Merge branch 'gallium-vertexelementcso'

Conflicts:
	src/gallium/auxiliary/cso_cache/cso_context.c
	src/gallium/auxiliary/util/u_blit.c
	src/gallium/auxiliary/util/u_gen_mipmap.c
	src/gallium/drivers/i915/i915_context.h
	src/gallium/drivers/i965/brw_draw_upload.c
	src/gallium/drivers/r300/r300_blit.c
	src/gallium/drivers/r300/r300_context.h
	src/gallium/drivers/r300/r300_render.c
	src/gallium/drivers/r300/r300_state_derived.c
	src/mesa/state_tracker/st_cb_bitmap.c
	src/mesa/state_tracker/st_cb_clear.c

4 files changed, 251 insertions, 213 deletions

diff --git a/src/gallium/drivers/i965/brw_context.h b/src/gallium/drivers/i965/brw_context.h
index 12cfa7b049c..f5b1a06576b 100644
--- a/src/gallium/drivers/i965/brw_context.h
+++ b/src/gallium/drivers/i965/brw_context.h
@@ -351,7 +351,7 @@ struct brw_vs_prog_data {
 
 /* Size == 0 if output either not written, or always [0,0,0,1]
  */
-struct brw_vs_ouput_sizes {
+struct brw_vs_output_sizes {
    GLubyte output_size[PIPE_MAX_SHADER_OUTPUTS];
 };
 
@@ -546,14 +546,13 @@ struct brw_context
       const struct brw_blend_state *blend;
       const struct brw_rasterizer_state *rast;
       const struct brw_depth_stencil_state *zstencil;
+      const struct brw_vertex_element_packet *velems;
 
       const struct brw_sampler *sampler[PIPE_MAX_SAMPLERS];
       unsigned num_samplers;
 
       struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
       struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
-      struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
-      unsigned num_vertex_elements;
       unsigned num_textures;
       unsigned num_vertex_buffers;
 
diff --git a/src/gallium/drivers/i965/brw_draw_upload.c b/src/gallium/drivers/i965/brw_draw_upload.c
index 9f136eec71c..0820ba20a08 100644
--- a/src/gallium/drivers/i965/brw_draw_upload.c
+++ b/src/gallium/drivers/i965/brw_draw_upload.c
@@ -42,141 +42,6 @@
 
 
 
-static unsigned brw_translate_surface_format( unsigned id )
-{
-   switch (id) {
-   case PIPE_FORMAT_R64_FLOAT:
-      return BRW_SURFACEFORMAT_R64_FLOAT;
-   case PIPE_FORMAT_R64G64_FLOAT:
-      return BRW_SURFACEFORMAT_R64G64_FLOAT;
-   case PIPE_FORMAT_R64G64B64_FLOAT:
-      return BRW_SURFACEFORMAT_R64G64B64_FLOAT;
-   case PIPE_FORMAT_R64G64B64A64_FLOAT:
-      return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT;
-
-   case PIPE_FORMAT_R32_FLOAT:
-      return BRW_SURFACEFORMAT_R32_FLOAT;
-   case PIPE_FORMAT_R32G32_FLOAT:
-      return BRW_SURFACEFORMAT_R32G32_FLOAT;
-   case PIPE_FORMAT_R32G32B32_FLOAT:
-      return BRW_SURFACEFORMAT_R32G32B32_FLOAT;
-   case PIPE_FORMAT_R32G32B32A32_FLOAT:
-      return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
-
-   case PIPE_FORMAT_R32_UNORM:
-      return BRW_SURFACEFORMAT_R32_UNORM;
-   case PIPE_FORMAT_R32G32_UNORM:
-      return BRW_SURFACEFORMAT_R32G32_UNORM;
-   case PIPE_FORMAT_R32G32B32_UNORM:
-      return BRW_SURFACEFORMAT_R32G32B32_UNORM;
-   case PIPE_FORMAT_R32G32B32A32_UNORM:
-      return BRW_SURFACEFORMAT_R32G32B32A32_UNORM;
-
-   case PIPE_FORMAT_R32_USCALED:
-      return BRW_SURFACEFORMAT_R32_USCALED;
-   case PIPE_FORMAT_R32G32_USCALED:
-      return BRW_SURFACEFORMAT_R32G32_USCALED;
-   case PIPE_FORMAT_R32G32B32_USCALED:
-      return BRW_SURFACEFORMAT_R32G32B32_USCALED;
-   case PIPE_FORMAT_R32G32B32A32_USCALED:
-      return BRW_SURFACEFORMAT_R32G32B32A32_USCALED;
-
-   case PIPE_FORMAT_R32_SNORM:
-      return BRW_SURFACEFORMAT_R32_SNORM;
-   case PIPE_FORMAT_R32G32_SNORM:
-      return BRW_SURFACEFORMAT_R32G32_SNORM;
-   case PIPE_FORMAT_R32G32B32_SNORM:
-      return BRW_SURFACEFORMAT_R32G32B32_SNORM;
-   case PIPE_FORMAT_R32G32B32A32_SNORM:
-      return BRW_SURFACEFORMAT_R32G32B32A32_SNORM;
-
-   case PIPE_FORMAT_R32_SSCALED:
-      return BRW_SURFACEFORMAT_R32_SSCALED;
-   case PIPE_FORMAT_R32G32_SSCALED:
-      return BRW_SURFACEFORMAT_R32G32_SSCALED;
-   case PIPE_FORMAT_R32G32B32_SSCALED:
-      return BRW_SURFACEFORMAT_R32G32B32_SSCALED;
-   case PIPE_FORMAT_R32G32B32A32_SSCALED:
-      return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED;
-
-   case PIPE_FORMAT_R16_UNORM:
-      return BRW_SURFACEFORMAT_R16_UNORM;
-   case PIPE_FORMAT_R16G16_UNORM:
-      return BRW_SURFACEFORMAT_R16G16_UNORM;
-   case PIPE_FORMAT_R16G16B16_UNORM:
-      return BRW_SURFACEFORMAT_R16G16B16_UNORM;
-   case PIPE_FORMAT_R16G16B16A16_UNORM:
-      return BRW_SURFACEFORMAT_R16G16B16A16_UNORM;
-
-   case PIPE_FORMAT_R16_USCALED:
-      return BRW_SURFACEFORMAT_R16_USCALED;
-   case PIPE_FORMAT_R16G16_USCALED:
-      return BRW_SURFACEFORMAT_R16G16_USCALED;
-   case PIPE_FORMAT_R16G16B16_USCALED:
-      return BRW_SURFACEFORMAT_R16G16B16_USCALED;
-   case PIPE_FORMAT_R16G16B16A16_USCALED:
-      return BRW_SURFACEFORMAT_R16G16B16A16_USCALED;
-
-   case PIPE_FORMAT_R16_SNORM:
-      return BRW_SURFACEFORMAT_R16_SNORM;
-   case PIPE_FORMAT_R16G16_SNORM:
-      return BRW_SURFACEFORMAT_R16G16_SNORM;
-   case PIPE_FORMAT_R16G16B16_SNORM:
-      return BRW_SURFACEFORMAT_R16G16B16_SNORM;
-   case PIPE_FORMAT_R16G16B16A16_SNORM:
-      return BRW_SURFACEFORMAT_R16G16B16A16_SNORM;
-
-   case PIPE_FORMAT_R16_SSCALED:
-      return BRW_SURFACEFORMAT_R16_SSCALED;
-   case PIPE_FORMAT_R16G16_SSCALED:
-      return BRW_SURFACEFORMAT_R16G16_SSCALED;
-   case PIPE_FORMAT_R16G16B16_SSCALED:
-      return BRW_SURFACEFORMAT_R16G16B16_SSCALED;
-   case PIPE_FORMAT_R16G16B16A16_SSCALED:
-      return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED;
-
-   case PIPE_FORMAT_R8_UNORM:
-      return BRW_SURFACEFORMAT_R8_UNORM;
-   case PIPE_FORMAT_R8G8_UNORM:
-      return BRW_SURFACEFORMAT_R8G8_UNORM;
-   case PIPE_FORMAT_R8G8B8_UNORM:
-      return BRW_SURFACEFORMAT_R8G8B8_UNORM;
-   case PIPE_FORMAT_R8G8B8A8_UNORM:
-      return BRW_SURFACEFORMAT_R8G8B8A8_UNORM;
-
-   case PIPE_FORMAT_R8_USCALED:
-      return BRW_SURFACEFORMAT_R8_USCALED;
-   case PIPE_FORMAT_R8G8_USCALED:
-      return BRW_SURFACEFORMAT_R8G8_USCALED;
-   case PIPE_FORMAT_R8G8B8_USCALED:
-      return BRW_SURFACEFORMAT_R8G8B8_USCALED;
-   case PIPE_FORMAT_R8G8B8A8_USCALED:
-      return BRW_SURFACEFORMAT_R8G8B8A8_USCALED;
-
-   case PIPE_FORMAT_R8_SNORM:
-      return BRW_SURFACEFORMAT_R8_SNORM;
-   case PIPE_FORMAT_R8G8_SNORM:
-      return BRW_SURFACEFORMAT_R8G8_SNORM;
-   case PIPE_FORMAT_R8G8B8_SNORM:
-      return BRW_SURFACEFORMAT_R8G8B8_SNORM;
-   case PIPE_FORMAT_R8G8B8A8_SNORM:
-      return BRW_SURFACEFORMAT_R8G8B8A8_SNORM;
-
-   case PIPE_FORMAT_R8_SSCALED:
-      return BRW_SURFACEFORMAT_R8_SSCALED;
-   case PIPE_FORMAT_R8G8_SSCALED:
-      return BRW_SURFACEFORMAT_R8G8_SSCALED;
-   case PIPE_FORMAT_R8G8B8_SSCALED:
-      return BRW_SURFACEFORMAT_R8G8B8_SSCALED;
-   case PIPE_FORMAT_R8G8B8A8_SSCALED:
-      return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED;
-
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
 static unsigned get_index_type(int type)
 {
    switch (type) {
@@ -315,75 +180,16 @@ static int brw_emit_vertex_buffers( struct brw_context *brw )
 
 
 
-
 static int brw_emit_vertex_elements(struct brw_context *brw)
 {
-   GLuint nr = brw->curr.num_vertex_elements;
-   GLuint i;
+   const struct brw_vertex_element_packet *brw_velems = brw->curr.velems;
+   unsigned size = brw_velems->header.length + 2;
 
+   /* why is this here */
    brw_emit_query_begin(brw);
 
-   /* If the VS doesn't read any inputs (calculating vertex position from
-    * a state variable for some reason, for example), emit a single pad
-    * VERTEX_ELEMENT struct and bail.
-    *
-    * The stale VB state stays in place, but they don't do anything unless
-    * a VE loads from them.
-    */
-   if (nr == 0) {
-      BEGIN_BATCH(3, IGNORE_CLIPRECTS);
-      OUT_BATCH((CMD_VERTEX_ELEMENT << 16) | 1);
-      OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT) |
-		BRW_VE0_VALID |
-		(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
-		(0 << BRW_VE0_SRC_OFFSET_SHIFT));
-      OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
-		(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
-		(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
-		(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
-      ADVANCE_BATCH();
-      return 0;
-   }
-
-   /* Now emit vertex element (VEP) state packets.
-    *
-    */
-   BEGIN_BATCH(1 + nr * 2, IGNORE_CLIPRECTS);
-   OUT_BATCH((CMD_VERTEX_ELEMENT << 16) | ((1 + nr * 2) - 2));
-   for (i = 0; i < nr; i++) {
-      const struct pipe_vertex_element *input = &brw->curr.vertex_element[i];
-      uint32_t format = brw_translate_surface_format( input->src_format );
-      uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
-      uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
-      uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
-      uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
-
-      switch (input->nr_components) {
-      case 0: comp0 = BRW_VE1_COMPONENT_STORE_0; /* fallthrough */
-      case 1: comp1 = BRW_VE1_COMPONENT_STORE_0; /* fallthrough */
-      case 2: comp2 = BRW_VE1_COMPONENT_STORE_0; /* fallthrough */
-      case 3: comp3 = BRW_VE1_COMPONENT_STORE_1_FLT;
-	 break;
-      }
-
-      OUT_BATCH((input->vertex_buffer_index << BRW_VE0_INDEX_SHIFT) |
-		BRW_VE0_VALID |
-		(format << BRW_VE0_FORMAT_SHIFT) |
-		(input->src_offset << BRW_VE0_SRC_OFFSET_SHIFT));
+   brw_batchbuffer_data(brw->batch, brw_velems, size * 4, IGNORE_CLIPRECTS);
 
-      if (BRW_IS_IGDNG(brw))
-          OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
-                    (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
-                    (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
-                    (comp3 << BRW_VE1_COMPONENT_3_SHIFT));
-      else
-          OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
-                    (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
-                    (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
-                    (comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
-                    ((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
-   }
-   ADVANCE_BATCH();
    return 0;
 }
 
@@ -396,10 +202,11 @@ static int brw_emit_vertices( struct brw_context *brw )
    if (ret)
       return ret;
 
+   /* XXX should separate this? */
    ret = brw_emit_vertex_elements( brw );
    if (ret)
       return ret;
-   
+
    return 0;
 }
 
@@ -407,7 +214,8 @@ static int brw_emit_vertices( struct brw_context *brw )
 const struct brw_tracked_state brw_vertices = {
    .dirty = {
       .mesa = (PIPE_NEW_INDEX_RANGE |
-               PIPE_NEW_VERTEX_BUFFER),
+               PIPE_NEW_VERTEX_BUFFER |
+               PIPE_NEW_VERTEX_ELEMENT),
       .brw = BRW_NEW_BATCH,
       .cache = 0,
    },
diff --git a/src/gallium/drivers/i965/brw_pipe_vertex.c b/src/gallium/drivers/i965/brw_pipe_vertex.c
index e3c48e31493..d6a840857ec 100644
--- a/src/gallium/drivers/i965/brw_pipe_vertex.c
+++ b/src/gallium/drivers/i965/brw_pipe_vertex.c
@@ -1,22 +1,251 @@
 #include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_structs.h"
 
+#include "util/u_memory.h"
+#include "util/u_format.h"
 
-static void brw_set_vertex_elements( struct pipe_context *pipe,
-				     unsigned count,
-				     const struct pipe_vertex_element *elements )
+
+static unsigned brw_translate_surface_format( unsigned id )
+{
+   switch (id) {
+   case PIPE_FORMAT_R64_FLOAT:
+      return BRW_SURFACEFORMAT_R64_FLOAT;
+   case PIPE_FORMAT_R64G64_FLOAT:
+      return BRW_SURFACEFORMAT_R64G64_FLOAT;
+   case PIPE_FORMAT_R64G64B64_FLOAT:
+      return BRW_SURFACEFORMAT_R64G64B64_FLOAT;
+   case PIPE_FORMAT_R64G64B64A64_FLOAT:
+      return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      return BRW_SURFACEFORMAT_R32_FLOAT;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      return BRW_SURFACEFORMAT_R32G32_FLOAT;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      return BRW_SURFACEFORMAT_R32G32B32_FLOAT;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
+
+   case PIPE_FORMAT_R32_UNORM:
+      return BRW_SURFACEFORMAT_R32_UNORM;
+   case PIPE_FORMAT_R32G32_UNORM:
+      return BRW_SURFACEFORMAT_R32G32_UNORM;
+   case PIPE_FORMAT_R32G32B32_UNORM:
+      return BRW_SURFACEFORMAT_R32G32B32_UNORM;
+   case PIPE_FORMAT_R32G32B32A32_UNORM:
+      return BRW_SURFACEFORMAT_R32G32B32A32_UNORM;
+
+   case PIPE_FORMAT_R32_USCALED:
+      return BRW_SURFACEFORMAT_R32_USCALED;
+   case PIPE_FORMAT_R32G32_USCALED:
+      return BRW_SURFACEFORMAT_R32G32_USCALED;
+   case PIPE_FORMAT_R32G32B32_USCALED:
+      return BRW_SURFACEFORMAT_R32G32B32_USCALED;
+   case PIPE_FORMAT_R32G32B32A32_USCALED:
+      return BRW_SURFACEFORMAT_R32G32B32A32_USCALED;
+
+   case PIPE_FORMAT_R32_SNORM:
+      return BRW_SURFACEFORMAT_R32_SNORM;
+   case PIPE_FORMAT_R32G32_SNORM:
+      return BRW_SURFACEFORMAT_R32G32_SNORM;
+   case PIPE_FORMAT_R32G32B32_SNORM:
+      return BRW_SURFACEFORMAT_R32G32B32_SNORM;
+   case PIPE_FORMAT_R32G32B32A32_SNORM:
+      return BRW_SURFACEFORMAT_R32G32B32A32_SNORM;
+
+   case PIPE_FORMAT_R32_SSCALED:
+      return BRW_SURFACEFORMAT_R32_SSCALED;
+   case PIPE_FORMAT_R32G32_SSCALED:
+      return BRW_SURFACEFORMAT_R32G32_SSCALED;
+   case PIPE_FORMAT_R32G32B32_SSCALED:
+      return BRW_SURFACEFORMAT_R32G32B32_SSCALED;
+   case PIPE_FORMAT_R32G32B32A32_SSCALED:
+      return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED;
+
+   case PIPE_FORMAT_R16_UNORM:
+      return BRW_SURFACEFORMAT_R16_UNORM;
+   case PIPE_FORMAT_R16G16_UNORM:
+      return BRW_SURFACEFORMAT_R16G16_UNORM;
+   case PIPE_FORMAT_R16G16B16_UNORM:
+      return BRW_SURFACEFORMAT_R16G16B16_UNORM;
+   case PIPE_FORMAT_R16G16B16A16_UNORM:
+      return BRW_SURFACEFORMAT_R16G16B16A16_UNORM;
+
+   case PIPE_FORMAT_R16_USCALED:
+      return BRW_SURFACEFORMAT_R16_USCALED;
+   case PIPE_FORMAT_R16G16_USCALED:
+      return BRW_SURFACEFORMAT_R16G16_USCALED;
+   case PIPE_FORMAT_R16G16B16_USCALED:
+      return BRW_SURFACEFORMAT_R16G16B16_USCALED;
+   case PIPE_FORMAT_R16G16B16A16_USCALED:
+      return BRW_SURFACEFORMAT_R16G16B16A16_USCALED;
+
+   case PIPE_FORMAT_R16_SNORM:
+      return BRW_SURFACEFORMAT_R16_SNORM;
+   case PIPE_FORMAT_R16G16_SNORM:
+      return BRW_SURFACEFORMAT_R16G16_SNORM;
+   case PIPE_FORMAT_R16G16B16_SNORM:
+      return BRW_SURFACEFORMAT_R16G16B16_SNORM;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      return BRW_SURFACEFORMAT_R16G16B16A16_SNORM;
+
+   case PIPE_FORMAT_R16_SSCALED:
+      return BRW_SURFACEFORMAT_R16_SSCALED;
+   case PIPE_FORMAT_R16G16_SSCALED:
+      return BRW_SURFACEFORMAT_R16G16_SSCALED;
+   case PIPE_FORMAT_R16G16B16_SSCALED:
+      return BRW_SURFACEFORMAT_R16G16B16_SSCALED;
+   case PIPE_FORMAT_R16G16B16A16_SSCALED:
+      return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED;
+
+   case PIPE_FORMAT_R8_UNORM:
+      return BRW_SURFACEFORMAT_R8_UNORM;
+   case PIPE_FORMAT_R8G8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8_UNORM;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8B8_UNORM;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R8_USCALED:
+      return BRW_SURFACEFORMAT_R8_USCALED;
+   case PIPE_FORMAT_R8G8_USCALED:
+      return BRW_SURFACEFORMAT_R8G8_USCALED;
+   case PIPE_FORMAT_R8G8B8_USCALED:
+      return BRW_SURFACEFORMAT_R8G8B8_USCALED;
+   case PIPE_FORMAT_R8G8B8A8_USCALED:
+      return BRW_SURFACEFORMAT_R8G8B8A8_USCALED;
+
+   case PIPE_FORMAT_R8_SNORM:
+      return BRW_SURFACEFORMAT_R8_SNORM;
+   case PIPE_FORMAT_R8G8_SNORM:
+      return BRW_SURFACEFORMAT_R8G8_SNORM;
+   case PIPE_FORMAT_R8G8B8_SNORM:
+      return BRW_SURFACEFORMAT_R8G8B8_SNORM;
+   case PIPE_FORMAT_R8G8B8A8_SNORM:
+      return BRW_SURFACEFORMAT_R8G8B8A8_SNORM;
+
+   case PIPE_FORMAT_R8_SSCALED:
+      return BRW_SURFACEFORMAT_R8_SSCALED;
+   case PIPE_FORMAT_R8G8_SSCALED:
+      return BRW_SURFACEFORMAT_R8G8_SSCALED;
+   case PIPE_FORMAT_R8G8B8_SSCALED:
+      return BRW_SURFACEFORMAT_R8G8B8_SSCALED;
+   case PIPE_FORMAT_R8G8B8A8_SSCALED:
+      return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED;
+
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static void brw_translate_vertex_elements(struct brw_context *brw,
+                                          struct brw_vertex_element_packet *brw_velems,
+                                          const struct pipe_vertex_element *attribs,
+                                          unsigned count)
+{
+   unsigned i;
+
+   /* If the VS doesn't read any inputs (calculating vertex position from
+    * a state variable for some reason, for example), emit a single pad
+    * VERTEX_ELEMENT struct and bail.
+    *
+    * The stale VB state stays in place, but they don't do anything unless
+    * a VE loads from them.
+    */
+   brw_velems->header.opcode = CMD_VERTEX_ELEMENT;
+
+   if (count == 0) {
+      brw_velems->header.length = 1;
+      brw_velems->ve[0].ve0.src_offset = 0;
+      brw_velems->ve[0].ve0.src_format = BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
+      brw_velems->ve[0].ve0.valid = 1;
+      brw_velems->ve[0].ve0.vertex_buffer_index = 0;
+      brw_velems->ve[0].ve1.dst_offset = 0;
+      brw_velems->ve[0].ve1.vfcomponent0 = BRW_VE1_COMPONENT_STORE_0;
+      brw_velems->ve[0].ve1.vfcomponent1 = BRW_VE1_COMPONENT_STORE_0;
+      brw_velems->ve[0].ve1.vfcomponent2 = BRW_VE1_COMPONENT_STORE_0;
+      brw_velems->ve[0].ve1.vfcomponent3 = BRW_VE1_COMPONENT_STORE_1_FLT;
+      return;
+   }
+
+
+   /* Now emit vertex element (VEP) state packets.
+    *
+    */
+   brw_velems->header.length = (1 + count * 2) - 2;
+   for (i = 0; i < count; i++) {
+      const struct pipe_vertex_element *input = &attribs[i];
+      unsigned nr_components = util_format_get_nr_components(input->src_format);
+
+      uint32_t format = brw_translate_surface_format( input->src_format );
+      uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
+      uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
+      uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
+      uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
+
+      switch (nr_components) {
+      case 0: comp0 = BRW_VE1_COMPONENT_STORE_0; /* fallthrough */
+      case 1: comp1 = BRW_VE1_COMPONENT_STORE_0; /* fallthrough */
+      case 2: comp2 = BRW_VE1_COMPONENT_STORE_0; /* fallthrough */
+      case 3: comp3 = BRW_VE1_COMPONENT_STORE_1_FLT;
+         break;
+      }
+
+      brw_velems->ve[i].ve0.src_offset = input->src_offset;
+      brw_velems->ve[i].ve0.src_format = format;
+      brw_velems->ve[i].ve0.valid = 1;
+      brw_velems->ve[i].ve0.vertex_buffer_index = input->vertex_buffer_index;
+      brw_velems->ve[i].ve1.vfcomponent0 = comp0;
+      brw_velems->ve[i].ve1.vfcomponent1 = comp1;
+      brw_velems->ve[i].ve1.vfcomponent2 = comp2;
+      brw_velems->ve[i].ve1.vfcomponent3 = comp3;
+
+      if (BRW_IS_IGDNG(brw))
+         brw_velems->ve[i].ve1.dst_offset = 0;
+      else
+         brw_velems->ve[i].ve1.dst_offset = i * 4;
+   }
+}
+
+static void* brw_create_vertex_elements_state( struct pipe_context *pipe,
+                                               unsigned count,
+                                               const struct pipe_vertex_element *attribs )
 {
+   /* note: for the brw_swtnl.c code (if ever we need draw fallback) we'd also need
+      to store the original data */
    struct brw_context *brw = brw_context(pipe);
+   struct brw_vertex_element_packet *velems;
+   assert(count <= BRW_VEP_MAX);
+   velems = (struct brw_vertex_element_packet *) MALLOC(sizeof(struct brw_vertex_element_packet));
+   if (velems) {
+      brw_translate_vertex_elements(brw, velems, attribs, count);
+   }
+   return velems;
+}
 
-   memcpy(brw->curr.vertex_element, elements, count * sizeof(elements[0]));
-   brw->curr.num_vertex_elements = count;
+static void brw_bind_vertex_elements_state(struct pipe_context *pipe,
+                                           void *velems)
+{
+   struct brw_context *brw = brw_context(pipe);
+   struct brw_vertex_element_packet *brw_velems = (struct brw_vertex_element_packet *) velems;
+
+   brw->curr.velems = brw_velems;
 
    brw->state.dirty.mesa |= PIPE_NEW_VERTEX_ELEMENT;
 }
 
+static void brw_delete_vertex_elements_state(struct pipe_context *pipe, void *velems)
+{
+   FREE( velems );
+}
+
 
 static void brw_set_vertex_buffers(struct pipe_context *pipe,
-				   unsigned count,
-				   const struct pipe_vertex_buffer *buffers)
+                                   unsigned count,
+                                   const struct pipe_vertex_buffer *buffers)
 {
    struct brw_context *brw = brw_context(pipe);
    unsigned i;
@@ -49,7 +278,9 @@ void
 brw_pipe_vertex_init( struct brw_context *brw )
 {
    brw->base.set_vertex_buffers = brw_set_vertex_buffers;
-   brw->base.set_vertex_elements = brw_set_vertex_elements;
+   brw->base.create_vertex_elements_state = brw_create_vertex_elements_state;
+   brw->base.bind_vertex_elements_state = brw_bind_vertex_elements_state;
+   brw->base.delete_vertex_elements_state = brw_delete_vertex_elements_state;
 }
 
 
diff --git a/src/gallium/drivers/i965/brw_structs.h b/src/gallium/drivers/i965/brw_structs.h
index bf10bc04de7..e97ddeb5e1c 100644
--- a/src/gallium/drivers/i965/brw_structs.h
+++ b/src/gallium/drivers/i965/brw_structs.h
@@ -28,7 +28,7 @@
   * Authors:
   *   Keith Whitwell <[email protected]>
   */
-        
+
 
 #ifndef BRW_STRUCTS_H
 #define BRW_STRUCTS_H
@@ -1149,7 +1149,7 @@ struct brw_vertex_element_state
       GLuint valid:1; 
       GLuint vertex_buffer_index:5; 
    } ve0;
-   
+
    struct
    {
       GLuint dst_offset:8;