translate_from(builder, vbuffer_ptr, velem->src_format);
+}
+
+static LLVMValueRef
+aos_to_soa(LLVMBuilderRef builder,
+           LLVMValueRef val0,
+           LLVMValueRef val1,
+           LLVMValueRef val2,
+           LLVMValueRef val3,
+           LLVMValueRef channel)
+{
+   LLVMValueRef ex, res;
+
+   ex = LLVMBuildExtractElement(builder, val0,
+                                channel, "");
+   res = LLVMBuildInsertElement(builder,
+                                LLVMConstNull(LLVMTypeOf(val0)),
+                                ex,
+                                LLVMConstInt(LLVMInt32Type(), 0, 0),
+                                "");
+
+   ex = LLVMBuildExtractElement(builder, val1,
+                                channel, "");
+   res = LLVMBuildInsertElement(builder,
+                                res, ex,
+                                LLVMConstInt(LLVMInt32Type(), 1, 0),
+                                "");
+
+   ex = LLVMBuildExtractElement(builder, val2,
+                                channel, "");
+   res = LLVMBuildInsertElement(builder,
+                                res, ex,
+                                LLVMConstInt(LLVMInt32Type(), 2, 0),
+                                "");
+
+   ex = LLVMBuildExtractElement(builder, val3,
+                                channel, "");
+   res = LLVMBuildInsertElement(builder,
+                                res, ex,
+                                LLVMConstInt(LLVMInt32Type(), 3, 0),
+                                "");
+
+   return res;
+}
+
+static void
+soa_to_aos(LLVMBuilderRef builder,
+           LLVMValueRef soa[NUM_CHANNELS],
+           LLVMValueRef aos[NUM_CHANNELS])
+{
+   LLVMValueRef comp;
+   int i = 0;
+
+   debug_assert(NUM_CHANNELS == 4);
+
+   aos[0] = LLVMConstNull(LLVMTypeOf(soa[0]));
+   aos[1] = aos[2] = aos[3] = aos[0];
+
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      LLVMValueRef channel = LLVMConstInt(LLVMInt32Type(), i, 0);
+
+      comp = LLVMBuildExtractElement(builder, soa[i],
+                                     LLVMConstInt(LLVMInt32Type(), 0, 0), "");
+      aos[0] = LLVMBuildInsertElement(builder, aos[0], comp, channel, "");
+
+      comp = LLVMBuildExtractElement(builder, soa[i],
+                                     LLVMConstInt(LLVMInt32Type(), 1, 0), "");
+      aos[1] = LLVMBuildInsertElement(builder, aos[1], comp, channel, "");
+
+      comp = LLVMBuildExtractElement(builder, soa[i],
+                                     LLVMConstInt(LLVMInt32Type(), 2, 0), "");
+      aos[2] = LLVMBuildInsertElement(builder, aos[2], comp, channel, "");
+
+      comp = LLVMBuildExtractElement(builder, soa[i],
+                                     LLVMConstInt(LLVMInt32Type(), 3, 0), "");
+      aos[3] = LLVMBuildInsertElement(builder, aos[3], comp, channel, "");
+
+   }
+}
+
+static void
+convert_to_soa(LLVMBuilderRef builder,
+               LLVMValueRef (*aos)[NUM_CHANNELS],
+               LLVMValueRef (*soa)[NUM_CHANNELS],
+               int num_attribs)
+{
+   int i;
+
+   debug_assert(NUM_CHANNELS == 4);
+
+   for (i = 0; i < num_attribs; ++i) {
+      LLVMValueRef val0 = aos[i][0];
+      LLVMValueRef val1 = aos[i][1];
+      LLVMValueRef val2 = aos[i][2];
+      LLVMValueRef val3 = aos[i][3];
+
+      soa[i][0] = aos_to_soa(builder, val0, val1, val2, val3,
+                             LLVMConstInt(LLVMInt32Type(), 0, 0));
+      soa[i][1] = aos_to_soa(builder, val0, val1, val2, val3,
+                             LLVMConstInt(LLVMInt32Type(), 1, 0));
+      soa[i][2] = aos_to_soa(builder, val0, val1, val2, val3,
+                             LLVMConstInt(LLVMInt32Type(), 2, 0));
+      soa[i][3] = aos_to_soa(builder, val0, val1, val2, val3,
+                             LLVMConstInt(LLVMInt32Type(), 3, 0));
+   }
+}
+
+static void
+store_aos(LLVMBuilderRef builder,
+          LLVMValueRef io_ptr,
+          LLVMValueRef index,
+          LLVMValueRef value)
+{
+   LLVMValueRef id_ptr = draw_jit_header_id(builder, io_ptr);
+   LLVMValueRef data_ptr = draw_jit_header_data(builder, io_ptr);
+   LLVMValueRef indices[3];
+
+   indices[0] = LLVMConstInt(LLVMInt32Type(), 0, 0);
+   indices[1] = index;
+   indices[2] = LLVMConstInt(LLVMInt32Type(), 0, 0);
+
+   /* undefined vertex */
+   LLVMBuildStore(builder, LLVMConstInt(LLVMInt32Type(),
+                                        0xffff, 0), id_ptr);
+
+#if DEBUG_STORE
+   lp_build_printf(builder, "    ---- %p storing attribute %d (io = %p)\n", data_ptr, index, io_ptr);
+#endif
+#if 0
+   /*lp_build_printf(builder, " ---- %p storing at %d (%p)  ", io_ptr, index, data_ptr);
+     print_vectorf(builder, value);*/
+   data_ptr = LLVMBuildBitCast(builder, data_ptr,
+                               LLVMPointerType(LLVMArrayType(LLVMVectorType(LLVMFloatType(), 4), 0), 0),
+                               "datavec");
+   data_ptr = LLVMBuildGEP(builder, data_ptr, indices, 2, "");
+
+   LLVMBuildStore(builder, value, data_ptr);
+#else
+   {
+      LLVMValueRef x, y, z, w;
+      LLVMValueRef idx0, idx1, idx2, idx3;
+      LLVMValueRef gep0, gep1, gep2, gep3;
+      data_ptr = LLVMBuildGEP(builder, data_ptr, indices, 3, "");
+
+      idx0 = LLVMConstInt(LLVMInt32Type(), 0, 0);
+      idx1 = LLVMConstInt(LLVMInt32Type(), 1, 0);
+      idx2 = LLVMConstInt(LLVMInt32Type(), 2, 0);
+      idx3 = LLVMConstInt(LLVMInt32Type(), 3, 0);
+
+      x = LLVMBuildExtractElement(builder, value,
+                                  idx0, "");
+      y = LLVMBuildExtractElement(builder, value,
+                                  idx1, "");
+      z = LLVMBuildExtractElement(builder, value,
+                                  idx2, "");
+      w = LLVMBuildExtractElement(builder, value,
+                                  idx3, "");
+
+      gep0 = LLVMBuildGEP(builder, data_ptr, &idx0, 1, "");
+      gep1 = LLVMBuildGEP(builder, data_ptr, &idx1, 1, "");
+      gep2 = LLVMBuildGEP(builder, data_ptr, &idx2, 1, "");
+      gep3 = LLVMBuildGEP(builder, data_ptr, &idx3, 1, "");
+
+      /*lp_build_printf(builder, "##### x = %f (%p), y = %f (%p), z = %f (%p), w = %f (%p)\n",
+        x, gep0, y, gep1, z, gep2, w, gep3);*/
+      LLVMBuildStore(builder, x, gep0);
+      LLVMBuildStore(builder, y, gep1);
+      LLVMBuildStore(builder, z, gep2);
+      LLVMBuildStore(builder, w, gep3);
+   }
+#endif
+}
+
+static void
+store_aos_array(LLVMBuilderRef builder,
+                LLVMValueRef io_ptr,
+                LLVMValueRef aos[NUM_CHANNELS],
+                int attrib,
+                int num_outputs)
+{
+   LLVMValueRef attr_index = LLVMConstInt(LLVMInt32Type(), attrib, 0);
+   LLVMValueRef ind0 = LLVMConstInt(LLVMInt32Type(), 0, 0);
+   LLVMValueRef ind1 = LLVMConstInt(LLVMInt32Type(), 1, 0);
+   LLVMValueRef ind2 = LLVMConstInt(LLVMInt32Type(), 2, 0);
+   LLVMValueRef ind3 = LLVMConstInt(LLVMInt32Type(), 3, 0);
+   LLVMValueRef io0_ptr, io1_ptr, io2_ptr, io3_ptr;
+
+   debug_assert(NUM_CHANNELS == 4);
+
+   io0_ptr = LLVMBuildGEP(builder, io_ptr,
+                          &ind0, 1, "");
+   io1_ptr = LLVMBuildGEP(builder, io_ptr,
+                          &ind1, 1, "");
+   io2_ptr = LLVMBuildGEP(builder, io_ptr,
+                          &ind2, 1, "");
+   io3_ptr = LLVMBuildGEP(builder, io_ptr,
+                          &ind3, 1, "");
+
+#if DEBUG_STORE
+   lp_build_printf(builder, "   io = %p, indexes[%d, %d, %d, %d]\n",
+                   io_ptr, ind0, ind1, ind2, ind3);
+#endif
+
+   store_aos(builder, io0_ptr, attr_index, aos[0]);
+   store_aos(builder, io1_ptr, attr_index, aos[1]);
+   store_aos(builder, io2_ptr, attr_index, aos[2]);
+   store_aos(builder, io3_ptr, attr_index, aos[3]);
+}
+
+static void
+convert_to_aos(LLVMBuilderRef builder,
+               LLVMValueRef io,
+               LLVMValueRef (*outputs)[NUM_CHANNELS],
+               int num_outputs,
+               int max_vertices)
+{
+   unsigned chan, attrib;
+
+#if DEBUG_STORE
+   lp_build_printf(builder, "   # storing begin\n");
+#endif
+   for (attrib = 0; attrib < num_outputs; ++attrib) {
+      LLVMValueRef soa[4];
+      LLVMValueRef aos[4];
+      for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+         if(outputs[attrib][chan]) {
+            LLVMValueRef out = LLVMBuildLoad(builder, outputs[attrib][chan], "");
+            lp_build_name(out, "output%u.%c", attrib, "xyzw"[chan]);
+            /*lp_build_printf(builder, "output %d : %d ",
+                            LLVMConstInt(LLVMInt32Type(), attrib, 0),
+                            LLVMConstInt(LLVMInt32Type(), chan, 0));
+              print_vectorf(builder, out);*/
+            soa[chan] = out;
+         } else
+            soa[chan] = 0;
+      }
+      soa_to_aos(builder, soa, aos);
+      store_aos_array(builder,
+                      io,
+                      aos,
+                      attrib,
+                      num_outputs);
+   }
+#if DEBUG_STORE
+   lp_build_printf(builder, "   # storing end\n");
+#endif
+}
+
+static void
+draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
+{
+   LLVMTypeRef arg_types[7];
+   LLVMTypeRef func_type;
+   LLVMValueRef context_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+   LLVMValueRef start, end, count, stride, step, io_itr;
+   LLVMValueRef io_ptr, vbuffers_ptr, vb_ptr;
+   struct draw_context *draw = llvm->draw;
+   unsigned i, j;
+   struct lp_build_context bld;
+   struct lp_build_loop_state lp_loop;
+   struct lp_type vs_type = lp_type_float_vec(32);
+   const int max_vertices = 4;
+   LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
+
+   arg_types[0] = llvm->context_ptr_type;           /* context */
+   arg_types[1] = llvm->vertex_header_ptr_type;     /* vertex_header */
+   arg_types[2] = llvm->buffer_ptr_type;            /* vbuffers */
+   arg_types[3] = LLVMInt32Type();                  /* start */
+   arg_types[4] = LLVMInt32Type();                  /* count */
+   arg_types[5] = LLVMInt32Type();                  /* stride */
+   arg_types[6] = llvm->vb_ptr_type;                /* pipe_vertex_buffer's */
+
+   func_type = LLVMFunctionType(LLVMVoidType(), arg_types, Elements(arg_types), 0);
+
+   variant->function = LLVMAddFunction(llvm->module, "draw_llvm_shader", func_type);
+   LLVMSetFunctionCallConv(variant->function, LLVMCCallConv);
+   for(i = 0; i < Elements(arg_types); ++i)
+      if(LLVMGetTypeKind(arg_types[i]) == LLVMPointerTypeKind)
+         LLVMAddAttribute(LLVMGetParam(variant->function, i), LLVMNoAliasAttribute);
+
+   context_ptr  = LLVMGetParam(variant->function, 0);
+   io_ptr       = LLVMGetParam(variant->function, 1);
+   vbuffers_ptr = LLVMGetParam(variant->function, 2);
+   start        = LLVMGetParam(variant->function, 3);
+   count        = LLVMGetParam(variant->function, 4);
+   stride       = LLVMGetParam(variant->function, 5);
+   vb_ptr       = LLVMGetParam(variant->function, 6);
+
+   lp_build_name(context_ptr, "context");
+   lp_build_name(io_ptr, "io");
+   lp_build_name(vbuffers_ptr, "vbuffers");
+   lp_build_name(start, "start");
+   lp_build_name(count, "count");
+   lp_build_name(stride, "stride");
+   lp_build_name(vb_ptr, "vb");
+
+   /*
+    * Function body
+    */
+
+   block = LLVMAppendBasicBlock(variant->function, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   lp_build_context_init(&bld, builder, vs_type);
+
+   end = lp_build_add(&bld, start, count);
+
+   step = LLVMConstInt(LLVMInt32Type(), max_vertices, 0);
+
+#if DEBUG_STORE
+   lp_build_printf(builder, "start = %d, end = %d, step = %d\n",
+                   start, end, step);
+#endif
+   lp_build_loop_begin(builder, start, &lp_loop);
+   {
+      LLVMValueRef inputs[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+      LLVMValueRef aos_attribs[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS] = { { 0 } };
+      LLVMValueRef io;
+      const LLVMValueRef (*ptr_aos)[NUM_CHANNELS];
+
+      io_itr = LLVMBuildSub(builder, lp_loop.counter, start, "");
+      io = LLVMBuildGEP(builder, io_ptr, &io_itr, 1, "");
+#if DEBUG_STORE
+      lp_build_printf(builder, " --- io %d = %p, loop counter %d\n",
+                      io_itr, io, lp_loop.counter);
+#endif
+      for (i = 0; i < NUM_CHANNELS; ++i) {
+         LLVMValueRef true_index = LLVMBuildAdd(
+            builder,
+            lp_loop.counter,
+            LLVMConstInt(LLVMInt32Type(), i, 0), "");
+         for (j = 0; j < draw->pt.nr_vertex_elements; ++j) {
+            struct pipe_vertex_element *velem = &draw->pt.vertex_element[j];
+            LLVMValueRef vb_index = LLVMConstInt(LLVMInt32Type(),
+                                                 velem->vertex_buffer_index,
+                                                 0);
+            LLVMValueRef vb = LLVMBuildGEP(builder, vb_ptr,
+                                           &vb_index, 1, "");
+            generate_fetch(builder, vbuffers_ptr,
+                           &aos_attribs[j][i], velem, vb, true_index);
+         }
+      }
+      convert_to_soa(builder, aos_attribs, inputs,
+                     draw->pt.nr_vertex_elements);
+
+      ptr_aos = (const LLVMValueRef (*)[NUM_CHANNELS]) inputs;
+      generate_vs(llvm,
+                  builder,
+                  outputs,
+                  ptr_aos,
+                  context_ptr);
+
+      convert_to_aos(builder, io, outputs,
+                     draw->vs.vertex_shader->info.num_outputs,
+                     max_vertices);
+   }
+   lp_build_loop_end_cond(builder, end, step, LLVMIntUGE, &lp_loop);
+
+   LLVMBuildRetVoid(builder);
+
+   LLVMDisposeBuilder(builder);
+
+   /*
+    * Translate the LLVM IR into machine code.
+    */
+#ifdef DEBUG
+   if(LLVMVerifyFunction(variant->function, LLVMPrintMessageAction)) {
+      LLVMDumpValue(variant->function);
+      assert(0);
+   }
+#endif
+
+   LLVMRunFunctionPassManager(llvm->pass, variant->function);
+
+   if (0) {
+      LLVMDumpValue(variant->function);
+      debug_printf("\n");
+   }
+   variant->jit_func = (draw_jit_vert_func)LLVMGetPointerToGlobal(llvm->draw->engine, variant->function);
+
+   if (0)
+      lp_disassemble(variant->jit_func);
+}
+
+
+static void
+draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
+{
+   LLVMTypeRef arg_types[7];
+   LLVMTypeRef func_type;
+   LLVMValueRef context_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+   LLVMValueRef fetch_elts, fetch_count, stride, step, io_itr;
+   LLVMValueRef io_ptr, vbuffers_ptr, vb_ptr;
+   struct draw_context *draw = llvm->draw;
+   unsigned i, j;
+   struct lp_build_context bld;
+   struct lp_build_loop_state lp_loop;
+   struct lp_type vs_type = lp_type_float_vec(32);
+   const int max_vertices = 4;
+   LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
+
+   arg_types[0] = llvm->context_ptr_type;               /* context */
+   arg_types[1] = llvm->vertex_header_ptr_type;         /* vertex_header */
+   arg_types[2] = llvm->buffer_ptr_type;                /* vbuffers */
+   arg_types[3] = LLVMPointerType(LLVMInt32Type(), 0);  /* fetch_elts * */
+   arg_types[4] = LLVMInt32Type();                      /* fetch_count */
+   arg_types[5] = LLVMInt32Type();                      /* stride */
+   arg_types[6] = llvm->vb_ptr_type;                    /* pipe_vertex_buffer's */
+
+   func_type = LLVMFunctionType(LLVMVoidType(), arg_types, Elements(arg_types), 0);
+
+   variant->function_elts = LLVMAddFunction(llvm->module, "draw_llvm_shader_elts", func_type);
+   LLVMSetFunctionCallConv(variant->function_elts, LLVMCCallConv);
+   for(i = 0; i < Elements(arg_types); ++i)
+      if(LLVMGetTypeKind(arg_types[i]) == LLVMPointerTypeKind)
+         LLVMAddAttribute(LLVMGetParam(variant->function_elts, i), LLVMNoAliasAttribute);
+
+   context_ptr  = LLVMGetParam(variant->function_elts, 0);
+   io_ptr       = LLVMGetParam(variant->function_elts, 1);
+   vbuffers_ptr = LLVMGetParam(variant->function_elts, 2);
+   fetch_elts   = LLVMGetParam(variant->function_elts, 3);
+   fetch_count  = LLVMGetParam(variant->function_elts, 4);
+   stride       = LLVMGetParam(variant->function_elts, 5);
+   vb_ptr       = LLVMGetParam(variant->function_elts, 6);
+
+   lp_build_name(context_ptr, "context");
+   lp_build_name(io_ptr, "io");
+   lp_build_name(vbuffers_ptr, "vbuffers");
+   lp_build_name(fetch_elts, "fetch_elts");
+   lp_build_name(fetch_count, "fetch_count");
+   lp_build_name(stride, "stride");
+   lp_build_name(vb_ptr, "vb");
+
+   /*
+    * Function body
+    */
+
+   block = LLVMAppendBasicBlock(variant->function_elts, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   lp_build_context_init(&bld, builder, vs_type);
+
+   step = LLVMConstInt(LLVMInt32Type(), max_vertices, 0);
+
+   lp_build_loop_begin(builder, LLVMConstInt(LLVMInt32Type(), 0, 0), &lp_loop);
+   {
+      LLVMValueRef inputs[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+      LLVMValueRef aos_attribs[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS] = { { 0 } };
+      LLVMValueRef io;
+      const LLVMValueRef (*ptr_aos)[NUM_CHANNELS];
+
+      io_itr = lp_loop.counter;
+      io = LLVMBuildGEP(builder, io_ptr, &io_itr, 1, "");
+#if DEBUG_STORE
+      lp_build_printf(builder, " --- io %d = %p, loop counter %d\n",
+                      io_itr, io, lp_loop.counter);
+#endif
+      for (i = 0; i < NUM_CHANNELS; ++i) {
+         LLVMValueRef true_index = LLVMBuildAdd(
+            builder,
+            lp_loop.counter,
+            LLVMConstInt(LLVMInt32Type(), i, 0), "");
+         LLVMValueRef fetch_ptr = LLVMBuildGEP(builder, fetch_elts,
+                                               &true_index, 1, "");
+         true_index = LLVMBuildLoad(builder, fetch_ptr, "fetch_elt");
+         for (j = 0; j < draw->pt.nr_vertex_elements; ++j) {
+            struct pipe_vertex_element *velem = &draw->pt.vertex_element[j];
+            LLVMValueRef vb_index = LLVMConstInt(LLVMInt32Type(),
+                                                 velem->vertex_buffer_index,
+                                                 0);
+            LLVMValueRef vb = LLVMBuildGEP(builder, vb_ptr,
+                                           &vb_index, 1, "");
+            generate_fetch(builder, vbuffers_ptr,
+                           &aos_attribs[j][i], velem, vb, true_index);
+         }
+      }
+      convert_to_soa(builder, aos_attribs, inputs,
+                     draw->pt.nr_vertex_elements);
+
+      ptr_aos = (const LLVMValueRef (*)[NUM_CHANNELS]) inputs;
+      generate_vs(llvm,
+                  builder,
+                  outputs,
+                  ptr_aos,
+                  context_ptr);
+
+      convert_to_aos(builder, io, outputs,
+                     draw->vs.vertex_shader->info.num_outputs,
+                     max_vertices);
+   }
+   lp_build_loop_end_cond(builder, fetch_count, step, LLVMIntUGE, &lp_loop);
+
+   LLVMBuildRetVoid(builder);
+
+   LLVMDisposeBuilder(builder);
+
+   /*
+    * Translate the LLVM IR into machine code.
+    */
+#ifdef DEBUG
+   if(LLVMVerifyFunction(variant->function_elts, LLVMPrintMessageAction)) {
+      LLVMDumpValue(variant->function_elts);
+      assert(0);
+   }
+#endif
+
+   LLVMRunFunctionPassManager(llvm->pass, variant->function_elts);
+
+   if (0) {
+      LLVMDumpValue(variant->function_elts);
+      debug_printf("\n");
+   }
+   variant->jit_func_elts = (draw_jit_vert_func_elts)LLVMGetPointerToGlobal(
+      llvm->draw->engine, variant->function_elts);
+
+   if (0)
+      lp_disassemble(variant->jit_func_elts);
+}
+
+void
+draw_llvm_make_variant_key(struct draw_llvm *llvm,
+                           struct draw_llvm_variant_key *key)
+{
+   memset(key, 0, sizeof(struct draw_llvm_variant_key));
+
+   key->nr_vertex_elements = llvm->draw->pt.nr_vertex_elements;
+
+   memcpy(key->vertex_element,
+          llvm->draw->pt.vertex_element,
+          sizeof(struct pipe_vertex_element) * key->nr_vertex_elements);
+
+   memcpy(&key->vs,
+          &llvm->draw->vs.vertex_shader->state,
+          sizeof(struct pipe_shader_state));
+}
diff --git a/src/gallium/auxiliary/draw/draw_llvm.h b/src/gallium/auxiliary/draw/draw_llvm.h
new file mode 100644
index 00000000000..58fee7f9d60
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_llvm.h
@@ -0,0 +1,159 @@
+#ifndef HAVE_LLVM_H
+#define HAVE_LLVM_H
+
+#include "draw/draw_private.h"
+
+#include "pipe/p_context.h"
+
+#include <llvm-c/Core.h>
+#include <llvm-c/Analysis.h>
+#include <llvm-c/Target.h>
+#include <llvm-c/ExecutionEngine.h>
+
+struct draw_jit_texture
+{
+   uint32_t width;
+   uint32_t height;
+   uint32_t stride;
+   const void *data;
+};
+
+enum {
+   DRAW_JIT_TEXTURE_WIDTH = 0,
+   DRAW_JIT_TEXTURE_HEIGHT,
+   DRAW_JIT_TEXTURE_STRIDE,
+   DRAW_JIT_TEXTURE_DATA
+};
+
+enum {
+   DRAW_JIT_VERTEX_VERTEX_ID = 0,
+   DRAW_JIT_VERTEX_CLIP,
+   DRAW_JIT_VERTEX_DATA
+};
+
+/**
+ * This structure is passed directly to the generated vertex shader.
+ *
+ * It contains the derived state.
+ *
+ * Changes here must be reflected in the draw_jit_context_* macros.
+ * Changes to the ordering should be avoided.
+ *
+ * Only use types with a clear size and padding here, in particular prefer the
+ * stdint.h types to the basic integer types.
+ */
+struct draw_jit_context
+{
+   const float *vs_constants;
+   const float *gs_constants;
+
+
+   struct draw_jit_texture textures[PIPE_MAX_SAMPLERS];
+};
+
+
+#define draw_jit_context_vs_constants(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 0, "vs_constants")
+
+#define draw_jit_context_gs_constants(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 1, "gs_constants")
+
+#define DRAW_JIT_CONTEXT_TEXTURES_INDEX 2
+
+#define draw_jit_context_textures(_builder, _ptr) \
+   lp_build_struct_get_ptr(_builder, _ptr, DRAW_JIT_CONTEXT_TEXTURES_INDEX, "textures")
+
+
+
+#define draw_jit_header_id(_builder, _ptr)              \
+   lp_build_struct_get_ptr(_builder, _ptr, 0, "id")
+
+#define draw_jit_header_clip(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 1, "clip")
+
+#define draw_jit_header_data(_builder, _ptr)            \
+   lp_build_struct_get_ptr(_builder, _ptr, 2, "data")
+
+
+#define draw_jit_vbuffer_stride(_builder, _ptr)         \
+   lp_build_struct_get(_builder, _ptr, 0, "stride")
+
+#define draw_jit_vbuffer_max_index(_builder, _ptr)      \
+   lp_build_struct_get(_builder, _ptr, 1, "max_index")
+
+#define draw_jit_vbuffer_offset(_builder, _ptr)         \
+   lp_build_struct_get(_builder, _ptr, 2, "buffer_offset")
+
+
+typedef void
+(*draw_jit_vert_func)(struct draw_jit_context *context,
+                      struct vertex_header *io,
+                      const char *vbuffers[PIPE_MAX_ATTRIBS],
+                      unsigned start,
+                      unsigned count,
+                      unsigned stride,
+                      struct pipe_vertex_buffer *vertex_buffers);
+
+
+typedef void
+(*draw_jit_vert_func_elts)(struct draw_jit_context *context,
+                           struct vertex_header *io,
+                           const char *vbuffers[PIPE_MAX_ATTRIBS],
+                           const unsigned *fetch_elts,
+                           unsigned fetch_count,
+                           unsigned stride,
+                           struct pipe_vertex_buffer *vertex_buffers);
+
+struct draw_llvm {
+   struct draw_context *draw;
+
+   struct draw_jit_context jit_context;
+
+   LLVMModuleRef module;
+   LLVMExecutionEngineRef engine;
+   LLVMModuleProviderRef provider;
+   LLVMTargetDataRef target;
+   LLVMPassManagerRef pass;
+
+   LLVMTypeRef context_ptr_type;
+   LLVMTypeRef vertex_header_ptr_type;
+   LLVMTypeRef buffer_ptr_type;
+   LLVMTypeRef vb_ptr_type;
+};
+
+struct draw_llvm_variant_key
+{
+   struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
+   unsigned                   nr_vertex_elements;
+   struct pipe_shader_state   vs;
+};
+
+struct draw_llvm_variant
+{
+   struct draw_llvm_variant_key key;
+   LLVMValueRef function;
+   LLVMValueRef function_elts;
+   draw_jit_vert_func jit_func;
+   draw_jit_vert_func_elts jit_func_elts;
+
+   struct draw_llvm_variant *next;
+};
+
+struct draw_llvm *
+draw_llvm_create(struct draw_context *draw);
+
+void
+draw_llvm_destroy(struct draw_llvm *llvm);
+
+struct draw_llvm_variant *
+draw_llvm_prepare(struct draw_llvm *llvm, int num_inputs);
+
+void
+draw_llvm_make_variant_key(struct draw_llvm *llvm,
+                           struct draw_llvm_variant_key *key);
+
+LLVMValueRef
+draw_llvm_translate_from(LLVMBuilderRef builder,
+                         LLVMValueRef vbuffer,
+                         enum pipe_format from_format);
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_llvm_translate.c b/src/gallium/auxiliary/draw/draw_llvm_translate.c
new file mode 100644
index 00000000000..d7da7ed357d
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_llvm_translate.c
@@ -0,0 +1,497 @@
+#include "draw_private.h"
+#include "draw_context.h"
+
+#include "draw_llvm.h"
+
+#include "gallivm/lp_bld_arit.h"
+#include "gallivm/lp_bld_struct.h"
+#include "gallivm/lp_bld_format.h"
+#include "gallivm/lp_bld_debug.h"
+
+#include "util/u_memory.h"
+#include "util/u_format.h"
+#include "pipe/p_state.h"
+
+
+#define DRAW_DBG 0
+
+static  LLVMValueRef
+from_64_float(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMDoubleType(), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   return LLVMBuildFPTrunc(builder, l, LLVMFloatType(), "");
+}
+
+static LLVMValueRef
+from_32_float(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMFloatType(), 0) , "");
+   return LLVMBuildLoad(builder, bc, "");
+}
+
+static INLINE LLVMValueRef
+from_8_uscaled(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, val, "");
+   return LLVMBuildUIToFP(builder, l, LLVMFloatType(), "");
+}
+
+static INLINE LLVMValueRef
+from_16_uscaled(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(16), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   return LLVMBuildUIToFP(builder, l, LLVMFloatType(), "");
+}
+
+static INLINE LLVMValueRef
+from_32_uscaled(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(32), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   return LLVMBuildUIToFP(builder, l, LLVMFloatType(), "");
+}
+
+static INLINE LLVMValueRef
+from_8_sscaled(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, val, "");
+   return LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+}
+
+static INLINE LLVMValueRef
+from_16_sscaled(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(16), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   return LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+}
+
+static INLINE LLVMValueRef
+from_32_sscaled(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(32), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   return LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+}
+
+
+static INLINE LLVMValueRef
+from_8_unorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, val, "");
+   LLVMValueRef uscaled = LLVMBuildUIToFP(builder, l, LLVMFloatType(), "");
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 255.), "");
+}
+
+static INLINE LLVMValueRef
+from_16_unorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(16), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   LLVMValueRef uscaled = LLVMBuildUIToFP(builder, l, LLVMFloatType(), "");
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 65535.), "");
+}
+
+static INLINE LLVMValueRef
+from_32_unorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(32), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   LLVMValueRef uscaled = LLVMBuildUIToFP(builder, l, LLVMFloatType(), "");
+
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 4294967295.), "");
+}
+
+static INLINE LLVMValueRef
+from_8_snorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, val, "");
+   LLVMValueRef uscaled = LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 127.0), "");
+}
+
+static INLINE LLVMValueRef
+from_16_snorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(16), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   LLVMValueRef uscaled = LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 32767.0f), "");
+}
+
+static INLINE LLVMValueRef
+from_32_snorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(32), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   LLVMValueRef uscaled = LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 2147483647.0), "");
+}
+
+static INLINE LLVMValueRef
+from_32_fixed(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef bc = LLVMBuildBitCast(builder, val,
+                                      LLVMPointerType(LLVMIntType(32), 0) , "");
+   LLVMValueRef l = LLVMBuildLoad(builder, bc, "");
+   LLVMValueRef uscaled = LLVMBuildSIToFP(builder, l, LLVMFloatType(), "");
+
+   return LLVMBuildFDiv(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 65536.0), "");
+}
+
+static LLVMValueRef
+to_64_float(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPExt(builder, l, LLVMDoubleType(), "");
+}
+
+static LLVMValueRef
+to_32_float(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   return LLVMBuildLoad(builder, fp, "");
+}
+
+static INLINE LLVMValueRef
+to_8_uscaled(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPToUI(builder, l, LLVMIntType(8), "");
+}
+
+static INLINE LLVMValueRef
+to_16_uscaled(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPToUI(builder, l, LLVMIntType(16), "");
+}
+
+static INLINE LLVMValueRef
+to_32_uscaled(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPToUI(builder, l, LLVMIntType(32), "");
+}
+
+static INLINE LLVMValueRef
+to_8_sscaled(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPToSI(builder, l, LLVMIntType(8), "");
+}
+
+static INLINE LLVMValueRef
+to_16_sscaled(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPToSI(builder, l, LLVMIntType(16), "");
+}
+
+static INLINE LLVMValueRef
+to_32_sscaled(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   return LLVMBuildFPToSI(builder, l, LLVMIntType(32), "");
+}
+
+static INLINE LLVMValueRef
+to_8_unorm(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   LLVMValueRef uscaled = LLVMBuildFPToUI(builder, l, LLVMIntType(8), "");
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 255.), "");
+}
+
+static INLINE LLVMValueRef
+to_16_unorm(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   LLVMValueRef uscaled = LLVMBuildFPToUI(builder, l, LLVMIntType(32), "");
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 65535.), "");
+}
+
+static INLINE LLVMValueRef
+to_32_unorm(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   LLVMValueRef uscaled = LLVMBuildFPToUI(builder, l, LLVMIntType(32), "");
+
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 4294967295.), "");
+}
+
+static INLINE LLVMValueRef
+to_8_snorm(LLVMBuilderRef builder, LLVMValueRef val)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, val, "");
+   LLVMValueRef uscaled = LLVMBuildFPToSI(builder, l, LLVMIntType(8), "");
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 127.0), "");
+}
+
+static INLINE LLVMValueRef
+to_16_snorm(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   LLVMValueRef uscaled = LLVMBuildFPToSI(builder, l, LLVMIntType(16), "");
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 32767.0f), "");
+}
+
+static INLINE LLVMValueRef
+to_32_snorm(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   LLVMValueRef uscaled = LLVMBuildFPToSI(builder, l, LLVMIntType(32), "");
+
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 2147483647.0), "");
+}
+
+static INLINE LLVMValueRef
+to_32_fixed(LLVMBuilderRef builder, LLVMValueRef fp)
+{
+   LLVMValueRef l = LLVMBuildLoad(builder, fp, "");
+   LLVMValueRef uscaled = LLVMBuildFPToSI(builder, l, LLVMIntType(32), "");
+
+   return LLVMBuildFMul(builder, uscaled,
+                        LLVMConstReal(LLVMFloatType(), 65536.0), "");
+}
+
+typedef LLVMValueRef (*from_func)(LLVMBuilderRef, LLVMValueRef);
+typedef  LLVMValueRef (*to_func)(LLVMBuilderRef, LLVMValueRef);
+
+/* so that underneath can avoid function calls which are prohibited
+ * for static initialization we need this conversion */
+enum ll_type {
+   LL_Double,
+   LL_Float,
+   LL_Int32,
+   LL_Int16,
+   LL_Int8
+};
+
+static INLINE LLVMTypeRef
+ll_type_to_llvm(enum ll_type type)
+{
+   switch (type) {
+   case LL_Double:
+      return LLVMDoubleType();
+   case LL_Float:
+      return LLVMFloatType();
+   case LL_Int32:
+      return LLVMInt32Type();
+   case LL_Int16:
+      return LLVMIntType(16);
+   case LL_Int8:
+      return LLVMIntType(8);
+   }
+   return LLVMIntType(8);
+}
+
+static INLINE int
+ll_type_size(enum ll_type type)
+{
+   switch (type) {
+   case LL_Double:
+      return 8;
+   case LL_Float:
+      return 4;
+   case LL_Int32:
+      return 4;
+   case LL_Int16:
+      return 2;
+   case LL_Int8:
+      return 1;
+   }
+   return 1;
+}
+
+struct draw_llvm_translate {
+   int format;
+   from_func from;
+   to_func to;
+   enum ll_type type;
+   int num_components;
+} translates[] =
+{
+   {PIPE_FORMAT_R64_FLOAT,          from_64_float, to_64_float, LL_Double, 1},
+   {PIPE_FORMAT_R64G64_FLOAT,       from_64_float, to_64_float, LL_Double, 2},
+   {PIPE_FORMAT_R64G64B64_FLOAT,    from_64_float, to_64_float, LL_Double, 3},
+   {PIPE_FORMAT_R64G64B64A64_FLOAT, from_64_float, to_64_float, LL_Double, 4},
+   {PIPE_FORMAT_R32_FLOAT,          from_32_float, to_32_float, LL_Float, 1},
+   {PIPE_FORMAT_R32G32_FLOAT,       from_32_float, to_32_float, LL_Float, 2},
+   {PIPE_FORMAT_R32G32B32_FLOAT,    from_32_float, to_32_float, LL_Float, 3},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, from_32_float, to_32_float, LL_Float, 4},
+
+   {PIPE_FORMAT_R32_UNORM,          from_32_unorm, to_32_unorm, LL_Int32, 1},
+   {PIPE_FORMAT_R32G32_UNORM,       from_32_unorm, to_32_unorm, LL_Int32, 2},
+   {PIPE_FORMAT_R32G32B32_UNORM,    from_32_unorm, to_32_unorm, LL_Int32, 3},
+   {PIPE_FORMAT_R32G32B32A32_UNORM, from_32_unorm, to_32_unorm, LL_Int32, 4},
+
+   {PIPE_FORMAT_R32_USCALED,          from_32_uscaled, to_32_uscaled, LL_Int32, 1},
+   {PIPE_FORMAT_R32G32_USCALED,       from_32_uscaled, to_32_uscaled, LL_Int32, 2},
+   {PIPE_FORMAT_R32G32B32_USCALED,    from_32_uscaled, to_32_uscaled, LL_Int32, 3},
+   {PIPE_FORMAT_R32G32B32A32_USCALED, from_32_uscaled, to_32_uscaled, LL_Int32, 4},
+
+   {PIPE_FORMAT_R32_SNORM,          from_32_snorm, to_32_snorm, LL_Int32, 1},
+   {PIPE_FORMAT_R32G32_SNORM,       from_32_snorm, to_32_snorm, LL_Int32, 2},
+   {PIPE_FORMAT_R32G32B32_SNORM,    from_32_snorm, to_32_snorm, LL_Int32, 3},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, from_32_snorm, to_32_snorm, LL_Int32, 4},
+
+   {PIPE_FORMAT_R32_SSCALED,          from_32_sscaled, to_32_sscaled, LL_Int32, 1},
+   {PIPE_FORMAT_R32G32_SSCALED,       from_32_sscaled, to_32_sscaled, LL_Int32, 2},
+   {PIPE_FORMAT_R32G32B32_SSCALED,    from_32_sscaled, to_32_sscaled, LL_Int32, 3},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, from_32_sscaled, to_32_sscaled, LL_Int32, 4},
+
+   {PIPE_FORMAT_R16_UNORM,          from_16_unorm, to_16_unorm, LL_Int16, 1},
+   {PIPE_FORMAT_R16G16_UNORM,       from_16_unorm, to_16_unorm, LL_Int16, 2},
+   {PIPE_FORMAT_R16G16B16_UNORM,    from_16_unorm, to_16_unorm, LL_Int16, 3},
+   {PIPE_FORMAT_R16G16B16A16_UNORM, from_16_unorm, to_16_unorm, LL_Int16, 4},
+
+   {PIPE_FORMAT_R16_USCALED,          from_16_uscaled, to_16_uscaled, LL_Int16, 1},
+   {PIPE_FORMAT_R16G16_USCALED,       from_16_uscaled, to_16_uscaled, LL_Int16, 2},
+   {PIPE_FORMAT_R16G16B16_USCALED,    from_16_uscaled, to_16_uscaled, LL_Int16, 3},
+   {PIPE_FORMAT_R16G16B16A16_USCALED, from_16_uscaled, to_16_uscaled, LL_Int16, 4},
+
+   {PIPE_FORMAT_R16_SNORM,          from_16_snorm, to_16_snorm, LL_Int16, 1},
+   {PIPE_FORMAT_R16G16_SNORM,       from_16_snorm, to_16_snorm, LL_Int16, 2},
+   {PIPE_FORMAT_R16G16B16_SNORM,    from_16_snorm, to_16_snorm, LL_Int16, 3},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, from_16_snorm, to_16_snorm, LL_Int16, 4},
+
+   {PIPE_FORMAT_R16_SSCALED,          from_16_sscaled, to_16_sscaled, LL_Int16, 1},
+   {PIPE_FORMAT_R16G16_SSCALED,       from_16_sscaled, to_16_sscaled, LL_Int16, 2},
+   {PIPE_FORMAT_R16G16B16_SSCALED,    from_16_sscaled, to_16_sscaled, LL_Int16, 3},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, from_16_sscaled, to_16_sscaled, LL_Int16, 4},
+
+   {PIPE_FORMAT_R8_UNORM,       from_8_unorm, to_8_unorm, LL_Int8, 1},
+   {PIPE_FORMAT_R8G8_UNORM,     from_8_unorm, to_8_unorm, LL_Int8, 2},
+   {PIPE_FORMAT_R8G8B8_UNORM,   from_8_unorm, to_8_unorm, LL_Int8, 3},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, from_8_unorm, to_8_unorm, LL_Int8, 4},
+
+   {PIPE_FORMAT_R8_USCALED,       from_8_uscaled, to_8_uscaled, LL_Int8, 1},
+   {PIPE_FORMAT_R8G8_USCALED,     from_8_uscaled, to_8_uscaled, LL_Int8, 2},
+   {PIPE_FORMAT_R8G8B8_USCALED,   from_8_uscaled, to_8_uscaled, LL_Int8, 3},
+   {PIPE_FORMAT_R8G8B8A8_USCALED, from_8_uscaled, to_8_uscaled, LL_Int8, 4},
+
+   {PIPE_FORMAT_R8_SNORM,       from_8_snorm, to_8_snorm, LL_Int8, 1},
+   {PIPE_FORMAT_R8G8_SNORM,     from_8_snorm, to_8_snorm, LL_Int8, 2},
+   {PIPE_FORMAT_R8G8B8_SNORM,   from_8_snorm, to_8_snorm, LL_Int8, 3},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, from_8_snorm, to_8_snorm, LL_Int8, 4},
+
+   {PIPE_FORMAT_R8_SSCALED,       from_8_sscaled, to_8_sscaled, LL_Int8, 1},
+   {PIPE_FORMAT_R8G8_SSCALED,     from_8_sscaled, to_8_sscaled, LL_Int8, 2},
+   {PIPE_FORMAT_R8G8B8_SSCALED,   from_8_sscaled, to_8_sscaled, LL_Int8, 3},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, from_8_sscaled, to_8_sscaled, LL_Int8, 4},
+
+   {PIPE_FORMAT_R32_FIXED,          from_32_fixed, to_32_fixed, LL_Int32, 1},
+   {PIPE_FORMAT_R32G32_FIXED,       from_32_fixed, to_32_fixed, LL_Int32, 2},
+   {PIPE_FORMAT_R32G32B32_FIXED,    from_32_fixed, to_32_fixed, LL_Int32, 3},
+   {PIPE_FORMAT_R32G32B32A32_FIXED, from_32_fixed, to_32_fixed, LL_Int32, 4},
+};
+
+
+static LLVMValueRef
+fetch(LLVMBuilderRef builder,
+      LLVMValueRef ptr, int val_size, int nr_components,
+      from_func func)
+{
+   int i;
+   int offset = 0;
+   LLVMValueRef res = LLVMConstNull(
+      LLVMVectorType(LLVMFloatType(), 4));
+   LLVMValueRef defaults[4];
+
+   defaults[0] = LLVMConstReal(LLVMFloatType(), 0);
+   defaults[1] = LLVMConstReal(LLVMFloatType(), 0);
+   defaults[2] = LLVMConstReal(LLVMFloatType(), 0);
+   defaults[3] = LLVMConstReal(LLVMFloatType(), 1);
+
+   for (i = 0; i < nr_components; ++i) {
+      LLVMValueRef src_index = LLVMConstInt(LLVMInt32Type(), offset, 0);
+      LLVMValueRef dst_index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef src_tmp;
+      LLVMValueRef component;
+
+      src_tmp = LLVMBuildGEP(builder, ptr, &src_index, 1, "src_tmp");
+
+      /* convert src_tmp to float */
+      component = func(builder, src_tmp);
+
+      /* vec.comp = component */
+      res = LLVMBuildInsertElement(builder,
+                                   res,
+                                   component,
+                                   dst_index, "");
+      offset += val_size;
+   }
+   for (; i < 4; ++i) {
+      LLVMValueRef dst_index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      res = LLVMBuildInsertElement(builder,
+                                   res,
+                                   defaults[i],
+                                   dst_index, "");
+   }
+   return res;
+}
+
+
+LLVMValueRef
+draw_llvm_translate_from(LLVMBuilderRef builder,
+                         LLVMValueRef vbuffer,
+                         enum pipe_format from_format)
+{
+   const struct util_format_description *format_desc;
+   LLVMValueRef zero;
+   int i;
+
+   /*
+    * The above can only cope with straight arrays: no bitfields,
+    * swizzles, or half floats.
+    */
+
+   for (i = 0; i < Elements(translates); ++i) {
+      if (translates[i].format == from_format) {
+         /*LLVMTypeRef type = ll_type_to_llvm(translates[i].type);*/
+         return fetch(builder,
+                      vbuffer,
+                      ll_type_size(translates[i].type),
+                      translates[i].num_components,
+                      translates[i].from);
+      }
+   }
+
+
+   /*
+    * This doesn't handle anything bigger than 32bits, or half floats
+    * yet.
+    *
+    * TODO: unify all this code into lp_build_fetch_rgba_aos().
+    */
+
+   format_desc = util_format_description(from_format);
+   zero = LLVMConstNull(LLVMInt32Type());
+   return lp_build_fetch_rgba_aos(builder, format_desc, vbuffer, zero, zero);
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c
index 83dc1a35f4c..64c35025081 100644
--- a/src/gallium/auxiliary/draw/draw_pipe.c
+++ b/src/gallium/auxiliary/draw/draw_pipe.c
@@ -66,6 +66,7 @@ boolean draw_pipeline_init( struct draw_context *draw )
    /* these defaults are oriented toward the needs of softpipe */
    draw->pipeline.wide_point_threshold = 1000000.0; /* infinity */
    draw->pipeline.wide_line_threshold = 1.0;
+   draw->pipeline.wide_point_sprites = FALSE;
    draw->pipeline.line_stipple = TRUE;
    draw->pipeline.point_sprite = TRUE;
 
@@ -225,7 +226,7 @@ static void do_triangle( struct draw_context *draw,
 
 
 /**
- * Code to run the pipeline on a fairly arbitary collection of vertices.
+ * Code to run the pipeline on a fairly arbitrary collection of vertices.
  * For drawing indexed primitives.
  *
  * Vertex headers must be pre-initialized with the
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
index 8f6ca15dfa2..4faf0a779ca 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -40,6 +40,7 @@
 #include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
+#include "util/u_sampler.h"
 
 #include "tgsi/tgsi_transform.h"
 #include "tgsi/tgsi_dump.h"
@@ -87,9 +88,10 @@ struct aaline_stage
    uint pos_slot;
 
    void *sampler_cso;
-   struct pipe_texture *texture;
+   struct pipe_resource *texture;
+   struct pipe_sampler_view *sampler_view;
    uint num_samplers;
-   uint num_textures;
+   uint num_sampler_views;
 
 
    /*
@@ -98,7 +100,7 @@ struct aaline_stage
    struct aaline_fragment_shader *fs;
    struct {
       void *sampler[PIPE_MAX_SAMPLERS];
-      struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+      struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
    } state;
 
    /*
@@ -111,10 +113,10 @@ struct aaline_stage
 
    void (*driver_bind_sampler_states)(struct pipe_context *, unsigned,
                                       void **);
-   void (*driver_set_sampler_textures)(struct pipe_context *, unsigned,
-                                       struct pipe_texture **);
 
-   struct pipe_context *pipe;
+   void (*driver_set_sampler_views)(struct pipe_context *,
+                                    unsigned,
+                                    struct pipe_sampler_view **);
 };
 
 
@@ -339,6 +341,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
 static boolean
 generate_aaline_fs(struct aaline_stage *aaline)
 {
+   struct pipe_context *pipe = aaline->stage.draw->pipe;
    const struct pipe_shader_state *orig_fs = &aaline->fs->state;
    struct pipe_shader_state aaline_fs;
    struct aa_transform_context transform;
@@ -371,7 +374,7 @@ generate_aaline_fs(struct aaline_stage *aaline)
    aaline->fs->sampler_unit = transform.freeSampler;
 
    aaline->fs->aaline_fs
-      = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs);
+      = aaline->driver_create_fs_state(pipe, &aaline_fs);
    if (aaline->fs->aaline_fs == NULL)
       goto fail;
 
@@ -391,9 +394,10 @@ fail:
 static boolean
 aaline_create_texture(struct aaline_stage *aaline)
 {
-   struct pipe_context *pipe = aaline->pipe;
+   struct pipe_context *pipe = aaline->stage.draw->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_texture texTemp;
+   struct pipe_resource texTemp;
+   struct pipe_sampler_view viewTempl;
    uint level;
 
    memset(&texTemp, 0, sizeof(texTemp));
@@ -403,28 +407,46 @@ aaline_create_texture(struct aaline_stage *aaline)
    texTemp.width0 = 1 << MAX_TEXTURE_LEVEL;
    texTemp.height0 = 1 << MAX_TEXTURE_LEVEL;
    texTemp.depth0 = 1;
+   texTemp.bind = PIPE_BIND_SAMPLER_VIEW;
 
-   aaline->texture = screen->texture_create(screen, &texTemp);
+   aaline->texture = screen->resource_create(screen, &texTemp);
    if (!aaline->texture)
       return FALSE;
 
+   u_sampler_view_default_template(&viewTempl,
+                                   aaline->texture,
+                                   aaline->texture->format);
+   aaline->sampler_view = pipe->create_sampler_view(pipe,
+                                                    aaline->texture,
+                                                    &viewTempl);
+   if (!aaline->sampler_view) {
+      return FALSE;
+   }
+
    /* Fill in mipmap images.
     * Basically each level is solid opaque, except for the outermost
     * texels which are zero.  Special case the 1x1 and 2x2 levels.
     */
    for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) {
       struct pipe_transfer *transfer;
+      struct pipe_box box;
       const uint size = u_minify(aaline->texture->width0, level);
       ubyte *data;
       uint i, j;
 
       assert(aaline->texture->width0 == aaline->texture->height0);
 
+      u_box_origin_2d( size, size, &box );
+
       /* This texture is new, no need to flush. 
        */
-      transfer = screen->get_tex_transfer(screen, aaline->texture, 0, level, 0,
-                                         PIPE_TRANSFER_WRITE, 0, 0, size, size);
-      data = screen->transfer_map(screen, transfer);
+      transfer = pipe->get_transfer(pipe,
+				    aaline->texture,
+				    u_subresource(0, level), 
+				    PIPE_TRANSFER_WRITE,
+				    &box);
+
+      data = pipe->transfer_map(pipe, transfer);
       if (data == NULL)
          return FALSE;
 
@@ -448,8 +470,8 @@ aaline_create_texture(struct aaline_stage *aaline)
       }
 
       /* unmap */
-      screen->transfer_unmap(screen, transfer);
-      screen->tex_transfer_destroy(transfer);
+      pipe->transfer_unmap(pipe, transfer);
+      pipe->transfer_destroy(pipe, transfer);
    }
    return TRUE;
 }
@@ -464,7 +486,7 @@ static boolean
 aaline_create_sampler(struct aaline_stage *aaline)
 {
    struct pipe_sampler_state sampler;
-   struct pipe_context *pipe = aaline->pipe;
+   struct pipe_context *pipe = aaline->stage.draw->pipe;
 
    memset(&sampler, 0, sizeof(sampler));
    sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
@@ -493,13 +515,14 @@ static boolean
 bind_aaline_fragment_shader(struct aaline_stage *aaline)
 {
    struct draw_context *draw = aaline->stage.draw;
+   struct pipe_context *pipe = draw->pipe;
 
    if (!aaline->fs->aaline_fs && 
        !generate_aaline_fs(aaline))
       return FALSE;
 
    draw->suspend_flushing = TRUE;
-   aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs);
+   aaline->driver_bind_fs_state(pipe, aaline->fs->aaline_fs);
    draw->suspend_flushing = FALSE;
 
    return TRUE;
@@ -639,8 +662,10 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header)
 {
    auto struct aaline_stage *aaline = aaline_stage(stage);
    struct draw_context *draw = stage->draw;
-   struct pipe_context *pipe = aaline->pipe;
+   struct pipe_context *pipe = draw->pipe;
+   const struct pipe_rasterizer_state *rast = draw->rasterizer;
    uint num_samplers;
+   void *r;
 
    assert(draw->rasterizer->line_smooth);
 
@@ -669,16 +694,21 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header)
 
    /* how many samplers? */
    /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
-   num_samplers = MAX2(aaline->num_textures, aaline->num_samplers);
+   num_samplers = MAX2(aaline->num_sampler_views, aaline->num_samplers);
    num_samplers = MAX2(num_samplers, aaline->fs->sampler_unit + 1);
 
    aaline->state.sampler[aaline->fs->sampler_unit] = aaline->sampler_cso;
-   pipe_texture_reference(&aaline->state.texture[aaline->fs->sampler_unit],
-                          aaline->texture);
+   pipe_sampler_view_reference(&aaline->state.sampler_views[aaline->fs->sampler_unit],
+                               aaline->sampler_view);
 
    draw->suspend_flushing = TRUE;
    aaline->driver_bind_sampler_states(pipe, num_samplers, aaline->state.sampler);
-   aaline->driver_set_sampler_textures(pipe, num_samplers, aaline->state.texture);
+   aaline->driver_set_sampler_views(pipe, num_samplers, aaline->state.sampler_views);
+
+   /* Disable triangle culling, stippling, unfilled mode etc. */
+   r = draw_get_rasterizer_no_cull(draw, rast->scissor, rast->flatshade);
+   pipe->bind_rasterizer_state(pipe, r);
+
    draw->suspend_flushing = FALSE;
 
    /* now really draw first line */
@@ -692,7 +722,7 @@ aaline_flush(struct draw_stage *stage, unsigned flags)
 {
    struct draw_context *draw = stage->draw;
    struct aaline_stage *aaline = aaline_stage(stage);
-   struct pipe_context *pipe = aaline->pipe;
+   struct pipe_context *pipe = draw->pipe;
 
    stage->line = aaline_first_line;
    stage->next->flush( stage->next, flags );
@@ -702,8 +732,15 @@ aaline_flush(struct draw_stage *stage, unsigned flags)
    aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs);
    aaline->driver_bind_sampler_states(pipe, aaline->num_samplers,
                                       aaline->state.sampler);
-   aaline->driver_set_sampler_textures(pipe, aaline->num_textures,
-                                       aaline->state.texture);
+   aaline->driver_set_sampler_views(pipe,
+                                    aaline->num_sampler_views,
+                                    aaline->state.sampler_views);
+
+   /* restore original rasterizer state */
+   if (draw->rast_handle) {
+      pipe->bind_rasterizer_state(pipe, draw->rast_handle);
+   }
+
    draw->suspend_flushing = FALSE;
 
    draw->extra_shader_outputs.slot = 0;
@@ -721,17 +758,22 @@ static void
 aaline_destroy(struct draw_stage *stage)
 {
    struct aaline_stage *aaline = aaline_stage(stage);
+   struct pipe_context *pipe = stage->draw->pipe;
    uint i;
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      pipe_texture_reference(&aaline->state.texture[i], NULL);
+      pipe_sampler_view_reference(&aaline->state.sampler_views[i], NULL);
    }
 
    if (aaline->sampler_cso)
-      aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso);
+      pipe->delete_sampler_state(pipe, aaline->sampler_cso);
 
    if (aaline->texture)
-      pipe_texture_reference(&aaline->texture, NULL);
+      pipe_resource_reference(&aaline->texture, NULL);
+
+   if (aaline->sampler_view) {
+      pipe_sampler_view_reference(&aaline->sampler_view, NULL);
+   }
 
    draw_free_temp_verts( stage );
 
@@ -787,13 +829,14 @@ aaline_create_fs_state(struct pipe_context *pipe,
 {
    struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
    struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader);
+
    if (aafs == NULL)
       return NULL;
 
    aafs->state = *fs;
 
    /* pass-through */
-   aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs);
+   aafs->driver_fs = aaline->driver_create_fs_state(pipe, fs);
 
    return aafs;
 }
@@ -808,8 +851,7 @@ aaline_bind_fs_state(struct pipe_context *pipe, void *fs)
    /* save current */
    aaline->fs = aafs;
    /* pass-through */
-   aaline->driver_bind_fs_state(aaline->pipe,
-                                (aafs ? aafs->driver_fs : NULL));
+   aaline->driver_bind_fs_state(pipe, (aafs ? aafs->driver_fs : NULL));
 }
 
 
@@ -818,11 +860,12 @@ aaline_delete_fs_state(struct pipe_context *pipe, void *fs)
 {
    struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
    struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs;
+
    /* pass-through */
-   aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs);
+   aaline->driver_delete_fs_state(pipe, aafs->driver_fs);
 
    if (aafs->aaline_fs)
-      aaline->driver_delete_fs_state(aaline->pipe, aafs->aaline_fs);
+      aaline->driver_delete_fs_state(pipe, aafs->aaline_fs);
 
    FREE(aafs);
 }
@@ -839,28 +882,29 @@ aaline_bind_sampler_states(struct pipe_context *pipe,
    aaline->num_samplers = num;
 
    /* pass-through */
-   aaline->driver_bind_sampler_states(aaline->pipe, num, sampler);
+   aaline->driver_bind_sampler_states(pipe, num, sampler);
 }
 
 
 static void
-aaline_set_sampler_textures(struct pipe_context *pipe,
-                            unsigned num, struct pipe_texture **texture)
+aaline_set_sampler_views(struct pipe_context *pipe,
+                         unsigned num,
+                         struct pipe_sampler_view **views)
 {
    struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
    uint i;
 
    /* save current */
    for (i = 0; i < num; i++) {
-      pipe_texture_reference(&aaline->state.texture[i], texture[i]);
+      pipe_sampler_view_reference(&aaline->state.sampler_views[i], views[i]);
    }
    for ( ; i < PIPE_MAX_SAMPLERS; i++) {
-      pipe_texture_reference(&aaline->state.texture[i], NULL);
+      pipe_sampler_view_reference(&aaline->state.sampler_views[i], NULL);
    }
-   aaline->num_textures = num;
+   aaline->num_sampler_views = num;
 
    /* pass-through */
-   aaline->driver_set_sampler_textures(aaline->pipe, num, texture);
+   aaline->driver_set_sampler_views(pipe, num, views);
 }
 
 
@@ -883,8 +927,6 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe)
    if (!aaline)
       goto fail;
 
-   aaline->pipe = pipe;
-
    /* create special texture, sampler state */
    if (!aaline_create_texture(aaline))
       goto fail;
@@ -898,7 +940,7 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe)
    aaline->driver_delete_fs_state = pipe->delete_fs_state;
 
    aaline->driver_bind_sampler_states = pipe->bind_fragment_sampler_states;
-   aaline->driver_set_sampler_textures = pipe->set_fragment_sampler_textures;
+   aaline->driver_set_sampler_views = pipe->set_fragment_sampler_views;
 
    /* override the driver's functions */
    pipe->create_fs_state = aaline_create_fs_state;
@@ -906,7 +948,7 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe)
    pipe->delete_fs_state = aaline_delete_fs_state;
 
    pipe->bind_fragment_sampler_states = aaline_bind_sampler_states;
-   pipe->set_fragment_sampler_textures = aaline_set_sampler_textures;
+   pipe->set_fragment_sampler_views = aaline_set_sampler_views;
    
    /* Install once everything is known to be OK:
     */
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
index 9f9fb4312c1..bba6f50c020 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -107,8 +107,6 @@ struct aapoint_stage
                                     const struct pipe_shader_state *);
    void (*driver_bind_fs_state)(struct pipe_context *, void *);
    void (*driver_delete_fs_state)(struct pipe_context *, void *);
-
-   struct pipe_context *pipe;
 };
 
 
@@ -499,6 +497,7 @@ generate_aapoint_fs(struct aapoint_stage *aapoint)
    struct pipe_shader_state aapoint_fs;
    struct aa_transform_context transform;
    const uint newLen = tgsi_num_tokens(orig_fs->tokens) + NUM_NEW_TOKENS;
+   struct pipe_context *pipe = aapoint->stage.draw->pipe;
 
    aapoint_fs = *orig_fs; /* copy to init */
    aapoint_fs.tokens = tgsi_alloc_tokens(newLen);
@@ -527,7 +526,7 @@ generate_aapoint_fs(struct aapoint_stage *aapoint)
 #endif
 
    aapoint->fs->aapoint_fs
-      = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs);
+      = aapoint->driver_create_fs_state(pipe, &aapoint_fs);
    if (aapoint->fs->aapoint_fs == NULL)
       goto fail;
 
@@ -549,13 +548,14 @@ static boolean
 bind_aapoint_fragment_shader(struct aapoint_stage *aapoint)
 {
    struct draw_context *draw = aapoint->stage.draw;
+   struct pipe_context *pipe = draw->pipe;
 
    if (!aapoint->fs->aapoint_fs &&
        !generate_aapoint_fs(aapoint))
       return FALSE;
 
    draw->suspend_flushing = TRUE;
-   aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs);
+   aapoint->driver_bind_fs_state(pipe, aapoint->fs->aapoint_fs);
    draw->suspend_flushing = FALSE;
 
    return TRUE;
@@ -679,6 +679,9 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
 {
    auto struct aapoint_stage *aapoint = aapoint_stage(stage);
    struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = draw->pipe;
+   const struct pipe_rasterizer_state *rast = draw->rasterizer;
+   void *r;
 
    assert(draw->rasterizer->point_smooth);
 
@@ -716,6 +719,14 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
       }
    }
 
+   draw->suspend_flushing = TRUE;
+
+   /* Disable triangle culling, stippling, unfilled mode etc. */
+   r = draw_get_rasterizer_no_cull(draw, rast->scissor, rast->flatshade);
+   pipe->bind_rasterizer_state(pipe, r);
+
+   draw->suspend_flushing = FALSE;
+
    /* now really draw first point */
    stage->point = aapoint_point;
    stage->point(stage, header);
@@ -727,7 +738,7 @@ aapoint_flush(struct draw_stage *stage, unsigned flags)
 {
    struct draw_context *draw = stage->draw;
    struct aapoint_stage *aapoint = aapoint_stage(stage);
-   struct pipe_context *pipe = aapoint->pipe;
+   struct pipe_context *pipe = draw->pipe;
 
    stage->point = aapoint_first_point;
    stage->next->flush( stage->next, flags );
@@ -735,6 +746,12 @@ aapoint_flush(struct draw_stage *stage, unsigned flags)
    /* restore original frag shader */
    draw->suspend_flushing = TRUE;
    aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs);
+
+   /* restore original rasterizer state */
+   if (draw->rast_handle) {
+      pipe->bind_rasterizer_state(pipe, draw->rast_handle);
+   }
+
    draw->suspend_flushing = FALSE;
 
    draw->extra_shader_outputs.slot = 0;
@@ -811,7 +828,7 @@ aapoint_create_fs_state(struct pipe_context *pipe,
    aafs->state = *fs;
 
    /* pass-through */
-   aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs);
+   aafs->driver_fs = aapoint->driver_create_fs_state(pipe, fs);
 
    return aafs;
 }
@@ -825,7 +842,7 @@ aapoint_bind_fs_state(struct pipe_context *pipe, void *fs)
    /* save current */
    aapoint->fs = aafs;
    /* pass-through */
-   aapoint->driver_bind_fs_state(aapoint->pipe,
+   aapoint->driver_bind_fs_state(pipe,
                                  (aafs ? aafs->driver_fs : NULL));
 }
 
@@ -837,10 +854,10 @@ aapoint_delete_fs_state(struct pipe_context *pipe, void *fs)
    struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs;
 
    /* pass-through */
-   aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs);
+   aapoint->driver_delete_fs_state(pipe, aafs->driver_fs);
 
    if (aafs->aapoint_fs)
-      aapoint->driver_delete_fs_state(aapoint->pipe, aafs->aapoint_fs);
+      aapoint->driver_delete_fs_state(pipe, aafs->aapoint_fs);
 
    FREE(aafs);
 }
@@ -857,8 +874,6 @@ draw_install_aapoint_stage(struct draw_context *draw,
 {
    struct aapoint_stage *aapoint;
 
-   pipe->draw = (void *) draw;
-
    /*
     * Create / install AA point drawing / prim stage
     */
@@ -866,8 +881,6 @@ draw_install_aapoint_stage(struct draw_context *draw,
    if (aapoint == NULL)
       return FALSE;
 
-   aapoint->pipe = pipe;
-
    /* save original driver functions */
    aapoint->driver_create_fs_state = pipe->create_fs_state;
    aapoint->driver_bind_fs_state = pipe->bind_fs_state;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
index d0d99aa331a..ef30db094fe 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -42,6 +42,7 @@
 #include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
+#include "util/u_sampler.h"
 
 #include "tgsi/tgsi_transform.h"
 #include "tgsi/tgsi_dump.h"
@@ -74,9 +75,10 @@ struct pstip_stage
    struct draw_stage stage;
 
    void *sampler_cso;
-   struct pipe_texture *texture;
+   struct pipe_resource *texture;
+   struct pipe_sampler_view *sampler_view;
    uint num_samplers;
-   uint num_textures;
+   uint num_sampler_views;
 
    /*
     * Currently bound state
@@ -84,7 +86,7 @@ struct pstip_stage
    struct pstip_fragment_shader *fs;
    struct {
       void *samplers[PIPE_MAX_SAMPLERS];
-      struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+      struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
       const struct pipe_poly_stipple *stipple;
    } state;
 
@@ -98,8 +100,9 @@ struct pstip_stage
 
    void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, void **);
 
-   void (*driver_set_sampler_textures)(struct pipe_context *, unsigned,
-                                       struct pipe_texture **);
+   void (*driver_set_sampler_views)(struct pipe_context *,
+                                    unsigned,
+                                    struct pipe_sampler_view **);
 
    void (*driver_set_polygon_stipple)(struct pipe_context *,
                                       const struct pipe_poly_stipple *);
@@ -374,19 +377,21 @@ pstip_update_texture(struct pstip_stage *pstip)
 {
    static const uint bit31 = 1 << 31;
    struct pipe_context *pipe = pstip->pipe;
-   struct pipe_screen *screen = pipe->screen;
    struct pipe_transfer *transfer;
    const uint *stipple = pstip->state.stipple->stipple;
    uint i, j;
    ubyte *data;
 
    /* XXX: want to avoid flushing just because we use stipple: 
+    *
+    * Flush should no longer be necessary if driver is properly
+    * interleaving drawing and transfers on a given context:
     */
    pipe->flush( pipe, PIPE_FLUSH_TEXTURE_CACHE, NULL );
 
-   transfer = screen->get_tex_transfer(screen, pstip->texture, 0, 0, 0,
-                                       PIPE_TRANSFER_WRITE, 0, 0, 32, 32);
-   data = screen->transfer_map(screen, transfer);
+   transfer = pipe_get_transfer(pipe, pstip->texture, 0, 0, 0,
+				    PIPE_TRANSFER_WRITE, 0, 0, 32, 32);
+   data = pipe->transfer_map(pipe, transfer);
 
    /*
     * Load alpha texture.
@@ -408,8 +413,8 @@ pstip_update_texture(struct pstip_stage *pstip)
    }
 
    /* unmap */
-   screen->transfer_unmap(screen, transfer);
-   screen->tex_transfer_destroy(transfer);
+   pipe->transfer_unmap(pipe, transfer);
+   pipe->transfer_destroy(pipe, transfer);
 }
 
 
@@ -421,7 +426,8 @@ pstip_create_texture(struct pstip_stage *pstip)
 {
    struct pipe_context *pipe = pstip->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_texture texTemp;
+   struct pipe_resource texTemp;
+   struct pipe_sampler_view viewTempl;
 
    memset(&texTemp, 0, sizeof(texTemp));
    texTemp.target = PIPE_TEXTURE_2D;
@@ -430,11 +436,22 @@ pstip_create_texture(struct pstip_stage *pstip)
    texTemp.width0 = 32;
    texTemp.height0 = 32;
    texTemp.depth0 = 1;
+   texTemp.bind = PIPE_BIND_SAMPLER_VIEW;
 
-   pstip->texture = screen->texture_create(screen, &texTemp);
+   pstip->texture = screen->resource_create(screen, &texTemp);
    if (pstip->texture == NULL)
       return FALSE;
 
+   u_sampler_view_default_template(&viewTempl,
+                                   pstip->texture,
+                                   pstip->texture->format);
+   pstip->sampler_view = pipe->create_sampler_view(pipe,
+                                                   pstip->texture,
+                                                   &viewTempl);
+   if (!pstip->sampler_view) {
+      return FALSE;
+   }
+
    return TRUE;
 }
 
@@ -513,19 +530,19 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header)
 
    /* how many samplers? */
    /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
-   num_samplers = MAX2(pstip->num_textures, pstip->num_samplers);
+   num_samplers = MAX2(pstip->num_sampler_views, pstip->num_samplers);
    num_samplers = MAX2(num_samplers, pstip->fs->sampler_unit + 1);
 
    /* plug in our sampler, texture */
    pstip->state.samplers[pstip->fs->sampler_unit] = pstip->sampler_cso;
-   pipe_texture_reference(&pstip->state.textures[pstip->fs->sampler_unit],
-                          pstip->texture);
+   pipe_sampler_view_reference(&pstip->state.sampler_views[pstip->fs->sampler_unit],
+                               pstip->sampler_view);
 
    assert(num_samplers <= PIPE_MAX_SAMPLERS);
 
    draw->suspend_flushing = TRUE;
    pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers);
-   pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures);
+   pstip->driver_set_sampler_views(pipe, num_samplers, pstip->state.sampler_views);
    draw->suspend_flushing = FALSE;
 
    /* now really draw first triangle */
@@ -549,8 +566,9 @@ pstip_flush(struct draw_stage *stage, unsigned flags)
    pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs);
    pstip->driver_bind_sampler_states(pipe, pstip->num_samplers,
                                      pstip->state.samplers);
-   pstip->driver_set_sampler_textures(pipe, pstip->num_textures,
-                                      pstip->state.textures);
+   pstip->driver_set_sampler_views(pipe,
+                                   pstip->num_sampler_views,
+                                   pstip->state.sampler_views);
    draw->suspend_flushing = FALSE;
 }
 
@@ -569,12 +587,16 @@ pstip_destroy(struct draw_stage *stage)
    uint i;
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      pipe_texture_reference(&pstip->state.textures[i], NULL);
+      pipe_sampler_view_reference(&pstip->state.sampler_views[i], NULL);
    }
 
    pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso);
 
-   pipe_texture_reference(&pstip->texture, NULL);
+   pipe_resource_reference(&pstip->texture, NULL);
+
+   if (pstip->sampler_view) {
+      pipe_sampler_view_reference(&pstip->sampler_view, NULL);
+   }
 
    draw_free_temp_verts( stage );
    FREE( stage );
@@ -680,24 +702,25 @@ pstip_bind_sampler_states(struct pipe_context *pipe,
 
 
 static void
-pstip_set_sampler_textures(struct pipe_context *pipe,
-                           unsigned num, struct pipe_texture **texture)
+pstip_set_sampler_views(struct pipe_context *pipe,
+                        unsigned num,
+                        struct pipe_sampler_view **views)
 {
    struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
    uint i;
 
    /* save current */
    for (i = 0; i < num; i++) {
-      pipe_texture_reference(&pstip->state.textures[i], texture[i]);
+      pipe_sampler_view_reference(&pstip->state.sampler_views[i], views[i]);
    }
    for (; i < PIPE_MAX_SAMPLERS; i++) {
-      pipe_texture_reference(&pstip->state.textures[i], NULL);
+      pipe_sampler_view_reference(&pstip->state.sampler_views[i], NULL);
    }
 
-   pstip->num_textures = num;
+   pstip->num_sampler_views = num;
 
    /* pass-through */
-   pstip->driver_set_sampler_textures(pstip->pipe, num, texture);
+   pstip->driver_set_sampler_views(pstip->pipe, num, views);
 }
 
 
@@ -754,7 +777,7 @@ draw_install_pstipple_stage(struct draw_context *draw,
    pstip->driver_delete_fs_state = pipe->delete_fs_state;
 
    pstip->driver_bind_sampler_states = pipe->bind_fragment_sampler_states;
-   pstip->driver_set_sampler_textures = pipe->set_fragment_sampler_textures;
+   pstip->driver_set_sampler_views = pipe->set_fragment_sampler_views;
    pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple;
 
    /* override the driver's functions */
@@ -763,7 +786,7 @@ draw_install_pstipple_stage(struct draw_context *draw,
    pipe->delete_fs_state = pstip_delete_fs_state;
 
    pipe->bind_fragment_sampler_states = pstip_bind_sampler_states;
-   pipe->set_fragment_sampler_textures = pstip_set_sampler_textures;
+   pipe->set_fragment_sampler_views = pstip_set_sampler_views;
    pipe->set_polygon_stipple = pstip_set_polygon_stipple;
 
    return TRUE;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c
index 153097e543e..2a50af7a414 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_validate.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c
@@ -100,6 +100,11 @@ draw_need_pipeline(const struct draw_context *draw,
       if (rasterizer->point_size > draw->pipeline.wide_point_threshold)
          return TRUE;
 
+      /* sprite points */
+      if (rasterizer->point_quad_rasterization
+          && draw->pipeline.wide_point_sprites)
+         return TRUE;
+
       /* AA points */
       if (rasterizer->point_smooth && draw->pipeline.aapoint)
          return TRUE;
@@ -154,6 +159,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
    boolean need_det = FALSE;
    boolean precalc_flat = FALSE;
    boolean wide_lines, wide_points;
+   const struct pipe_rasterizer_state *rast = draw->rasterizer;
 
    /* Set the validate's next stage to the rasterize stage, so that it
     * can be found later if needed for flushing.
@@ -161,15 +167,17 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
    stage->next = next;
 
    /* drawing wide lines? */
-   wide_lines = (draw->rasterizer->line_width > draw->pipeline.wide_line_threshold
-                 && !draw->rasterizer->line_smooth);
+   wide_lines = (rast->line_width > draw->pipeline.wide_line_threshold
+                 && !rast->line_smooth);
 
    /* drawing large points? */
-   if (draw->rasterizer->sprite_coord_enable && draw->pipeline.point_sprite)
+   if (rast->sprite_coord_enable && draw->pipeline.point_sprite)
       wide_points = TRUE;
-   else if (draw->rasterizer->point_smooth && draw->pipeline.aapoint)
+   else if (rast->point_smooth && draw->pipeline.aapoint)
       wide_points = FALSE;
-   else if (draw->rasterizer->point_size > draw->pipeline.wide_point_threshold)
+   else if (rast->point_size > draw->pipeline.wide_point_threshold)
+      wide_points = TRUE;
+   else if (rast->point_quad_rasterization && draw->pipeline.wide_point_sprites)
       wide_points = TRUE;
    else
       wide_points = FALSE;
@@ -181,12 +189,12 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
     * shorter pipelines for lines & points.
     */
 
-   if (draw->rasterizer->line_smooth && draw->pipeline.aaline) {
+   if (rast->line_smooth && draw->pipeline.aaline) {
       draw->pipeline.aaline->next = next;
       next = draw->pipeline.aaline;
    }
 
-   if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) {
+   if (rast->point_smooth && draw->pipeline.aapoint) {
       draw->pipeline.aapoint->next = next;
       next = draw->pipeline.aapoint;
    }
@@ -197,44 +205,44 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
       precalc_flat = TRUE;
    }
 
-   if (wide_points || draw->rasterizer->sprite_coord_enable) {
+   if (wide_points || rast->sprite_coord_enable) {
       draw->pipeline.wide_point->next = next;
       next = draw->pipeline.wide_point;
    }
 
-   if (draw->rasterizer->line_stipple_enable && draw->pipeline.line_stipple) {
+   if (rast->line_stipple_enable && draw->pipeline.line_stipple) {
       draw->pipeline.stipple->next = next;
       next = draw->pipeline.stipple;
       precalc_flat = TRUE;		/* only needed for lines really */
    }
 
-   if (draw->rasterizer->poly_stipple_enable
+   if (rast->poly_stipple_enable
        && draw->pipeline.pstipple) {
       draw->pipeline.pstipple->next = next;
       next = draw->pipeline.pstipple;
    }
 
-   if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
-       draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) {
+   if (rast->fill_cw != PIPE_POLYGON_MODE_FILL ||
+       rast->fill_ccw != PIPE_POLYGON_MODE_FILL) {
       draw->pipeline.unfilled->next = next;
       next = draw->pipeline.unfilled;
       precalc_flat = TRUE;		/* only needed for triangles really */
       need_det = TRUE;
    }
 
-   if (draw->rasterizer->flatshade && precalc_flat) {
+   if (rast->flatshade && precalc_flat) {
       draw->pipeline.flatshade->next = next;
       next = draw->pipeline.flatshade;
    }
 	 
-   if (draw->rasterizer->offset_cw ||
-       draw->rasterizer->offset_ccw) {
+   if (rast->offset_cw ||
+       rast->offset_ccw) {
       draw->pipeline.offset->next = next;
       next = draw->pipeline.offset;
       need_det = TRUE;
    }
 
-   if (draw->rasterizer->light_twoside) {
+   if (rast->light_twoside) {
       draw->pipeline.twoside->next = next;
       next = draw->pipeline.twoside;
       need_det = TRUE;
@@ -247,7 +255,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
     * to less work emitting vertices, smaller vertex buffers, etc.
     * It's difficult to say whether this will be true in general.
     */
-   if (need_det || draw->rasterizer->cull_mode) {
+   if (need_det || rast->cull_mode) {
       draw->pipeline.cull->next = next;
       next = draw->pipeline.cull;
    }
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
index 27099579618..abbf6247ab8 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -137,7 +137,7 @@ emit_vertex( struct vbuf_stage *vbuf,
        */
       /* Note: we really do want data[0] here, not data[pos]: 
        */
-      vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0);
+      vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0, ~0);
       vbuf->translate->run(vbuf->translate, 0, 1, 0, vbuf->vertex_ptr);
 
       if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr);
@@ -235,41 +235,18 @@ vbuf_start_prim( struct vbuf_stage *vbuf, uint prim )
    for (i = 0; i < vbuf->vinfo->num_attribs; i++) {
       unsigned emit_sz = 0;
       unsigned src_buffer = 0;
-      unsigned output_format;
+      enum pipe_format output_format;
       unsigned src_offset = (vbuf->vinfo->attrib[i].src_index * 4 * sizeof(float) );
 
-      switch (vbuf->vinfo->attrib[i].emit) {
-      case EMIT_4F:
-	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
-	 emit_sz = 4 * sizeof(float);
-	 break;
-      case EMIT_3F:
-	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
-	 emit_sz = 3 * sizeof(float);
-	 break;
-      case EMIT_2F:
-	 output_format = PIPE_FORMAT_R32G32_FLOAT;
-	 emit_sz = 2 * sizeof(float);
-	 break;
-      case EMIT_1F:
-	 output_format = PIPE_FORMAT_R32_FLOAT;
-	 emit_sz = 1 * sizeof(float);
-	 break;
-      case EMIT_1F_PSIZE:
-	 output_format = PIPE_FORMAT_R32_FLOAT;
-	 emit_sz = 1 * sizeof(float);
+      output_format = draw_translate_vinfo_format(vbuf->vinfo->attrib[i].emit);
+      emit_sz = draw_translate_vinfo_size(vbuf->vinfo->attrib[i].emit);
+
+      /* doesn't handle EMIT_OMIT */
+      assert(emit_sz != 0);
+
+      if (vbuf->vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
 	 src_buffer = 1;
 	 src_offset = 0;
-	 break;
-      case EMIT_4UB:
-	 output_format = PIPE_FORMAT_A8R8G8B8_UNORM;
-	 emit_sz = 4 * sizeof(ubyte);
-         break;
-      default:
-	 assert(0);
-	 output_format = PIPE_FORMAT_NONE;
-	 emit_sz = 0;
-	 break;
       }
 
       hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
@@ -294,7 +271,7 @@ vbuf_start_prim( struct vbuf_stage *vbuf, uint prim )
       translate_key_sanitize(&hw_key);
       vbuf->translate = translate_cache_find(vbuf->cache, &hw_key);
 
-      vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0);
+      vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0, ~0);
    }
 
    vbuf->point_size = vbuf->stage.draw->rasterizer->point_size;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
index 3073c870825..ab167065815 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
@@ -28,6 +28,7 @@
 /* Authors:  Keith Whitwell <[email protected]>
  */
 
+#include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
 #include "util/u_math.h"
@@ -53,7 +54,6 @@ static INLINE struct wideline_stage *wideline_stage( struct draw_stage *stage )
 
 /**
  * Draw a wide line by drawing a quad (two triangles).
- * XXX need to disable polygon stipple.
  */
 static void wideline_line( struct draw_stage *stage,
                            struct prim_header *header )
@@ -142,9 +142,40 @@ static void wideline_line( struct draw_stage *stage,
 }
 
 
+static void wideline_first_line( struct draw_stage *stage, 
+                                 struct prim_header *header )
+{
+   struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = draw->pipe;
+   const struct pipe_rasterizer_state *rast = draw->rasterizer;
+   void *r;
+
+   /* Disable triangle culling, stippling, unfilled mode etc. */
+   r = draw_get_rasterizer_no_cull(draw, rast->scissor, rast->flatshade);
+   draw->suspend_flushing = TRUE;
+   pipe->bind_rasterizer_state(pipe, r);
+   draw->suspend_flushing = FALSE;
+
+   stage->line = wideline_line;
+
+   wideline_line(stage, header);
+}
+
+
 static void wideline_flush( struct draw_stage *stage, unsigned flags )
 {
+   struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = draw->pipe;
+
+   stage->line = wideline_first_line;
    stage->next->flush( stage->next, flags );
+
+   /* restore original rasterizer state */
+   if (draw->rast_handle) {
+      draw->suspend_flushing = TRUE;
+      pipe->bind_rasterizer_state(pipe, draw->rast_handle);
+      draw->suspend_flushing = FALSE;
+   }
 }
 
 
@@ -171,7 +202,7 @@ struct draw_stage *draw_wide_line_stage( struct draw_context *draw )
    wide->stage.name = "wide-line";
    wide->stage.next = NULL;
    wide->stage.point = draw_pipe_passthrough_point;
-   wide->stage.line = wideline_line;
+   wide->stage.line = wideline_first_line;
    wide->stage.tri = draw_pipe_passthrough_tri;
    wide->stage.flush = wideline_flush;
    wide->stage.reset_stipple_counter = wideline_reset_stipple_counter;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index fdabce7d443..a86fe19586c 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -52,6 +52,7 @@
  */
 
 
+#include "pipe/p_context.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "pipe/p_defines.h"
@@ -113,7 +114,10 @@ static void set_texcoords(const struct widepoint_stage *wide,
       /* put gl_PointCoord into the extra vertex slot */
       uint slot = wide->stage.draw->extra_shader_outputs.slot;
       v->data[slot][0] = tc[0];
-      v->data[slot][1] = tc[1];
+      if (wide->texcoord_mode == PIPE_SPRITE_COORD_LOWER_LEFT)
+         v->data[slot][1] = 1.0f - tc[1];
+      else
+         v->data[slot][1] = tc[1];
       v->data[slot][2] = 0.0F;
       v->data[slot][3] = 1.0F;
    }
@@ -128,10 +132,9 @@ static void set_texcoords(const struct widepoint_stage *wide,
 static void widepoint_point( struct draw_stage *stage,
                              struct prim_header *header )
 {
-   /* XXX should take point_quad_rasterization into account? */
    const struct widepoint_stage *wide = widepoint_stage(stage);
    const unsigned pos = draw_current_shader_position_output(stage->draw);
-   const boolean sprite = (boolean) stage->draw->rasterizer->sprite_coord_enable;
+   const boolean sprite = (boolean) stage->draw->rasterizer->point_quad_rasterization;
    float half_size;
    float left_adj, right_adj, bot_adj, top_adj;
 
@@ -213,33 +216,42 @@ static void widepoint_first_point( struct draw_stage *stage,
 {
    struct widepoint_stage *wide = widepoint_stage(stage);
    struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = draw->pipe;
+   const struct pipe_rasterizer_state *rast = draw->rasterizer;
+   void *r;
 
-   wide->half_point_size = 0.5f * draw->rasterizer->point_size;
+   wide->half_point_size = 0.5f * rast->point_size;
    wide->xbias = 0.0;
    wide->ybias = 0.0;
 
-   if (draw->rasterizer->gl_rasterization_rules) {
+   if (rast->gl_rasterization_rules) {
       wide->xbias = 0.125;
    }
 
+   /* Disable triangle culling, stippling, unfilled mode etc. */
+   r = draw_get_rasterizer_no_cull(draw, rast->scissor, rast->flatshade);
+   draw->suspend_flushing = TRUE;
+   pipe->bind_rasterizer_state(pipe, r);
+   draw->suspend_flushing = FALSE;
+
    /* XXX we won't know the real size if it's computed by the vertex shader! */
-   if ((draw->rasterizer->point_size > draw->pipeline.wide_point_threshold) ||
-       (draw->rasterizer->sprite_coord_enable && draw->pipeline.point_sprite)) {
+   if ((rast->point_size > draw->pipeline.wide_point_threshold) ||
+       (rast->point_quad_rasterization && draw->pipeline.point_sprite)) {
       stage->point = widepoint_point;
    }
    else {
       stage->point = draw_pipe_passthrough_point;
    }
 
-   if (draw->rasterizer->sprite_coord_enable) {
+   if (rast->point_quad_rasterization) {
       /* find vertex shader texcoord outputs */
       const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
       uint i, j = 0;
-      wide->texcoord_mode = draw->rasterizer->sprite_coord_mode;
+      wide->texcoord_mode = rast->sprite_coord_mode;
       for (i = 0; i < vs->info.num_outputs; i++) {
          if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) {
             wide->texcoord_slot[j] = i;
-            wide->texcoord_enable[j] = (draw->rasterizer->sprite_coord_enable >> j) & 1;
+            wide->texcoord_enable[j] = (rast->sprite_coord_enable >> j) & 1;
             j++;
          }
       }
@@ -259,7 +271,7 @@ static void widepoint_first_point( struct draw_stage *stage,
    }
 
    wide->psize_slot = -1;
-   if (draw->rasterizer->point_size_per_vertex) {
+   if (rast->point_size_per_vertex) {
       /* find PSIZ vertex output */
       const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
       uint i;
@@ -277,9 +289,19 @@ static void widepoint_first_point( struct draw_stage *stage,
 
 static void widepoint_flush( struct draw_stage *stage, unsigned flags )
 {
+   struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = draw->pipe;
+
    stage->point = widepoint_first_point;
    stage->next->flush( stage->next, flags );
    stage->draw->extra_shader_outputs.slot = 0;
+
+   /* restore original rasterizer state */
+   if (draw->rast_handle) {
+      draw->suspend_flushing = TRUE;
+      pipe->bind_rasterizer_state(pipe, draw->rast_handle);
+      draw->suspend_flushing = FALSE;
+   }
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 1e6e01af9e2..a2bfb693c09 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -46,6 +46,10 @@
 
 #include "tgsi/tgsi_scan.h"
 
+#ifdef HAVE_LLVM
+#include <llvm-c/ExecutionEngine.h>
+#endif
+
 
 struct pipe_context;
 struct draw_vertex_shader;
@@ -82,6 +86,8 @@ struct vertex_header {
  */
 struct draw_context
 {
+   struct pipe_context *pipe;
+
    /** Drawing/primitive pipeline stages */
    struct {
       struct draw_stage *first;  /**< one of the following */
@@ -105,6 +111,7 @@ struct draw_context
 
       float wide_point_threshold; /**< convert pnts to tris if larger than this */
       float wide_line_threshold;  /**< convert lines to tris if wider than this */
+      boolean wide_point_sprites; /**< convert points to tris for sprite mode */
       boolean line_stipple;       /**< do line stipple? */
       boolean point_sprite;       /**< convert points to quads for sprites? */
 
@@ -125,6 +132,7 @@ struct draw_context
          struct draw_pt_middle_end *fetch_emit;
          struct draw_pt_middle_end *fetch_shade_emit;
          struct draw_pt_middle_end *general;
+         struct draw_pt_middle_end *llvm;
       } middle;
 
       struct {
@@ -144,6 +152,7 @@ struct draw_context
          const void *elts;
          /** bytes per index (0, 1, 2 or 4) */
          unsigned eltSize;
+         int eltBias;
          unsigned min_index;
          unsigned max_index;
          
@@ -174,8 +183,14 @@ struct draw_context
 
    double mrd;  /**< minimum resolvable depth value, for polygon offset */
 
-   /* pipe state that we need: */
+   /** Current rasterizer state given to us by the driver */
    const struct pipe_rasterizer_state *rasterizer;
+   /** Driver CSO handle for the current rasterizer state */
+   void *rast_handle;
+
+   /** Rasterizer CSOs without culling/stipple/etc */
+   void *rasterizer_no_cull[2][2];
+
    struct pipe_viewport_state viewport;
    boolean identity_viewport;
 
@@ -237,9 +252,17 @@ struct draw_context
 
    unsigned instance_id;
 
+#ifdef HAVE_LLVM
+   LLVMExecutionEngineRef engine;
+#endif
+
    void *driver_private;
 };
 
+/*******************************************************************************
+ * Draw common initialization code
+ */
+boolean draw_init(struct draw_context *draw);
 
 /*******************************************************************************
  * Vertex shader code:
@@ -345,5 +368,10 @@ void draw_do_flush( struct draw_context *draw, unsigned flags );
 
 
 
+void *
+draw_get_rasterizer_no_cull( struct draw_context *draw,
+                             boolean scissor,
+                             boolean flatshade );
+
 
 #endif /* DRAW_PRIVATE_H */
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index 6d90a6c42fd..b853f3a89f8 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -37,6 +37,13 @@
 #include "util/u_math.h"
 #include "util/u_prim.h"
 
+
+DEBUG_GET_ONCE_BOOL_OPTION(draw_fse, "DRAW_FSE", FALSE)
+DEBUG_GET_ONCE_BOOL_OPTION(draw_no_fse, "DRAW_NO_FSE", FALSE)
+#ifdef HAVE_LLVM
+DEBUG_GET_ONCE_BOOL_OPTION(draw_use_llvm, "DRAW_USE_LLVM", TRUE)
+#endif
+
 static unsigned trim( unsigned count, unsigned first, unsigned incr )
 {
    if (count < first)
@@ -90,12 +97,16 @@ draw_pt_arrays(struct draw_context *draw,
       opt |= PT_SHADE;
    }
       
-   if (opt == 0) 
-      middle = draw->pt.middle.fetch_emit;
-   else if (opt == PT_SHADE && !draw->pt.no_fse)
-      middle = draw->pt.middle.fetch_shade_emit;
-   else
-      middle = draw->pt.middle.general;
+   if (draw->pt.middle.llvm) {
+      middle = draw->pt.middle.llvm;
+   } else {
+      if (opt == 0)
+         middle = draw->pt.middle.fetch_emit;
+      else if (opt == PT_SHADE && !draw->pt.no_fse)
+         middle = draw->pt.middle.fetch_shade_emit;
+      else
+         middle = draw->pt.middle.general;
+   }
 
 
    /* Pick the right frontend
@@ -111,6 +122,7 @@ draw_pt_arrays(struct draw_context *draw,
    frontend->run(frontend, 
                  draw_pt_elt_func(draw),
                  draw_pt_elt_ptr(draw, start),
+                 draw->pt.user.eltBias,
                  count);
 
    frontend->finish( frontend );
@@ -121,8 +133,8 @@ draw_pt_arrays(struct draw_context *draw,
 
 boolean draw_pt_init( struct draw_context *draw )
 {
-   draw->pt.test_fse = debug_get_bool_option("DRAW_FSE", FALSE);
-   draw->pt.no_fse = debug_get_bool_option("DRAW_NO_FSE", FALSE);
+   draw->pt.test_fse = debug_get_option_draw_fse();
+   draw->pt.no_fse = debug_get_option_draw_no_fse();
 
    draw->pt.front.vcache = draw_pt_vcache( draw );
    if (!draw->pt.front.vcache)
@@ -144,12 +156,22 @@ boolean draw_pt_init( struct draw_context *draw )
    if (!draw->pt.middle.general)
       return FALSE;
 
+#if HAVE_LLVM
+   if (debug_get_option_draw_use_llvm())
+      draw->pt.middle.llvm = draw_pt_fetch_pipeline_or_emit_llvm( draw );
+#endif
+
    return TRUE;
 }
 
 
 void draw_pt_destroy( struct draw_context *draw )
 {
+   if (draw->pt.middle.llvm) {
+      draw->pt.middle.llvm->destroy( draw->pt.middle.llvm );
+      draw->pt.middle.llvm = NULL;
+   }
+
    if (draw->pt.middle.general) {
       draw->pt.middle.general->destroy( draw->pt.middle.general );
       draw->pt.middle.general = NULL;
@@ -215,8 +237,11 @@ draw_print_arrays(struct draw_context *draw, uint prim, int start, uint count)
             break;
          default:
             assert(0);
+            return;
          }
-         debug_printf("Element[%u + %u] -> Vertex %u:\n", start, i, ii);
+         ii += draw->pt.user.eltBias;
+         debug_printf("Element[%u + %u] + %i -> Vertex %u:\n", start, i,
+                      draw->pt.user.eltBias, ii);
       }
       else {
          /* non-indexed arrays */
@@ -307,9 +332,8 @@ draw_arrays_instanced(struct draw_context *draw,
       tgsi_dump(draw->vs.vertex_shader->state.tokens, 0);
       debug_printf("Elements:\n");
       for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
-         debug_printf("  format=%s comps=%u\n",
-                      util_format_name(draw->pt.vertex_element[i].src_format),
-                      draw->pt.vertex_element[i].nr_components);
+         debug_printf("  format=%s\n",
+                      util_format_name(draw->pt.vertex_element[i].src_format));
       }
       debug_printf("Buffers:\n");
       for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index d5e0d92a605..3e3ea320cc0 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -67,6 +67,7 @@ struct draw_pt_front_end {
    void (*run)( struct draw_pt_front_end *,
                 pt_elt_func elt_func,
                 const void *elt_ptr,
+                int elt_bias,
                 unsigned count );
 
    void (*finish)( struct draw_pt_front_end * );
@@ -147,6 +148,7 @@ struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw);
 struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw );
 struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw );
 struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw);
+struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit_llvm(struct draw_context *draw);
 
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h
index 4ca5b520204..3c44f7c11ee 100644
--- a/src/gallium/auxiliary/draw/draw_pt_decompose.h
+++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h
@@ -105,40 +105,20 @@ static void FUNC( ARGS,
 
 
    case PIPE_PRIM_QUADS:
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 4) {
-            QUAD( (i + 1),
-                  (i + 2),
-                  (i + 3),
-                  (i + 0) );
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 4) {
-            QUAD( (i + 0),
-                  (i + 1),
-                  (i + 2),
-                  (i + 3));
-         }
+      for (i = 0; i+3 < count; i += 4) {
+         QUAD( (i + 0),
+               (i + 1),
+               (i + 2),
+               (i + 3));
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 2) {
-            QUAD( (i + 1),
-                  (i + 3),
-                  (i + 2),
-                  (i + 0) );
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 2) {
-            QUAD( (i + 2),
-                  (i + 0),
-                  (i + 1),
-                  (i + 3));
-         }
+      for (i = 0; i+3 < count; i += 2) {
+         QUAD( (i + 2),
+               (i + 0),
+               (i + 1),
+               (i + 3));
       }
       break;
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c
index ae357b51226..ad48fa39a4f 100644
--- a/src/gallium/auxiliary/draw/draw_pt_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_emit.c
@@ -86,40 +86,15 @@ void draw_pt_emit_prepare( struct pt_emit *emit,
       unsigned output_format;
       unsigned src_offset = (vinfo->attrib[i].src_index * 4 * sizeof(float) );
 
+      output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
+      emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
 
-         
-      switch (vinfo->attrib[i].emit) {
-      case EMIT_4F:
-	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
-	 emit_sz = 4 * sizeof(float);
-	 break;
-      case EMIT_3F:
-	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
-	 emit_sz = 3 * sizeof(float);
-	 break;
-      case EMIT_2F:
-	 output_format = PIPE_FORMAT_R32G32_FLOAT;
-	 emit_sz = 2 * sizeof(float);
-	 break;
-      case EMIT_1F:
-	 output_format = PIPE_FORMAT_R32_FLOAT;
-	 emit_sz = 1 * sizeof(float);
-	 break;
-      case EMIT_1F_PSIZE:
-	 output_format = PIPE_FORMAT_R32_FLOAT;
-	 emit_sz = 1 * sizeof(float);
+      /* doesn't handle EMIT_OMIT */
+      assert(emit_sz != 0);
+
+      if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
 	 src_buffer = 1;
 	 src_offset = 0;
-	 break;
-      case EMIT_4UB:
-	 output_format = PIPE_FORMAT_A8R8G8B8_UNORM;
-	 emit_sz = 4 * sizeof(ubyte);
-         break;
-      default:
-	 assert(0);
-	 output_format = PIPE_FORMAT_NONE;
-	 emit_sz = 0;
-	 break;
       }
 
       hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
@@ -196,12 +171,14 @@ void draw_pt_emit( struct pt_emit *emit,
    translate->set_buffer(translate, 
 			 0, 
 			 vertex_data,
-			 stride );
+			 stride,
+			 ~0);
 
    translate->set_buffer(translate, 
 			 1, 
 			 &draw->rasterizer->point_size,
-			 0);
+			 0,
+			 ~0);
 
    translate->run( translate,
 		   0, 
@@ -257,11 +234,11 @@ void draw_pt_emit_linear(struct pt_emit *emit,
       goto fail;
 
    translate->set_buffer(translate, 0,
-			 vertex_data, stride);
+			 vertex_data, stride, count - 1);
 
    translate->set_buffer(translate, 1,
 			 &draw->rasterizer->point_size,
-			 0);
+			 0, ~0);
 
    translate->run(translate,
                   0,
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
index 252be5053e4..a1347221b5d 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -149,7 +149,8 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch,
 	 fetch->translate->set_buffer(fetch->translate,
 				      draw->pt.nr_vertex_buffers,
 				      &vh,
-				      0);
+				      0,
+				      ~0);
       }
    }
 
@@ -172,7 +173,8 @@ void draw_pt_fetch_run( struct pt_fetch *fetch,
 			    i, 
 			    ((char *)draw->pt.user.vbuffer[i] + 
 			     draw->pt.vertex_buffer[i].buffer_offset),
-			    draw->pt.vertex_buffer[i].stride );
+			    draw->pt.vertex_buffer[i].stride,
+			    draw->pt.vertex_buffer[i].max_index);
    }
 
    translate->run_elts( translate,
@@ -198,7 +200,8 @@ void draw_pt_fetch_run_linear( struct pt_fetch *fetch,
 			    i,
 			    ((char *)draw->pt.user.vbuffer[i] +
 			     draw->pt.vertex_buffer[i].buffer_offset),
-			    draw->pt.vertex_buffer[i].stride );
+			    draw->pt.vertex_buffer[i].stride,
+			    draw->pt.vertex_buffer[i].max_index);
    }
 
    translate->run( translate,
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
index 2a604470e9a..d7735bf1ac9 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
@@ -129,41 +129,16 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
       unsigned input_offset = src->src_offset;
       unsigned output_format;
 
-      switch (vinfo->attrib[i].emit) {
-      case EMIT_4UB:
-	 output_format = PIPE_FORMAT_R8G8B8A8_UNORM;
-	 emit_sz = 4 * sizeof(unsigned char);
-         break;
-      case EMIT_4F:
-	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
-	 emit_sz = 4 * sizeof(float);
-         break;
-      case EMIT_3F:
-	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
-	 emit_sz = 3 * sizeof(float);
-         break;
-      case EMIT_2F:
-	 output_format = PIPE_FORMAT_R32G32_FLOAT;
-	 emit_sz = 2 * sizeof(float);
-         break;
-      case EMIT_1F:
-	 output_format = PIPE_FORMAT_R32_FLOAT;
-	 emit_sz = 1 * sizeof(float);
-         break;
-      case EMIT_1F_PSIZE:
+      output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
+      emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
+
+      if (vinfo->attrib[i].emit == EMIT_OMIT)
+	 continue;
+
+      if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
 	 input_format = PIPE_FORMAT_R32_FLOAT;
 	 input_buffer = draw->pt.nr_vertex_buffers;
 	 input_offset = 0;
-	 output_format = PIPE_FORMAT_R32_FLOAT;
-	 emit_sz = 1 * sizeof(float);
-         break;
-      case EMIT_OMIT:
-         continue;
-      default:
-         assert(0);
-	 output_format = PIPE_FORMAT_NONE;
-	 emit_sz = 0;
-	 continue;
       }
 
       key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
@@ -193,7 +168,8 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
       feme->translate->set_buffer(feme->translate, 
 				  draw->pt.nr_vertex_buffers, 
 				  &feme->point_size,
-				  0);
+				  0,
+				  ~0);
    }
    
    feme->point_size = draw->rasterizer->point_size;
@@ -203,7 +179,8 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
                                   i, 
                                   ((char *)draw->pt.user.vbuffer[i] + 
                                    draw->pt.vertex_buffer[i].buffer_offset),
-                                  draw->pt.vertex_buffer[i].stride );
+                                  draw->pt.vertex_buffer[i].stride,
+                                  draw->pt.vertex_buffer[i].max_index);
    }
 
    *max_vertices = (draw->render->max_vertex_buffer_bytes / 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
index 1aecb510777..cbb5b6c9605 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
@@ -130,31 +130,10 @@ static void fse_prepare( struct draw_pt_middle_end *middle,
       unsigned dst_offset = 0;
 
       for (i = 0; i < vinfo->num_attribs; i++) {
-         unsigned emit_sz = 0;
-
-         switch (vinfo->attrib[i].emit) {
-         case EMIT_4F:
-            emit_sz = 4 * sizeof(float);
-            break;
-         case EMIT_3F:
-            emit_sz = 3 * sizeof(float);
-            break;
-         case EMIT_2F:
-            emit_sz = 2 * sizeof(float);
-            break;
-         case EMIT_1F:
-            emit_sz = 1 * sizeof(float);
-            break;
-         case EMIT_1F_PSIZE:
-            emit_sz = 1 * sizeof(float);
-            break;
-         case EMIT_4UB:
-            emit_sz = 4 * sizeof(ubyte);
-            break;
-         default:
-            assert(0);
-            break;
-         }
+         unsigned emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
+
+         /* doesn't handle EMIT_OMIT */
+         assert(emit_sz != 0);
 
          /* The elements in the key correspond to vertex shader output
           * numbers, not to positions in the hw vertex description --
@@ -188,7 +167,8 @@ static void fse_prepare( struct draw_pt_middle_end *middle,
                                i, 
                                ((const ubyte *) draw->pt.user.vbuffer[i] + 
                                 draw->pt.vertex_buffer[i].buffer_offset),
-                              draw->pt.vertex_buffer[i].stride );
+                              draw->pt.vertex_buffer[i].stride,
+                              draw->pt.vertex_buffer[i].max_index );
    }
 
    *max_vertices = (draw->render->max_vertex_buffer_bytes / 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
new file mode 100644
index 00000000000..d2a492f2b4c
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
@@ -0,0 +1,416 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMWare, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+#include "draw/draw_vs.h"
+#include "draw/draw_gs.h"
+#include "draw/draw_llvm.h"
+
+
+struct llvm_middle_end {
+   struct draw_pt_middle_end base;
+   struct draw_context *draw;
+
+   struct pt_emit *emit;
+   struct pt_fetch *fetch;
+   struct pt_post_vs *post_vs;
+
+
+   unsigned vertex_data_offset;
+   unsigned vertex_size;
+   unsigned prim;
+   unsigned opt;
+
+   struct draw_llvm *llvm;
+   struct draw_llvm_variant *variants;
+   struct draw_llvm_variant *current_variant;
+   int nr_variants;
+};
+
+
+static void
+llvm_middle_end_prepare( struct draw_pt_middle_end *middle,
+                         unsigned prim,
+                         unsigned opt,
+                         unsigned *max_vertices )
+{
+   struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+   struct draw_geometry_shader *gs = draw->gs.geometry_shader;
+   struct draw_llvm_variant_key key;
+   struct draw_llvm_variant *variant = NULL;
+   unsigned i;
+   unsigned instance_id_index = ~0;
+
+   /* Add one to num_outputs because the pipeline occasionally tags on
+    * an additional texcoord, eg for AA lines.
+    */
+   unsigned nr = MAX2( vs->info.num_inputs,
+		       vs->info.num_outputs + 1 );
+
+   /* Scan for instanceID system value.
+    */
+   for (i = 0; i < vs->info.num_inputs; i++) {
+      if (vs->info.input_semantic_name[i] == TGSI_SEMANTIC_INSTANCEID) {
+         instance_id_index = i;
+         break;
+      }
+   }
+
+   fpme->prim = prim;
+   fpme->opt = opt;
+
+   /* Always leave room for the vertex header whether we need it or
+    * not.  It's hard to get rid of it in particular because of the
+    * viewport code in draw_pt_post_vs.c.
+    */
+   fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float);
+
+
+
+   draw_pt_fetch_prepare( fpme->fetch,
+                          vs->info.num_inputs,
+                          fpme->vertex_size,
+                          instance_id_index );
+   if (opt & PT_SHADE) {
+      vs->prepare(vs, draw);
+      draw_geometry_shader_prepare(gs, draw);
+   }
+
+
+   /* XXX: it's not really gl rasterization rules we care about here,
+    * but gl vs dx9 clip spaces.
+    */
+   draw_pt_post_vs_prepare( fpme->post_vs,
+			    (boolean)draw->bypass_clipping,
+			    (boolean)(draw->identity_viewport),
+			    (boolean)draw->rasterizer->gl_rasterization_rules,
+			    (draw->vs.edgeflag_output ? true : false) );
+
+   if (!(opt & PT_PIPELINE)) {
+      draw_pt_emit_prepare( fpme->emit,
+			    prim,
+                            max_vertices );
+
+      *max_vertices = MAX2( *max_vertices,
+                            DRAW_PIPE_MAX_VERTICES );
+   }
+   else {
+      *max_vertices = DRAW_PIPE_MAX_VERTICES;
+   }
+
+   /* return even number */
+   *max_vertices = *max_vertices & ~1;
+
+   draw_llvm_make_variant_key(fpme->llvm, &key);
+
+   variant = fpme->variants;
+   while(variant) {
+      if(memcmp(&variant->key, &key, sizeof key) == 0)
+         break;
+
+      variant = variant->next;
+   }
+
+   if (!variant) {
+      variant = draw_llvm_prepare(fpme->llvm, nr);
+      variant->next = fpme->variants;
+      fpme->variants = variant;
+      ++fpme->nr_variants;
+   }
+   fpme->current_variant = variant;
+
+   /*XXX we only support one constant buffer */
+   fpme->llvm->jit_context.vs_constants =
+      draw->pt.user.vs_constants[0];
+   fpme->llvm->jit_context.gs_constants =
+      draw->pt.user.gs_constants[0];
+}
+
+
+
+static void llvm_middle_end_run( struct draw_pt_middle_end *middle,
+                                 const unsigned *fetch_elts,
+                                 unsigned fetch_count,
+                                 const ushort *draw_elts,
+                                 unsigned draw_count )
+{
+   struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   unsigned opt = fpme->opt;
+   unsigned alloc_count = align( fetch_count, 4 );
+
+   struct vertex_header *pipeline_verts =
+      (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count);
+
+   if (!pipeline_verts) {
+      /* Not much we can do here - just skip the rendering.
+       */
+      assert(0);
+      return;
+   }
+
+   fpme->current_variant->jit_func_elts( &fpme->llvm->jit_context,
+                                         pipeline_verts,
+                                         (const char **)draw->pt.user.vbuffer,
+                                         fetch_elts,
+                                         fetch_count,
+                                         fpme->vertex_size,
+                                         draw->pt.vertex_buffer );
+
+   if (draw_pt_post_vs_run( fpme->post_vs,
+			    pipeline_verts,
+			    fetch_count,
+			    fpme->vertex_size ))
+   {
+      opt |= PT_PIPELINE;
+   }
+
+   /* Do we need to run the pipeline?
+    */
+   if (opt & PT_PIPELINE) {
+      draw_pipeline_run( fpme->draw,
+                         fpme->prim,
+                         pipeline_verts,
+                         fetch_count,
+                         fpme->vertex_size,
+                         draw_elts,
+                         draw_count );
+   }
+   else {
+      draw_pt_emit( fpme->emit,
+		    (const float (*)[4])pipeline_verts->data,
+		    fetch_count,
+		    fpme->vertex_size,
+		    draw_elts,
+		    draw_count );
+   }
+
+
+   FREE(pipeline_verts);
+}
+
+
+static void llvm_middle_end_linear_run( struct draw_pt_middle_end *middle,
+                                       unsigned start,
+                                       unsigned count)
+{
+   struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   unsigned opt = fpme->opt;
+   unsigned alloc_count = align( count, 4 );
+
+   struct vertex_header *pipeline_verts =
+      (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count);
+
+   if (!pipeline_verts) {
+      /* Not much we can do here - just skip the rendering.
+       */
+      assert(0);
+      return;
+   }
+
+#if 0
+   debug_printf("#### Pipeline = %p (data = %p)\n",
+                pipeline_verts, pipeline_verts->data);
+#endif
+   fpme->current_variant->jit_func( &fpme->llvm->jit_context,
+                                    pipeline_verts,
+                                    (const char **)draw->pt.user.vbuffer,
+                                    start,
+                                    count,
+                                    fpme->vertex_size,
+                                    draw->pt.vertex_buffer );
+
+   if (draw_pt_post_vs_run( fpme->post_vs,
+			    pipeline_verts,
+			    count,
+			    fpme->vertex_size ))
+   {
+      opt |= PT_PIPELINE;
+   }
+
+   /* Do we need to run the pipeline?
+    */
+   if (opt & PT_PIPELINE) {
+      draw_pipeline_run_linear( fpme->draw,
+                                fpme->prim,
+                                pipeline_verts,
+                                count,
+                                fpme->vertex_size);
+   }
+   else {
+      draw_pt_emit_linear( fpme->emit,
+                           (const float (*)[4])pipeline_verts->data,
+                           fpme->vertex_size,
+                           count );
+   }
+
+   FREE(pipeline_verts);
+}
+
+
+
+static boolean
+llvm_middle_end_linear_run_elts( struct draw_pt_middle_end *middle,
+                                 unsigned start,
+                                 unsigned count,
+                                 const ushort *draw_elts,
+                                 unsigned draw_count )
+{
+   struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   unsigned opt = fpme->opt;
+   unsigned alloc_count = align( count, 4 );
+
+   struct vertex_header *pipeline_verts =
+      (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count);
+
+   if (!pipeline_verts)
+      return FALSE;
+
+   fpme->current_variant->jit_func( &fpme->llvm->jit_context,
+                                    pipeline_verts,
+                                    (const char **)draw->pt.user.vbuffer,
+                                    start,
+                                    count,
+                                    fpme->vertex_size,
+                                    draw->pt.vertex_buffer );
+
+   if (draw_pt_post_vs_run( fpme->post_vs,
+			    pipeline_verts,
+			    count,
+			    fpme->vertex_size ))
+   {
+      opt |= PT_PIPELINE;
+   }
+
+   /* Do we need to run the pipeline?
+    */
+   if (opt & PT_PIPELINE) {
+      draw_pipeline_run( fpme->draw,
+                         fpme->prim,
+                         pipeline_verts,
+                         count,
+                         fpme->vertex_size,
+                         draw_elts,
+                         draw_count );
+   }
+   else {
+      draw_pt_emit( fpme->emit,
+		    (const float (*)[4])pipeline_verts->data,
+		    count,
+		    fpme->vertex_size,
+		    draw_elts,
+		    draw_count );
+   }
+
+   FREE(pipeline_verts);
+   return TRUE;
+}
+
+
+
+static void llvm_middle_end_finish( struct draw_pt_middle_end *middle )
+{
+   /* nothing to do */
+}
+
+static void llvm_middle_end_destroy( struct draw_pt_middle_end *middle )
+{
+   struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
+
+   if (fpme->fetch)
+      draw_pt_fetch_destroy( fpme->fetch );
+
+   if (fpme->emit)
+      draw_pt_emit_destroy( fpme->emit );
+
+   if (fpme->post_vs)
+      draw_pt_post_vs_destroy( fpme->post_vs );
+
+   if (fpme->llvm)
+      draw_llvm_destroy( fpme->llvm );
+
+   FREE(middle);
+}
+
+
+struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit_llvm( struct draw_context *draw )
+{
+   struct llvm_middle_end *fpme = 0;
+
+   if (!draw->engine)
+      return NULL;
+
+   fpme = CALLOC_STRUCT( llvm_middle_end );
+   if (!fpme)
+      goto fail;
+
+   fpme->base.prepare         = llvm_middle_end_prepare;
+   fpme->base.run             = llvm_middle_end_run;
+   fpme->base.run_linear      = llvm_middle_end_linear_run;
+   fpme->base.run_linear_elts = llvm_middle_end_linear_run_elts;
+   fpme->base.finish          = llvm_middle_end_finish;
+   fpme->base.destroy         = llvm_middle_end_destroy;
+
+   fpme->draw = draw;
+
+   fpme->fetch = draw_pt_fetch_create( draw );
+   if (!fpme->fetch)
+      goto fail;
+
+   fpme->post_vs = draw_pt_post_vs_create( draw );
+   if (!fpme->post_vs)
+      goto fail;
+
+   fpme->emit = draw_pt_emit_create( draw );
+   if (!fpme->emit)
+      goto fail;
+
+   fpme->llvm = draw_llvm_create(draw);
+   if (!fpme->llvm)
+      goto fail;
+
+   fpme->variants = NULL;
+   fpme->current_variant = NULL;
+   fpme->nr_variants = 0;
+
+   return &fpme->base;
+
+ fail:
+   if (fpme)
+      llvm_middle_end_destroy( &fpme->base );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
index 9728d5c2bdf..5525dfc748d 100644
--- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c
+++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
@@ -108,6 +108,11 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs,
    for (j = 0; j < count; j++) {
       float *position = out->data[pos];
 
+#if 0
+      debug_printf("%d) io = %p, data = %p = [%f, %f, %f, %f]\n",
+                   j, out, position, position[0], position[1], position[2], position[3]);
+#endif
+
       out->clip[0] = position[0];
       out->clip[1] = position[1];
       out->clip[2] = position[2];
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
index f0aec5febab..a292346be95 100644
--- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
+++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
@@ -6,6 +6,7 @@ static unsigned trim( unsigned count, unsigned first, unsigned incr )
 static void FUNC(struct draw_pt_front_end *frontend,
                  pt_elt_func get_elt,
                  const void *elts,
+                 int elt_bias,
                  unsigned count)
 {
    struct varray_frontend *varray = (struct varray_frontend *)frontend;
@@ -14,6 +15,8 @@ static void FUNC(struct draw_pt_front_end *frontend,
    unsigned j;
    unsigned first, incr;
 
+   assert(elt_bias == 0);
+
    draw_pt_split_prim(varray->input_prim, &first, &incr);
    
    /* Sanitize primitive length:
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
index 757c4874545..37ffbac4f92 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -329,6 +329,7 @@ static INLINE void
 vcache_check_run( struct draw_pt_front_end *frontend, 
                   pt_elt_func get_elt,
                   const void *elts,
+                  int elt_bias,
                   unsigned draw_count )
 {
    struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; 
@@ -346,7 +347,7 @@ vcache_check_run( struct draw_pt_front_end *frontend,
                        vcache->fetch_max,
                        draw_count);
       
-   if (max_index >= DRAW_PIPE_MAX_VERTICES ||
+   if (elt_bias + max_index >= DRAW_PIPE_MAX_VERTICES ||
        fetch_count >= UNDEFINED_VERTEX_ID ||
        fetch_count > draw_count) {
       if (0) debug_printf("fail\n");
@@ -362,8 +363,11 @@ vcache_check_run( struct draw_pt_front_end *frontend,
    }
 
 
+   assert((elt_bias >= 0 && min_index + elt_bias >= min_index) ||
+          (elt_bias <  0 && min_index + elt_bias <  min_index));
+
    if (min_index == 0 &&
-       index_size == 2) 
+       index_size == 2)
    {
       transformed_elts = (const ushort *)elts;
    }
@@ -433,7 +437,7 @@ vcache_check_run( struct draw_pt_front_end *frontend,
 
    if (fetch_count < UNDEFINED_VERTEX_ID)
       ok = vcache->middle->run_linear_elts( vcache->middle,
-                                            min_index, /* start */
+                                            min_index + elt_bias, /* start */
                                             fetch_count,
                                             transformed_elts,
                                             draw_count );
@@ -447,7 +451,7 @@ vcache_check_run( struct draw_pt_front_end *frontend,
                 fetch_count, draw_count);
 
  fail:
-   vcache_run( frontend, get_elt, elts, draw_count );
+   vcache_run( frontend, get_elt, elts, elt_bias, draw_count );
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
index 62822a3d562..f7a63de3ba9 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
@@ -3,6 +3,7 @@
 static void FUNC( struct draw_pt_front_end *frontend, 
                   pt_elt_func get_elt,
                   const void *elts,
+                  int elt_bias,
                   unsigned count )
 {
    struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
@@ -20,7 +21,7 @@ static void FUNC( struct draw_pt_front_end *frontend,
    case PIPE_PRIM_POINTS:
       for (i = 0; i < count; i ++) {
 	 POINT( vcache,
-                get_elt(elts, i + 0) );
+                get_elt(elts, i + 0) + elt_bias );
       }
       break;
 
@@ -28,8 +29,8 @@ static void FUNC( struct draw_pt_front_end *frontend,
       for (i = 0; i+1 < count; i += 2) {
          LINE( vcache, 
                DRAW_PIPE_RESET_STIPPLE,
-               get_elt(elts, i + 0),
-               get_elt(elts, i + 1));
+               get_elt(elts, i + 0) + elt_bias,
+               get_elt(elts, i + 1) + elt_bias);
       }
       break;
 
@@ -40,14 +41,14 @@ static void FUNC( struct draw_pt_front_end *frontend,
          for (i = 1; i < count; i++, flags = 0) {
             LINE( vcache, 
                   flags,
-                  get_elt(elts, i - 1),
-                  get_elt(elts, i ));
+                  get_elt(elts, i - 1) + elt_bias,
+                  get_elt(elts, i ) + elt_bias);
          }
 
 	 LINE( vcache, 
                flags,
-               get_elt(elts, i - 1),
-               get_elt(elts, 0 ));
+               get_elt(elts, i - 1) + elt_bias,
+               get_elt(elts, 0 ) + elt_bias);
       }
       break;
 
@@ -56,8 +57,8 @@ static void FUNC( struct draw_pt_front_end *frontend,
       for (i = 1; i < count; i++, flags = 0) {
          LINE( vcache, 
                flags,
-               get_elt(elts, i - 1),
-               get_elt(elts, i ));
+               get_elt(elts, i - 1) + elt_bias,
+               get_elt(elts, i ) + elt_bias);
       }
       break;
 
@@ -65,9 +66,9 @@ static void FUNC( struct draw_pt_front_end *frontend,
       for (i = 0; i+2 < count; i += 3) {
          TRIANGLE( vcache,
                    DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                   get_elt(elts, i + 0),
-                   get_elt(elts, i + 1),
-                   get_elt(elts, i + 2 ));
+                   get_elt(elts, i + 0) + elt_bias,
+                   get_elt(elts, i + 1) + elt_bias,
+                   get_elt(elts, i + 2 ) + elt_bias);
       }
       break;
 
@@ -76,18 +77,18 @@ static void FUNC( struct draw_pt_front_end *frontend,
          for (i = 0; i+2 < count; i++) {
             TRIANGLE( vcache,
                       DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                      get_elt(elts, i + 0),
-                      get_elt(elts, i + 1 + (i&1)),
-                      get_elt(elts, i + 2 - (i&1)));
+                      get_elt(elts, i + 0) + elt_bias,
+                      get_elt(elts, i + 1 + (i&1)) + elt_bias,
+                      get_elt(elts, i + 2 - (i&1)) + elt_bias);
          }
       }
       else {
          for (i = 0; i+2 < count; i++) {
             TRIANGLE( vcache,
                       DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                      get_elt(elts, i + 0 + (i&1)),
-                      get_elt(elts, i + 1 - (i&1)),
-                      get_elt(elts, i + 2 ));
+                      get_elt(elts, i + 0 + (i&1)) + elt_bias,
+                      get_elt(elts, i + 1 - (i&1)) + elt_bias,
+                      get_elt(elts, i + 2 ) + elt_bias);
          }
       }
       break;
@@ -98,18 +99,18 @@ static void FUNC( struct draw_pt_front_end *frontend,
             for (i = 0; i+2 < count; i++) {
                TRIANGLE( vcache,
                          DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                         get_elt(elts, i + 1),
-                         get_elt(elts, i + 2),
-                         get_elt(elts, 0 ));
+                         get_elt(elts, i + 1) + elt_bias,
+                         get_elt(elts, i + 2) + elt_bias,
+                         get_elt(elts, 0 ) + elt_bias);
             }
          }
          else {
             for (i = 0; i+2 < count; i++) {
                TRIANGLE( vcache,
                          DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                         get_elt(elts, 0),
-                         get_elt(elts, i + 1),
-                         get_elt(elts, i + 2 ));
+                         get_elt(elts, 0) + elt_bias,
+                         get_elt(elts, i + 1) + elt_bias,
+                         get_elt(elts, i + 2 ) + elt_bias);
             }
          }
       }
@@ -118,39 +119,21 @@ static void FUNC( struct draw_pt_front_end *frontend,
 
    case PIPE_PRIM_QUADS:
       for (i = 0; i+3 < count; i += 4) {
-         if (flatfirst) {
-            QUAD( vcache,
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 2),
-                  get_elt(elts, i + 3) );
-         }
-         else {
-            QUAD( vcache,
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 2),
-                  get_elt(elts, i + 3) );
-         }
+         QUAD( vcache,
+               get_elt(elts, i + 0) + elt_bias,
+               get_elt(elts, i + 1) + elt_bias,
+               get_elt(elts, i + 2) + elt_bias,
+               get_elt(elts, i + 3) + elt_bias );
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 0; i+3 < count; i += 2) {
-         if (flatfirst) {
-            QUAD( vcache,
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 3),
-                  get_elt(elts, i + 2) );
-         }
-         else {
-            QUAD( vcache,
-                  get_elt(elts, i + 2),
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 3) );
-         }
+         QUAD( vcache,
+               get_elt(elts, i + 2) + elt_bias,
+               get_elt(elts, i + 0) + elt_bias,
+               get_elt(elts, i + 1) + elt_bias,
+               get_elt(elts, i + 3) + elt_bias );
       }
       break;
 
@@ -183,16 +166,16 @@ static void FUNC( struct draw_pt_front_end *frontend,
             if (flatfirst) {
                TRIANGLE( vcache,
                          flags,
-                         get_elt(elts, 0),
-                         get_elt(elts, i + 1),
-                         get_elt(elts, i + 2) );
+                         get_elt(elts, 0) + elt_bias,
+                         get_elt(elts, i + 1) + elt_bias,
+                         get_elt(elts, i + 2) + elt_bias );
             }
             else {
                TRIANGLE( vcache,
                          flags,
-                         get_elt(elts, i + 1),
-                         get_elt(elts, i + 2),
-                         get_elt(elts, 0));
+                         get_elt(elts, i + 1) + elt_bias,
+                         get_elt(elts, i + 2) + elt_bias,
+                         get_elt(elts, 0) + elt_bias);
             }
 	 }
       }
diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c
index 3214213e445..a4f5e882c0a 100644
--- a/src/gallium/auxiliary/draw/draw_vertex.c
+++ b/src/gallium/auxiliary/draw/draw_vertex.c
@@ -48,30 +48,12 @@ draw_compute_vertex_size(struct vertex_info *vinfo)
    uint i;
 
    vinfo->size = 0;
-   for (i = 0; i < vinfo->num_attribs; i++) {
-      switch (vinfo->attrib[i].emit) {
-      case EMIT_OMIT:
-         break;
-      case EMIT_4UB:
-         /* fall-through */
-      case EMIT_1F_PSIZE:
-         /* fall-through */
-      case EMIT_1F:
-         vinfo->size += 1;
-         break;
-      case EMIT_2F:
-         vinfo->size += 2;
-         break;
-      case EMIT_3F:
-         vinfo->size += 3;
-         break;
-      case EMIT_4F:
-         vinfo->size += 4;
-         break;
-      default:
-         assert(0);
-      }
-   }
+   for (i = 0; i < vinfo->num_attribs; i++)
+      vinfo->size += draw_translate_vinfo_size(vinfo->attrib[i].emit);
+
+   assert(vinfo->size % 4 == 0);
+   /* in dwords */
+   vinfo->size /= 4;
 }
 
 
@@ -120,6 +102,13 @@ draw_dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data)
          debug_printf("%u ", *data++);
          debug_printf("%u ", *data++);
          break;
+      case EMIT_4UB_BGRA:
+         debug_printf("EMIT_4UB_BGRA:\t");
+         debug_printf("%u ", *data++);
+         debug_printf("%u ", *data++);
+         debug_printf("%u ", *data++);
+         debug_printf("%u ", *data++);
+         break;
       default:
          assert(0);
       }
diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h
index 8c3c7befbc7..3af31ffe126 100644
--- a/src/gallium/auxiliary/draw/draw_vertex.h
+++ b/src/gallium/auxiliary/draw/draw_vertex.h
@@ -54,7 +54,8 @@ enum attrib_emit {
    EMIT_2F,
    EMIT_3F,
    EMIT_4F,
-   EMIT_4UB  /**< XXX may need variations for RGBA vs BGRA, etc */
+   EMIT_4UB, /**< is RGBA like the rest */
+   EMIT_4UB_BGRA
 };
 
 
@@ -141,9 +142,11 @@ void draw_dump_emitted_vertex(const struct vertex_info *vinfo,
                               const uint8_t *data);
 
 
-static INLINE unsigned draw_translate_vinfo_format(unsigned format )
+static INLINE enum pipe_format draw_translate_vinfo_format(enum attrib_emit emit)
 {
-   switch (format) {
+   switch (emit) {
+   case EMIT_OMIT:
+      return PIPE_FORMAT_NONE;
    case EMIT_1F:
    case EMIT_1F_PSIZE:
       return PIPE_FORMAT_R32_FLOAT;
@@ -155,10 +158,36 @@ static INLINE unsigned draw_translate_vinfo_format(unsigned format )
       return PIPE_FORMAT_R32G32B32A32_FLOAT;
    case EMIT_4UB:
       return PIPE_FORMAT_R8G8B8A8_UNORM;
+   case EMIT_4UB_BGRA:
+      return PIPE_FORMAT_B8G8R8A8_UNORM;
    default:
+      assert(!"unexpected format");
       return PIPE_FORMAT_NONE;
    }
 }
 
+static INLINE enum attrib_emit draw_translate_vinfo_size(enum attrib_emit emit)
+{
+   switch (emit) {
+   case EMIT_OMIT:
+      return 0;
+   case EMIT_1F:
+   case EMIT_1F_PSIZE:
+      return 1 * sizeof(float);
+   case EMIT_2F:
+      return 2 * sizeof(float);
+   case EMIT_3F:
+      return 3 * sizeof(float);
+   case EMIT_4F:
+      return 4 * sizeof(float);
+   case EMIT_4UB:
+      return 4 * sizeof(unsigned char);
+   case EMIT_4UB_BGRA:
+      return 4 * sizeof(unsigned char);
+   default:
+      assert(!"unexpected format");
+      return 0;
+   }
+}
 
 #endif /* DRAW_VERTEX_H */
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index 90858380221..c2832eefa2a 100644
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -46,7 +46,7 @@
 #include "tgsi/tgsi_dump.h"
 #include "tgsi/tgsi_exec.h"
 
-
+DEBUG_GET_ONCE_BOOL_OPTION(gallium_dump_vs, "GALLIUM_DUMP_VS", FALSE)
 
 void
 draw_vs_set_constants(struct draw_context *draw,
@@ -91,14 +91,11 @@ draw_create_vertex_shader(struct draw_context *draw,
       tgsi_dump(shader->tokens, 0);
    }
 
-   vs = draw_create_vs_llvm( draw, shader );
+   vs = draw_create_vs_sse( draw, shader );
    if (!vs) {
-      vs = draw_create_vs_sse( draw, shader );
+      vs = draw_create_vs_ppc( draw, shader );
       if (!vs) {
-         vs = draw_create_vs_ppc( draw, shader );
-         if (!vs) {
-            vs = draw_create_vs_exec( draw, shader );
-         }
+         vs = draw_create_vs_exec( draw, shader );
       }
    }
 
@@ -160,7 +157,7 @@ draw_delete_vertex_shader(struct draw_context *draw,
 boolean 
 draw_vs_init( struct draw_context *draw )
 {
-   draw->dump_vs = debug_get_bool_option("GALLIUM_DUMP_VS", FALSE);
+   draw->dump_vs = debug_get_option_gallium_dump_vs();
 
    draw->vs.machine = tgsi_exec_machine_create();
    if (!draw->vs.machine)
diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h
index d095c9bad1d..6c7e94db433 100644
--- a/src/gallium/auxiliary/draw/draw_vs.h
+++ b/src/gallium/auxiliary/draw/draw_vs.h
@@ -80,7 +80,8 @@ struct draw_vs_varient {
    void (*set_buffer)( struct draw_vs_varient *,
                       unsigned i,
                       const void *ptr,
-                      unsigned stride );
+                      unsigned stride,
+                      unsigned max_stride );
 
    void (PIPE_CDECL *run_linear)( struct draw_vs_varient *shader,
                                   unsigned start,
@@ -163,17 +164,14 @@ struct draw_vertex_shader *
 draw_create_vs_ppc(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);
-
 
 
 struct draw_vs_varient_key;
 struct draw_vertex_shader;
 
-struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs,
-                                                 const struct draw_vs_varient_key *key );
+struct draw_vs_varient *
+draw_vs_create_varient_aos_sse( struct draw_vertex_shader *vs,
+                                const struct draw_vs_varient_key *key );
 
 
 
@@ -191,8 +189,9 @@ struct translate *draw_vs_get_fetch( struct draw_context *draw,
 struct translate *draw_vs_get_emit( struct draw_context *draw,
                                     struct translate_key *key );
 
-struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
-                                                 const struct draw_vs_varient_key *key );
+struct draw_vs_varient *
+draw_vs_create_varient_generic( struct draw_vertex_shader *vs,
+                                const struct draw_vs_varient_key *key );
 
 
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c
index e7121f36541..19f49e34c8b 100644
--- a/src/gallium/auxiliary/draw/draw_vs_aos.c
+++ b/src/gallium/auxiliary/draw/draw_vs_aos.c
@@ -2089,13 +2089,21 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient,
 }
 
 
+/** cast wrapper */
+static INLINE struct draw_vs_varient_aos_sse *
+draw_vs_varient_aos_sse(struct draw_vs_varient *varient)
+{
+   return (struct draw_vs_varient_aos_sse *) varient;
+}
+
 
 static void vaos_set_buffer( struct draw_vs_varient *varient,
                              unsigned buf,
                              const void *ptr,
-                             unsigned stride )
+                             unsigned stride,
+                             unsigned max_stride)
 {
-   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+   struct draw_vs_varient_aos_sse *vaos = draw_vs_varient_aos_sse(varient);
 
    if (buf < vaos->nr_vb) {
       vaos->buffer[buf].base_ptr = (char *)ptr;
@@ -2112,7 +2120,7 @@ static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient,
                                       unsigned count,
                                       void *output_buffer )
 {
-   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+   struct draw_vs_varient_aos_sse *vaos = draw_vs_varient_aos_sse(varient);
    struct aos_machine *machine = vaos->draw->vs.aos_machine;
    unsigned i;
 
@@ -2136,7 +2144,7 @@ static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient,
                                         unsigned count,
                                         void *output_buffer )
 {
-   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+   struct draw_vs_varient_aos_sse *vaos = draw_vs_varient_aos_sse(varient);
    struct aos_machine *machine = vaos->draw->vs.aos_machine;
    unsigned i;
 
@@ -2165,7 +2173,7 @@ static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient,
 
 static void vaos_destroy( struct draw_vs_varient *varient )
 {
-   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+   struct draw_vs_varient_aos_sse *vaos = draw_vs_varient_aos_sse(varient);
 
    FREE( vaos->buffer );
 
@@ -2241,13 +2249,14 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs,
 }
 
 
-struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs,
-                                                 const struct draw_vs_varient_key *key )
+struct draw_vs_varient *
+draw_vs_create_varient_aos_sse( struct draw_vertex_shader *vs,
+                                const struct draw_vs_varient_key *key )
 {
    struct draw_vs_varient *varient = varient_aos_sse( vs, key );
 
    if (varient == NULL) {
-      varient = draw_vs_varient_generic( vs, key );
+      varient = draw_vs_create_varient_generic( vs, key );
    }
 
    return varient;
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c
index ece1ddde0cb..8f8bbe7cb88 100644
--- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c
+++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c
@@ -401,13 +401,11 @@ static boolean emit_output( struct aos_compilation *cp,
       emit_store_R32G32B32A32(cp, ptr, dataXMM);
       break;
    case EMIT_4UB:
-      if (1) {
-         emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W));
-         emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM);
-      }
-      else {
-         emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM);
-      }
+      emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM);
+      break;
+   case EMIT_4UB_BGRA:
+      emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W));
+      emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM);
       break;
    default:
       AOS_ERROR(cp, "unhandled output format");
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index 7deca2b69d9..bc34d390dae 100644
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -203,7 +203,7 @@ draw_create_vs_exec(struct draw_context *draw,
    vs->base.prepare = vs_exec_prepare;
    vs->base.run_linear = vs_exec_run_linear;
    vs->base.delete = vs_exec_delete;
-   vs->base.create_varient = draw_vs_varient_generic;
+   vs->base.create_varient = draw_vs_create_varient_generic;
    vs->machine = draw->vs.machine;
 
    return &vs->base;
diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c
deleted file mode 100644
index 5f7a645f5d8..00000000000
--- a/src/gallium/auxiliary/draw/draw_vs_llvm.c
+++ /dev/null
@@ -1,131 +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:
-  *   Zack Rusin
-  *   Keith Whitwell <[email protected]>
-  *   Brian Paul
-  */
-
-#include "util/u_memory.h"
-#include "pipe/p_shader_tokens.h"
-#include "draw_private.h"
-#include "draw_context.h"
-#include "draw_vs.h"
-
-#include "tgsi/tgsi_parse.h"
-
-#ifdef MESA_LLVM
-
-struct draw_llvm_vertex_shader {
-   struct draw_vertex_shader base;
-   struct tgsi_exec_machine *machine;
-};
-
-
-static void
-vs_llvm_prepare( struct draw_vertex_shader *base,
-		 struct draw_context *draw )
-{
-}
-
-
-static void
-vs_llvm_run_linear( struct draw_vertex_shader *base,
-		   const float (*input)[4],
-		   float (*output)[4],
-                   const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
-		   unsigned count,
-		   unsigned input_stride,
-		   unsigned output_stride )
-{
-   struct draw_llvm_vertex_shader *shader =
-      (struct draw_llvm_vertex_shader *)base;
-}
-
-
-
-static void
-vs_llvm_delete( struct draw_vertex_shader *base )
-{
-   struct draw_llvm_vertex_shader *shader =
-      (struct draw_llvm_vertex_shader *)base;
-
-   /* Do something to free compiled shader:
-    */
-
-   FREE( (void*) shader->base.state.tokens );
-   FREE( shader );
-}
-
-
-
-
-struct draw_vertex_shader *
-draw_create_vs_llvm(struct draw_context *draw,
-		    const struct pipe_shader_state *templ)
-{
-   struct draw_llvm_vertex_shader *vs;
-
-   vs = CALLOC_STRUCT( draw_llvm_vertex_shader );
-   if (vs == NULL)
-      return NULL;
-
-   /* we make a private copy of the tokens */
-   vs->base.state.tokens = tgsi_dup_tokens(templ->tokens);
-   if (!vs->base.state.tokens) {
-      FREE(vs);
-      return NULL;
-   }
-
-   tgsi_scan_shader(vs->base.state.tokens, &vs->base.info);
-
-   vs->base.draw = draw;
-   vs->base.prepare = vs_llvm_prepare;
-   vs->base.create_varient = draw_vs_varient_generic;
-   vs->base.run_linear = vs_llvm_run_linear;
-   vs->base.delete = vs_llvm_delete;
-   vs->machine = draw->vs.machine;
-
-   return &vs->base;
-}
-
-
-
-
-
-#else
-
-struct draw_vertex_shader *
-draw_create_vs_llvm(struct draw_context *draw,
-                          const struct pipe_shader_state *shader)
-{
-   return NULL;
-}
-
-#endif
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index 54e6423388f..14c95082a9d 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -165,9 +165,9 @@ draw_create_vs_sse(struct draw_context *draw,
 
    vs->base.draw = draw;
    if (1)
-      vs->base.create_varient = draw_vs_varient_aos_sse;
+      vs->base.create_varient = draw_vs_create_varient_aos_sse;
    else
-      vs->base.create_varient = draw_vs_varient_generic;
+      vs->base.create_varient = draw_vs_create_varient_generic;
    vs->base.prepare = vs_sse_prepare;
    vs->base.run_linear = vs_sse_run_linear;
    vs->base.delete = vs_sse_delete;
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index 5ed706cb4ff..6eb26927f27 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -66,14 +66,16 @@ struct draw_vs_varient_generic {
 static void vsvg_set_buffer( struct draw_vs_varient *varient,
                              unsigned buffer,
                              const void *ptr,
-                             unsigned stride )
+                             unsigned stride,
+                             unsigned max_index )
 {
    struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
 
    vsvg->fetch->set_buffer(vsvg->fetch, 
                            buffer, 
                            ptr, 
-                           stride);
+                           stride,
+                           max_index );
 }
 
 
@@ -172,12 +174,14 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
    vsvg->emit->set_buffer( vsvg->emit,
                            0, 
                            temp_buffer,
-                           temp_vertex_stride );
+                           temp_vertex_stride,
+                           ~0 );
 
    vsvg->emit->set_buffer( vsvg->emit, 
                            1,
                            &vsvg->draw->rasterizer->point_size,
-                           0);
+                           0,
+                           ~0 );
 
    vsvg->emit->run( vsvg->emit,
                     0, count,
@@ -232,12 +236,14 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    vsvg->emit->set_buffer( vsvg->emit,
                            0, 
                            temp_buffer,
-                           temp_vertex_stride );
+                           temp_vertex_stride,
+                           ~0 );
    
    vsvg->emit->set_buffer( vsvg->emit, 
                            1,
                            &vsvg->draw->rasterizer->point_size,
-                           0);
+                           0,
+                           ~0 );
    
    vsvg->emit->run( vsvg->emit,
                     0, count,
@@ -257,8 +263,9 @@ static void vsvg_destroy( struct draw_vs_varient *varient )
 }
 
 
-struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
-                                                 const struct draw_vs_varient_key *key )
+struct draw_vs_varient *
+draw_vs_create_varient_generic( struct draw_vertex_shader *vs,
+                                const struct draw_vs_varient_key *key )
 {
    unsigned i;
    struct translate_key fetch, emit;