i965g: still working on compilation

38 files changed, 789 insertions, 680 deletions

diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h
index 8a7ee0c7e4f..6754001e885 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h
@@ -61,6 +61,9 @@ struct tgsi_shader_info
    boolean uses_kill;  /**< KIL or KILP instruction used? */
    boolean uses_fogcoord; /**< fragment shader uses fog coord? */
    boolean uses_frontfacing; /**< fragment shader uses front/back-face flag? */
+
+   uint texture_max;
+   uint texture_mask;
 };
 
 
diff --git a/src/gallium/drivers/i965/Makefile b/src/gallium/drivers/i965/Makefile
index 40e8aa8786d..c3dbad72ae6 100644
--- a/src/gallium/drivers/i965/Makefile
+++ b/src/gallium/drivers/i965/Makefile
@@ -28,10 +28,7 @@ C_SOURCES = \
 	brw_pipe_blend.c \
 	brw_pipe_depth.c \
 	brw_pipe_fb.c \
-	brw_pipe_flush.c \
 	brw_pipe_query.c \
-	brw_pipe_shader.c \
-	brw_screen_surface.c \
 	brw_sf.c \
 	brw_sf_emit.c \
 	brw_sf_state.c \
@@ -40,8 +37,6 @@ C_SOURCES = \
 	brw_state_dump.c \
 	brw_state_upload.c \
 	brw_swtnl.c \
-	brw_tex.c \
-	brw_tex_layout.c \
 	brw_urb.c \
 	brw_util.c \
 	brw_vs.c \
@@ -60,8 +55,12 @@ C_SOURCES = \
 	brw_wm_sampler_state.c \
 	brw_wm_state.c \
 	brw_wm_surface_state.c \
+	brw_screen_surface.c \
+	brw_screen_texture.c \
 	brw_bo.c \
 	brw_batchbuffer.c \
+	brw_pipe_shader.c \
+	brw_pipe_flush.c \
 	intel_tex_layout.c 
 
 include ../../Makefile.template
diff --git a/src/gallium/drivers/i965/brw_batchbuffer.c b/src/gallium/drivers/i965/brw_batchbuffer.c
index 8bcac76ede1..45fbd59273a 100644
--- a/src/gallium/drivers/i965/brw_batchbuffer.c
+++ b/src/gallium/drivers/i965/brw_batchbuffer.c
@@ -105,13 +105,13 @@ _brw_batchbuffer_flush(struct brw_batchbuffer *batch, const char *file,
    }
 
 
-   if (INTEL_DEBUG & DEBUG_BATCH)
-      fprintf(stderr, "%s:%d: Batchbuffer flush with %db used\n", file, line,
+   if (BRW_DEBUG & DEBUG_BATCH)
+      debug_printf("%s:%d: Batchbuffer flush with %db used\n", file, line,
 	      used);
 
    /* Emit a flush if the bufmgr doesn't do it for us. */
    if (intel->always_flush_cache || !intel->ttm) {
-      *(GLuint *) (batch->ptr) = intel->vtbl.flush_cmd();
+      *(GLuint *) (batch->ptr) = ((CMD_MI_FLUSH << 16) | BRW_FLUSH_STATE_CACHE);
       batch->ptr += 4;
       used = batch->ptr - batch->map;
    }
@@ -136,15 +136,15 @@ _brw_batchbuffer_flush(struct brw_batchbuffer *batch, const char *file,
       
    batch->sws->bo_exec(batch->buf, used, NULL, 0, 0 );
       
-   if (INTEL_DEBUG & DEBUG_BATCH) {
+   if (BRW_DEBUG & DEBUG_BATCH) {
       dri_bo_map(batch->buf, GL_FALSE);
       intel_decode(batch->buf->virtual, used / 4, batch->buf->offset,
 		   brw->brw_screen->pci_id);
       dri_bo_unmap(batch->buf);
    }
 
-   if (INTEL_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "waiting for idle\n");
+   if (BRW_DEBUG & DEBUG_SYNC) {
+      debug_printf("waiting for idle\n");
       dri_bo_map(batch->buf, GL_TRUE);
       dri_bo_unmap(batch->buf);
    }
@@ -166,7 +166,7 @@ brw_batchbuffer_emit_reloc(struct brw_batchbuffer *batch,
    int ret;
 
    if (batch->ptr - batch->map > batch->buf->size)
-      _mesa_printf ("bad relocation ptr %p map %p offset %d size %d\n",
+      debug_printf ("bad relocation ptr %p map %p offset %d size %d\n",
 		    batch->ptr, batch->map, batch->ptr - batch->map, batch->buf->size);
 
    ret = batch->sws->bo_emit_reloc(batch->buf,
diff --git a/src/gallium/drivers/i965/brw_context.h b/src/gallium/drivers/i965/brw_context.h
index df43d8ba4d5..10c1cf6f33a 100644
--- a/src/gallium/drivers/i965/brw_context.h
+++ b/src/gallium/drivers/i965/brw_context.h
@@ -190,6 +190,8 @@ struct brw_fragment_shader {
 #define PIPE_NEW_FRAMEBUFFER_DIMENSIONS 0x10000
 #define PIPE_NEW_DEPTH_BUFFER           0x20000
 #define PIPE_NEW_COLOR_BUFFERS          0x40000
+#define PIPE_NEW_QUERY                  0x80000
+#define PIPE_NEW_SCISSOR                0x100000
 
 
 
@@ -204,7 +206,7 @@ struct brw_fragment_shader {
 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
 #define BRW_NEW_PSP                     0x800
 #define BRW_NEW_WM_SURFACES		0x1000
-#define BRW_NEW_FENCE                   0x2000
+#define BRW_NEW_xxx                     0x2000 /* was FENCE */
 #define BRW_NEW_INDICES			0x4000
 #define BRW_NEW_VERTICES		0x8000
 /**
@@ -373,6 +375,7 @@ struct brw_cache_item {
 
 struct brw_cache {
    struct brw_context *brw;
+   struct brw_winsys_screen *sws;
 
    struct brw_cache_item **items;
    GLuint size, n_items;
@@ -380,6 +383,7 @@ struct brw_cache {
    GLuint key_size[BRW_MAX_CACHE];		/* for fixed-size keys */
    GLuint aux_size[BRW_MAX_CACHE];
    char *name[BRW_MAX_CACHE];
+   
 
    /* Record of the last BOs chosen for each cache_id.  Used to set
     * brw->state.dirty.cache when a new cache item is chosen.
@@ -448,7 +452,7 @@ struct brw_query_object {
    int last_index;
 
    /* Total count of pixels from previous BOs */
-   unsigned int count;
+   uint64_t result;
 };
 
 
@@ -477,11 +481,18 @@ struct brw_context
       const struct brw_rasterizer_state *rast;
       const struct brw_depth_stencil_state *zstencil;
 
+      const struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+      const struct pipe_sampler *sampler[PIPE_MAX_SAMPLERS];
+      unsigned num_textures;
+      unsigned num_samplers;
+      
+
       struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
       struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
       unsigned num_vertex_elements;
       unsigned num_vertex_buffers;
 
+      struct pipe_scissor_state scissor;
       struct pipe_framebuffer_state fb;
       struct pipe_viewport_state vp;
       struct pipe_clip_state ucp;
@@ -492,6 +503,8 @@ struct brw_context
       struct brw_blend_constant_color bcc;
       struct brw_polygon_stipple bps;
 
+      
+
       /**
        * Index buffer for this draw_prims call.
        *
@@ -688,6 +701,7 @@ struct brw_context
       struct brw_winsys_buffer *bo;
       int index;
       GLboolean active;
+      int stats_wm;
    } query;
 
    struct {
diff --git a/src/gallium/drivers/i965/brw_eu_emit.c b/src/gallium/drivers/i965/brw_eu_emit.c
index f6b8843e01e..f7fa5203488 100644
--- a/src/gallium/drivers/i965/brw_eu_emit.c
+++ b/src/gallium/drivers/i965/brw_eu_emit.c
@@ -1262,7 +1262,7 @@ void brw_SAMPLE(struct brw_compile *p,
    GLboolean need_stall = 0;
    
    if (writemask == 0) {
-      /*_mesa_printf("%s: zero writemask??\n", __FUNCTION__); */
+      /*debug_printf("%s: zero writemask??\n", __FUNCTION__); */
       return;
    }
    
@@ -1294,7 +1294,7 @@ void brw_SAMPLE(struct brw_compile *p,
 
       if (newmask != writemask) {
 	 need_stall = 1;
-         /* _mesa_printf("need stall %x %x\n", newmask , writemask); */
+         /* debug_printf("need stall %x %x\n", newmask , writemask); */
       }
       else {
 	 struct brw_reg m1 = brw_message_reg(msg_reg_nr);
diff --git a/src/gallium/drivers/i965/brw_pipe_fb.c b/src/gallium/drivers/i965/brw_pipe_fb.c
index 63917172271..c65f9bc3747 100644
--- a/src/gallium/drivers/i965/brw_pipe_fb.c
+++ b/src/gallium/drivers/i965/brw_pipe_fb.c
@@ -53,7 +53,7 @@ static void brw_set_viewport_state( struct pipe_context *pipe,
 void brw_pipe_framebuffer_init( struct brw_context *brw )
 {
    brw->base.set_framebuffer_state = brw_set_framebuffer_state;
-   brw->base.set_framebuffer_state = brw_set_framebuffer_state;
+   brw->base.set_viewport_state = brw_set_viewport_state;
 }
 
 void brw_pipe_framebuffer_cleanup( struct brw_context *brw )
diff --git a/src/gallium/drivers/i965/brw_pipe_flush.c b/src/gallium/drivers/i965/brw_pipe_flush.c
index 65e71515178..fb4a784de9d 100644
--- a/src/gallium/drivers/i965/brw_pipe_flush.c
+++ b/src/gallium/drivers/i965/brw_pipe_flush.c
@@ -52,14 +52,7 @@ static void brw_note_fence( struct brw_context *brw, GLuint fence )
  */
 static GLuint brw_flush_cmd( void )
 {
-   struct brw_mi_flush flush;
-
-   return ;
-
-   flush.opcode = CMD_MI_FLUSH;
-   flush.pad = 0;
-   flush.flags = BRW_FLUSH_STATE_CACHE;
-   return *(GLuint *)&flush;
+   return ((CMD_MI_FLUSH << 16) | BRW_FLUSH_STATE_CACHE);
 }
 
 
diff --git a/src/gallium/drivers/i965/brw_pipe_query.c b/src/gallium/drivers/i965/brw_pipe_query.c
index a2da1373bfd..18a9b71af07 100644
--- a/src/gallium/drivers/i965/brw_pipe_query.c
+++ b/src/gallium/drivers/i965/brw_pipe_query.c
@@ -46,25 +46,38 @@
 #include "brw_reg.h"
 
 /** Waits on the query object's BO and totals the results for this query */
-static void
-brw_queryobj_get_results(struct brw_query_object *query)
+static boolean
+brw_query_get_result(struct pipe_context *pipe,
+		     struct pipe_query *q,
+		     boolean wait,
+		     uint64_t *result)
 {
-   int i;
-   uint64_t *results;
-
-   if (query->bo == NULL)
-      return;
+   struct brw_context *brw = brw_context(pipe);
+   struct brw_query_object *query = (struct brw_query_object *)q;
 
    /* Map and count the pixels from the current query BO */
-   dri_bo_map(query->bo, GL_FALSE);
-   results = query->bo->virtual;
-   for (i = query->first_index; i <= query->last_index; i++) {
-      query->Base.Result += results[i * 2 + 1] - results[i * 2];
+   if (query->bo) {
+      int i;
+      uint64_t *map;
+      
+      if (brw->sws->bo_is_busy(query->bo) && !wait)
+	 return FALSE;
+      
+      map = brw->sws->bo_map(query->bo, GL_FALSE);
+      if (map == NULL)
+	 return FALSE;
+      
+      for (i = query->first_index; i <= query->last_index; i++) {
+	 query->result += map[i * 2 + 1] - map[i * 2];
+      }
+
+      brw->sws->bo_unmap(query->bo);
+      brw->sws->bo_unreference(query->bo);
+      query->bo = NULL;
    }
-   dri_bo_unmap(query->bo);
 
-   brw->sws->bo_unreference(query->bo);
-   query->bo = NULL;
+   *result = query->result;
+   return TRUE;
 }
 
 static struct pipe_query *
@@ -72,12 +85,12 @@ brw_query_create(struct pipe_context *pipe, unsigned type )
 {
    struct brw_query_object *query;
 
-   switch (query->type) {
+   switch (type) {
    case PIPE_QUERY_OCCLUSION_COUNTER:
       query = CALLOC_STRUCT( brw_query_object );
       if (query == NULL)
 	 return NULL;
-      return &query->Base;
+      return (struct pipe_query *)query;
       
    default:
       return NULL;
@@ -87,6 +100,7 @@ brw_query_create(struct pipe_context *pipe, unsigned type )
 static void
 brw_query_destroy(struct pipe_context *pipe, struct pipe_query *q)
 {
+   struct brw_context *brw = brw_context(pipe);
    struct brw_query_object *query = (struct brw_query_object *)q;
 
    brw->sws->bo_unreference(query->bo);
@@ -94,24 +108,25 @@ brw_query_destroy(struct pipe_context *pipe, struct pipe_query *q)
 }
 
 static void
-brw_begin_query(struct pipe_context *pipe, struct pipe_query *q)
+brw_query_begin(struct pipe_context *pipe, struct pipe_query *q)
 {
    struct brw_context *brw = brw_context(pipe);
    struct brw_query_object *query = (struct brw_query_object *)q;
 
    /* Reset our driver's tracking of query state. */
    brw->sws->bo_unreference(query->bo);
+   query->result = 0;
    query->bo = NULL;
    query->first_index = -1;
    query->last_index = -1;
 
    insert_at_head(&brw->query.active_head, query);
-   brw->stats_wm++;
-   brw->dirty.mesa |= PIPE_NEW_QUERY;
+   brw->query.stats_wm++;
+   brw->state.dirty.mesa |= PIPE_NEW_QUERY;
 }
 
 static void
-brw_end_query(struct pipe_context *pipe, struct pipe_query *q)
+brw_query_end(struct pipe_context *pipe, struct pipe_query *q)
 {
    struct brw_context *brw = brw_context(pipe);
    struct brw_query_object *query = (struct brw_query_object *)q;
@@ -129,27 +144,13 @@ brw_end_query(struct pipe_context *pipe, struct pipe_query *q)
    }
 
    remove_from_list(query);
-   brw->stats_wm--;
-   brw->dirty.mesa |= PIPE_NEW_QUERY;
+   brw->query.stats_wm--;
+   brw->state.dirty.mesa |= PIPE_NEW_QUERY;
 }
 
-static void brw_wait_query(struct pipe_context *pipe, struct pipe_query *q)
-{
-   struct brw_query_object *query = (struct brw_query_object *)q;
-
-   brw_queryobj_get_results(query);
-   query->Base.Ready = GL_TRUE;
-}
-
-static void brw_check_query(struct pipe_context *pipe, struct pipe_query *q)
-{
-   struct brw_query_object *query = (struct brw_query_object *)q;
-
-   if (query->bo == NULL || !drm_intel_bo_busy(query->bo)) {
-      brw_queryobj_get_results(query);
-      query->Base.Ready = GL_TRUE;
-   }
-}
+/***********************************************************************
+ * Internal functions and callbacks to implement queries 
+ */
 
 /** Called to set up the query BO and account for its aperture space */
 void
@@ -201,8 +202,17 @@ brw_emit_query_begin(struct brw_context *brw)
 
    foreach(query, &brw->query.active_head) {
       if (query->bo != brw->query.bo) {
+	 uint64_t tmp;
+	 
+	 /* Propogate the results from this buffer to all of the
+	  * active queries, as the bo is going away.
+	  */
 	 if (query->bo != NULL)
-	    brw_queryobj_get_results(query);
+	    brw_query_get_result( &brw->base, 
+				  (struct pipe_query *)query,
+				  FALSE,
+				  &tmp );
+
 	 brw->sws->bo_reference(brw->query.bo);
 	 query->bo = brw->query.bo;
 	 query->first_index = brw->query.index;
@@ -235,12 +245,18 @@ brw_emit_query_end(struct brw_context *brw)
    brw->query.index++;
 }
 
-void brw_init_queryobj_functions(struct dd_function_table *functions)
+void brw_pipe_query_init( struct brw_context *brw )
 {
-   functions->NewQueryObject = brw_new_query_object;
-   functions->DeleteQuery = brw_delete_query;
-   functions->BeginQuery = brw_begin_query;
-   functions->EndQuery = brw_end_query;
-   functions->CheckQuery = brw_check_query;
-   functions->WaitQuery = brw_wait_query;
+   brw->base.create_query = brw_query_create;
+   brw->base.destroy_query = brw_query_destroy;
+   brw->base.begin_query = brw_query_begin;
+   brw->base.end_query = brw_query_end;
+   brw->base.get_query_result = brw_query_get_result;
+}
+
+
+void brw_pipe_query_cleanup( struct brw_context *brw )
+{
+   /* Unreference brw->query.bo ??
+    */
 }
diff --git a/src/gallium/drivers/i965/brw_pipe_sampler.c b/src/gallium/drivers/i965/brw_pipe_sampler.c
index b3069f08c00..bc20eef6fb2 100644
--- a/src/gallium/drivers/i965/brw_pipe_sampler.c
+++ b/src/gallium/drivers/i965/brw_pipe_sampler.c
@@ -14,6 +14,87 @@ static void *brw_create_sampler_state( struct pipe_context *pipe,
 {
    struct brw_sampler_state *sampler = CALLOC_STRUCT(brw_sampler_state);
 
+   switch (key->minfilter) {
+   case GL_NEAREST:
+      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
+      sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
+      break;
+   case GL_LINEAR:
+      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
+      sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
+      break;
+   case GL_NEAREST_MIPMAP_NEAREST:
+      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
+      sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
+      break;
+   case GL_LINEAR_MIPMAP_NEAREST:
+      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
+      sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
+      break;
+   case GL_NEAREST_MIPMAP_LINEAR:
+      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
+      sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
+      break;
+   case GL_LINEAR_MIPMAP_LINEAR:
+      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
+      sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
+      break;
+   default:
+      break;
+   }
+
+   /* Set Anisotropy: 
+    */
+   if (key->max_aniso > 1.0) {
+      sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; 
+      sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC;
+
+      if (key->max_aniso > 2.0) {
+	 sampler->ss3.max_aniso = MIN2((key->max_aniso - 2) / 2,
+				       BRW_ANISORATIO_16);
+      }
+   }
+   else {
+      switch (key->magfilter) {
+      case GL_NEAREST:
+	 sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
+	 break;
+      case GL_LINEAR:
+	 sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
+	 break;
+      default:
+	 break;
+      }
+   }
+
+   sampler->ss1.r_wrap_mode = translate_wrap_mode(key->wrap_r);
+   sampler->ss1.s_wrap_mode = translate_wrap_mode(key->wrap_s);
+   sampler->ss1.t_wrap_mode = translate_wrap_mode(key->wrap_t);
+
+   /* Set LOD bias: 
+    */
+   sampler->ss0.lod_bias = S_FIXED(CLAMP(key->lod_bias, -16, 15), 6);
+
+   sampler->ss0.lod_preclamp = 1; /* OpenGL mode */
+   sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */
+
+   /* Set shadow function: 
+    */
+   if (key->comparemode == GL_COMPARE_R_TO_TEXTURE_ARB) {
+      /* Shadowing is "enabled" by emitting a particular sampler
+       * message (sample_c).  So need to recompile WM program when
+       * shadow comparison is enabled on each/any texture unit.
+       */
+      sampler->ss0.shadow_function =
+	 intel_translate_shadow_compare_func(key->comparefunc);
+   }
+
+   /* Set BaseMipLevel, MaxLOD, MinLOD: 
+    */
+   sampler->ss0.base_level = U_FIXED(0, 1);
+
+   sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(key->maxlod, 0), 13), 6);
+   sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(key->minlod, 0), 13), 6);
 
    return (void *)sampler;
 }
diff --git a/src/gallium/drivers/i965/brw_screen_surface.c b/src/gallium/drivers/i965/brw_screen_surface.c
index 544be6a0896..e0df6cc629b 100644
--- a/src/gallium/drivers/i965/brw_screen_surface.c
+++ b/src/gallium/drivers/i965/brw_screen_surface.c
@@ -1,27 +1,131 @@
-   /* _NEW_BUFFERS */
-   if (IS_965(brw->brw_screen->pci_id) &&
-       !IS_G4X(brw->brw_screen->pci_id)) {
-      for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
-	 struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
-	 struct intel_renderbuffer *irb = intel_renderbuffer(rb);
-
-	 /* The original gen4 hardware couldn't set up WM surfaces pointing
-	  * at an offset within a tile, which can happen when rendering to
-	  * anything but the base level of a texture or the +X face/0 depth.
-	  * This was fixed with the 4 Series hardware.
-	  *
-	  * For these original chips, you would have to make the depth and
-	  * color destination surfaces include information on the texture
-	  * type, LOD, face, and various limits to use them as a destination.
-	  * I would have done this, but there's also a nasty requirement that
-	  * the depth and the color surfaces all be of the same LOD, which
-	  * may be a worse requirement than this alignment.  (Also, we may
-	  * want to just demote the texture to untiled, instead).
-	  */
-	 if (irb->region && 
-	     irb->region->tiling != I915_TILING_NONE &&
-	     (irb->region->draw_offset & 4095)) {
-	    DBG("FALLBACK: non-tile-aligned destination for tiled FBO\n");
-	    return GL_TRUE;
-	 }
+
+#include "pipe/p_screen.h"
+#include "brw_screen.h"
+
+struct brw_surface_id {
+   unsigned face:3;
+   unsigned zslice:13;
+   unsigned level:16;
+};
+
+static boolean need_linear_view( struct brw_screen *brw_screen,
+				 struct brw_texture *brw_texture,
+				 unsigned face,
+				 unsigned level,
+				 unsigned zslice )
+{
+#if 0
+   /* XXX: what about IDGNG?
+    */
+   if (!BRW_IS_G4X(brw->brw_screen->pci_id))
+   {
+      struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
+      struct intel_renderbuffer *irb = intel_renderbuffer(rb);
+
+      /* The original gen4 hardware couldn't set up WM surfaces pointing
+       * at an offset within a tile, which can happen when rendering to
+       * anything but the base level of a texture or the +X face/0 depth.
+       * This was fixed with the 4 Series hardware.
+       *
+       * For these original chips, you would have to make the depth and
+       * color destination surfaces include information on the texture
+       * type, LOD, face, and various limits to use them as a destination.
+       *
+       * This is easy in Gallium as surfaces are all backed by
+       * textures, but there's also a nasty requirement that the depth
+       * and the color surfaces all be of the same LOD, which is
+       * harder to get around as we can't look at a surface in
+       * isolation and decide if it's legal.
+       *
+       * Instead, end up being pessimistic and say that for i965,
+       * ... ??
+       */
+      if (brw_tex->tiling != I915_TILING_NONE &&
+	  (brw_tex_image_offset(brw_tex, face, level, zslize) & 4095)) {
+	 if (BRW_DEBUG & DEBUG_VIEW)
+	    debug_printf("%s: need surface view for non-aligned tex image\n",
+			 __FUNCTION__);
+	 return GL_TRUE;
       }
+   }
+#endif
+
+   /* Tiled 3d textures don't have subsets that look like 2d surfaces:
+    */
+   
+   /* Everything else should be fine to render to in-place:
+    */
+   return GL_FALSE;
+}
+
+/* Look at all texture views and figure out if any of them need to be
+ * back-copied into the texture for sampling
+ */
+void brw_update_texture( struct pipe_screen *screen,
+			 struct pipe_texture *texture )
+{
+   /* currently nothing to do */
+}
+
+
+static struct pipe_surface *create_linear_view( struct brw_screen *brw_screen,
+						struct brw_texture *brw_tex,
+						struct brw_surface_id id )
+{
+   
+}
+
+static struct pipe_surface *create_in_place_view( struct brw_screen *brw_screen,
+						  struct brw_texture *brw_tex,
+						  struct brw_surface_id id )
+{
+   struct brw_surface *surface = CALLOC_STRUCT(brw_surface);
+   surface->id = id;
+   
+}
+
+/* Get a surface which is view into a texture 
+ */
+struct pipe_surface *brw_get_tex_surface(struct pipe_screen *screen,
+					 struct pipe_texture *texture,
+					 unsigned face, unsigned level,
+					 unsigned zslice,
+					 unsigned usage )
+{
+   struct brw_screen *bscreen = brw_screen(screen);
+   struct brw_surface_id id;
+
+   id.face = face;
+   id.level = level;
+   id.zslice = zslice;
+
+   if (need_linear_view(brw_screen, brw_tex, id)) 
+      type = BRW_VIEW_LINEAR;
+   else
+      type = BRW_VIEW_IN_PLACE;
+
+   
+   foreach (surface, texture->views[type]) {
+      if (id.value == surface->id.value)
+	 return surface;
+   }
+
+   switch (type) {
+   case BRW_VIEW_LINEAR:
+      surface = create_linear_view( texture, id, type );
+      break;
+   case BRW_VIEW_IN_PLACE:
+      surface = create_in_place_view( texture, id, type );
+      break;
+   default:
+      return NULL;
+   }
+
+   insert_at_head( texture->views[type], surface );
+   return surface;
+}
+
+
+void brw_tex_surface_destroy( struct pipe_surface *surface )
+{
+}
diff --git a/src/gallium/drivers/i965/brw_tex_layout.c b/src/gallium/drivers/i965/brw_screen_texture.c
index 50c30878c6f..50c30878c6f 100644
--- a/src/gallium/drivers/i965/brw_tex_layout.c
+++ b/src/gallium/drivers/i965/brw_screen_texture.c
diff --git a/src/gallium/drivers/i965/brw_sf.c b/src/gallium/drivers/i965/brw_sf.c
index 1b73b3fd51c..013d839e37d 100644
--- a/src/gallium/drivers/i965/brw_sf.c
+++ b/src/gallium/drivers/i965/brw_sf.c
@@ -29,11 +29,12 @@
   *   Keith Whitwell <[email protected]>
   */
   
+#include "pipe/p_state.h"
 
 #include "brw_batchbuffer.h"
-
 #include "brw_defines.h"
 #include "brw_context.h"
+#include "brw_pipe_rast.h"
 #include "brw_eu.h"
 #include "brw_util.h"
 #include "brw_sf.h"
@@ -45,7 +46,6 @@ static void compile_sf_prog( struct brw_context *brw,
    struct brw_sf_compile c;
    const GLuint *program;
    GLuint program_size;
-   GLuint i, idx;
 
    memset(&c, 0, sizeof(c));
 
@@ -54,7 +54,7 @@ static void compile_sf_prog( struct brw_context *brw,
    brw_init_compile(brw, &c.func);
 
    c.key = *key;
-   c.nr_attrs = util_count_bits(c.key.attrs);
+   c.nr_attrs = c.key.nr_attrs;
    c.nr_attr_regs = (c.nr_attrs+1)/2;
    c.nr_setup_attrs = c.key.nr_attrs;
    c.nr_setup_regs = (c.nr_setup_attrs+1)/2;
@@ -62,21 +62,6 @@ static void compile_sf_prog( struct brw_context *brw,
    c.prog_data.urb_read_length = c.nr_attr_regs;
    c.prog_data.urb_entry_size = c.nr_setup_regs * 2;
 
-   /* Construct map from attribute number to position in the vertex.
-    */
-   for (i = idx = 0; i < VERT_RESULT_MAX; i++) 
-      if (c.key.attrs & (1<<i)) {
-	 c.attr_to_idx[i] = idx;
-	 c.idx_to_attr[idx] = i;
-	 if (i >= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) {
-            c.point_attrs[i].CoordReplace = 
-               ctx->Point.CoordReplace[i - VERT_RESULT_TEX0];
-	 }
-         else {
-            c.point_attrs[i].CoordReplace = GL_FALSE;
-         }
-	 idx++;
-      }
    
    /* Which primitive?  Or all three? 
     */
@@ -122,7 +107,7 @@ static void compile_sf_prog( struct brw_context *brw,
 
 /* Calculate interpolants for triangle and line rasterization.
  */
-static void upload_sf_prog(struct brw_context *brw)
+static int upload_sf_prog(struct brw_context *brw)
 {
    struct brw_sf_prog_key key;
 
@@ -131,46 +116,49 @@ static void upload_sf_prog(struct brw_context *brw)
    /* Populate the key, noting state dependencies:
     */
    /* CACHE_NEW_VS_PROG */
-   key.attrs = brw->vs.prog_data->nr_outputs_written; 
+   key.nr_attrs = brw->curr.vertex_shader->info.file_max[TGSI_FILE_OUTPUT] + 1;
+
+
+   /* XXX: this is probably where the mapping between vertex shader
+    * outputs and fragment shader inputs should be handled.  Assume
+    * for now 1:1 correspondance.
+    *
+    * XXX: scan frag shader inputs to work out linear vs. perspective
+    * interpolation below.
+    *
+    * XXX: as long as we're hard-wiring, is eg. position required to
+    * be linear?
+    */
+   key.linear_attrs = 0;
+   key.persp_attrs = (1 << key.nr_attrs) - 1;
 
    /* BRW_NEW_REDUCED_PRIMITIVE */
    switch (brw->reduced_primitive) {
-   case GL_TRIANGLES: 
-      /* NOTE: We just use the edgeflag attribute as an indicator that
-       * unfilled triangles are active.  We don't actually do the
-       * edgeflag testing here, it is already done in the clip
-       * program.
+   case PIPE_PRIM_TRIANGLES: 
+      /* PIPE_NEW_RAST
        */
-      if (key.attrs & (1<<VERT_RESULT_EDGE))
+      if (brw->curr.rast->templ.fill_cw != PIPE_POLYGON_MODE_FILL ||
+	  brw->curr.rast->templ.fill_ccw != PIPE_POLYGON_MODE_FILL)
 	 key.primitive = SF_UNFILLED_TRIS;
       else
 	 key.primitive = SF_TRIANGLES;
       break;
-   case GL_LINES: 
+   case PIPE_PRIM_LINES: 
       key.primitive = SF_LINES; 
       break;
-   case GL_POINTS: 
+   case PIPE_PRIM_POINTS: 
       key.primitive = SF_POINTS; 
       break;
    }
 
-   key.do_point_sprite = ctx->Point.PointSprite;
-   key.SpriteOrigin = ctx->Point.SpriteOrigin;
-   /* _NEW_LIGHT */
-   key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
-   key.do_twoside_color = (ctx->Light.Enabled && ctx->Light.Model.TwoSide);
+   key.do_point_sprite = brw->curr.rast->templ.point_sprite;
+   key.sprite_origin_lower_left = 0; /* XXX: ctx->Point.SpriteOrigin - fix rast state */
+   key.do_flat_shading = brw->curr.rast->templ.flatshade;
+   key.do_twoside_color = brw->curr.rast->templ.light_twoside;
 
-   /* _NEW_HINT */
-   key.linear_color = 0;
-
-   /* _NEW_POLYGON */
    if (key.do_twoside_color) {
-      /* If we're rendering to a FBO, we have to invert the polygon
-       * face orientation, just as we invert the viewport in
-       * sf_unit_create_from_key().  ctx->DrawBuffer->Name will be
-       * nonzero if we're rendering to such an FBO.
-       */
-      key.frontface_ccw = (ctx->Polygon.FrontFace == GL_CCW) ^ (ctx->DrawBuffer->Name != 0);
+      key.frontface_ccw = (brw->curr.rast->templ.front_winding == 
+			   PIPE_WINDING_CCW);
    }
 
    brw->sws->bo_unreference(brw->sf.prog_bo);
@@ -180,14 +168,16 @@ static void upload_sf_prog(struct brw_context *brw)
 				      &brw->sf.prog_data);
    if (brw->sf.prog_bo == NULL)
       compile_sf_prog( brw, &key );
+
+   return 0;
 }
 
 
 const struct brw_tracked_state brw_sf_prog = {
    .dirty = {
-      .mesa  = (_NEW_HINT | _NEW_LIGHT | _NEW_POLYGON | _NEW_POINT),
+      .mesa  = (PIPE_NEW_RAST | PIPE_NEW_VERTEX_SHADER),
       .brw   = (BRW_NEW_REDUCED_PRIMITIVE),
-      .cache = CACHE_NEW_VS_PROG
+      .cache = 0
    },
    .prepare = upload_sf_prog
 };
diff --git a/src/gallium/drivers/i965/brw_sf.h b/src/gallium/drivers/i965/brw_sf.h
index c99116b8b1b..0b7003dc5ef 100644
--- a/src/gallium/drivers/i965/brw_sf.h
+++ b/src/gallium/drivers/i965/brw_sf.h
@@ -49,14 +49,21 @@ struct brw_sf_prog_key {
     */
    GLuint persp_attrs:32;
    GLuint linear_attrs:32;
+   GLuint point_coord_replace_attrs:32;
 
+   GLuint nr_attrs:8;
    GLuint primitive:2;
    GLuint do_twoside_color:1;
    GLuint do_flat_shading:1;
    GLuint frontface_ccw:1;
    GLuint do_point_sprite:1;
    GLuint sprite_origin_lower_left:1;
-   GLuint pad:25;
+   GLuint pad:17;
+
+   GLuint attr_col0:8;
+   GLuint attr_col1:8;
+   GLuint attr_bfc0:8;
+   GLuint attr_bfc1:8;
 };
 
 struct brw_sf_point_tex {
@@ -101,9 +108,7 @@ struct brw_sf_compile {
    GLuint nr_setup_attrs;
    GLuint nr_setup_regs;
 
-   GLubyte attr_to_idx[VERT_RESULT_MAX];   
-   GLubyte idx_to_attr[VERT_RESULT_MAX];   
-   struct brw_sf_point_tex point_attrs[VERT_RESULT_MAX];
+   GLuint point_coord_replace_mask;
 };
 
  
diff --git a/src/gallium/drivers/i965/brw_sf_emit.c b/src/gallium/drivers/i965/brw_sf_emit.c
index 4acb2b7d72e..db52c9553e7 100644
--- a/src/gallium/drivers/i965/brw_sf_emit.c
+++ b/src/gallium/drivers/i965/brw_sf_emit.c
@@ -43,17 +43,12 @@ static struct brw_reg get_vert_attr(struct brw_sf_compile *c,
 				    struct brw_reg vert,
 				    GLuint attr)
 {
-   GLuint off = c->attr_to_idx[attr] / 2;
-   GLuint sub = c->attr_to_idx[attr] % 2;
+   GLuint off = attr / 2;
+   GLuint sub = attr % 2;
 
    return brw_vec4_grf(vert.nr + off, sub * 4);
 }
 
-static GLboolean have_attr(struct brw_sf_compile *c,
-			   GLuint attr)
-{
-   return (c->key.attrs & (1<<attr)) ? 1 : 0;
-}
 
 /*********************************************************************** 
  * Twoside lighting
@@ -62,15 +57,16 @@ static void copy_bfc( struct brw_sf_compile *c,
 		      struct brw_reg vert )
 {
    struct brw_compile *p = &c->func;
-   GLuint i;
 
-   for (i = 0; i < 2; i++) {
-      if (have_attr(c, VERT_RESULT_COL0+i) &&
-	  have_attr(c, VERT_RESULT_BFC0+i))
-	 brw_MOV(p, 
-		 get_vert_attr(c, vert, VERT_RESULT_COL0+i), 
-		 get_vert_attr(c, vert, VERT_RESULT_BFC0+i));
-   }
+   if (c->key.attr_col0 && c->key.attr_bfc0)
+      brw_MOV(p, 
+	      get_vert_attr(c, vert, c->key.attr_col0), 
+	      get_vert_attr(c, vert, c->key.attr_bfc0));
+
+   if (c->key.attr_col1 && c->key.attr_bfc1)
+      brw_MOV(p, 
+	      get_vert_attr(c, vert, c->key.attr_col1), 
+	      get_vert_attr(c, vert, c->key.attr_bfc1));
 }
 
 
@@ -89,8 +85,8 @@ static void do_twoside_color( struct brw_sf_compile *c )
     * for user-supplied vertex programs, as t_vp_build.c always does
     * the right thing.
     */
-   if (!(have_attr(c, VERT_RESULT_COL0) && have_attr(c, VERT_RESULT_BFC0)) &&
-       !(have_attr(c, VERT_RESULT_COL1) && have_attr(c, VERT_RESULT_BFC1)))
+   if (!(c->key.attr_col0 && c->key.attr_bfc0) &&
+       !(c->key.attr_col1 && c->key.attr_bfc1))
       return;
    
    /* Need to use BRW_EXECUTE_4 and also do an 4-wide compare in order
@@ -126,14 +122,17 @@ static void copy_colors( struct brw_sf_compile *c,
 		     struct brw_reg src)
 {
    struct brw_compile *p = &c->func;
-   GLuint i;
 
-   for (i = VERT_RESULT_COL0; i <= VERT_RESULT_COL1; i++) {
-      if (have_attr(c,i))
-	 brw_MOV(p, 
-		 get_vert_attr(c, dst, i), 
-		 get_vert_attr(c, src, i));
-   }
+   if (c->key.attr_col0)
+      brw_MOV(p, 
+	      get_vert_attr(c, dst, c->key.attr_col0), 
+	      get_vert_attr(c, src, c->key.attr_col0));
+
+   if (c->key.attr_col1)
+      brw_MOV(p, 
+	      get_vert_attr(c, dst, c->key.attr_col1), 
+	      get_vert_attr(c, src, c->key.attr_col1));
+
 }
 
 
@@ -146,10 +145,16 @@ static void do_flatshade_triangle( struct brw_sf_compile *c )
 {
    struct brw_compile *p = &c->func;
    struct brw_reg ip = brw_ip_reg();
-   GLuint nr = util_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS);
    GLuint jmpi = 1;
+   GLuint nr = 0;
 
-   if (!nr)
+   if (c->key.attr_col0)
+      nr++;
+
+   if (c->key.attr_col1)
+      nr++;
+
+   if (nr == 0)
       return;
 
    /* Already done in clip program:
@@ -184,10 +189,16 @@ static void do_flatshade_line( struct brw_sf_compile *c )
 {
    struct brw_compile *p = &c->func;
    struct brw_reg ip = brw_ip_reg();
-   GLuint nr = util_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS);
    GLuint jmpi = 1;
+   GLuint nr = 0;
+
+   if (c->key.attr_col0)
+      nr++;
+
+   if (c->key.attr_col1)
+      nr++;
 
-   if (!nr)
+   if (nr == 0)
       return;
 
    /* Already done in clip program: 
@@ -319,10 +330,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c,
    *pc_linear = 0;
    *pc = 0xf;
       
-   if (persp_mask & (1 << c->idx_to_attr[reg*2])) 
+   if (persp_mask & (1 << (reg*2))) 
       *pc_persp = 0xf;
 
-   if (linear_mask & (1 << c->idx_to_attr[reg*2])) 
+   if (linear_mask & (1 << (reg*2))) 
       *pc_linear = 0xf;
 
    /* Maybe only processs one attribute on the final round:
@@ -330,10 +341,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c,
    if (reg*2+1 < c->nr_setup_attrs) {
       *pc |= 0xf0;
 
-      if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) 
+      if (persp_mask & (1 << (reg*2+1))) 
 	 *pc_persp |= 0xf0;
 
-      if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) 
+      if (linear_mask & (1 << (reg*2+1))) 
 	 *pc_linear |= 0xf0;
    }
 
@@ -513,24 +524,28 @@ void brw_emit_point_sprite_setup( struct brw_sf_compile *c, GLboolean allocate)
       alloc_regs(c);
 
    copy_z_inv_w(c);
+
    for (i = 0; i < c->nr_setup_regs; i++)
    {
-      struct brw_sf_point_tex *tex = &c->point_attrs[c->idx_to_attr[2*i]];
+      /* XXX: only seems to check point_coord_replace_attrs for every
+       * second attribute?!?
+       */
+      boolean coord_replace = !!(c->key.point_coord_replace_attrs & (1<<(2*i)));
       struct brw_reg a0 = offset(c->vert[0], i);
       GLushort pc, pc_persp, pc_linear;
       GLboolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear);
             
       if (pc_persp)
       {				
-	  if (!tex->CoordReplace) {
-	      brw_set_predicate_control_flag_value(p, pc_persp);
-	      brw_MUL(p, a0, a0, c->inv_w[0]);
-	  }
+	 if (coord_replace) {
+	    brw_set_predicate_control_flag_value(p, pc_persp);
+	    brw_MUL(p, a0, a0, c->inv_w[0]);
+	 }
       }
 
-      if (tex->CoordReplace) {
-	  /* Caculate 1.0/PointWidth */
-	  brw_math(&c->func,
+      if (coord_replace) {
+	 /* Caculate 1.0/PointWidth */
+	 brw_math(&c->func,
 		  c->tmp,
 		  BRW_MATH_FUNCTION_INV,
 		  BRW_MATH_SATURATE_NONE,
@@ -539,33 +554,37 @@ void brw_emit_point_sprite_setup( struct brw_sf_compile *c, GLboolean allocate)
 		  BRW_MATH_DATA_SCALAR,
 		  BRW_MATH_PRECISION_FULL);
 
-	  if (c->key.SpriteOrigin == GL_LOWER_LEFT) {
-	   	brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]);
-		brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0));
-	  	brw_MUL(p, c->m2Cy, c->tmp, negate(c->inv_w[0]));
-		brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0));
-	  } else {
-	   	brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]);
-		brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0));
-	  	brw_MUL(p, c->m2Cy, c->tmp, c->inv_w[0]);
-		brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0));
-	  }
-      } else {
-	  brw_MOV(p, c->m1Cx, brw_imm_ud(0));
-	  brw_MOV(p, c->m2Cy, brw_imm_ud(0));
+	 if (c->key.sprite_origin_lower_left) {
+	    brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]);
+	    brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0));
+	    brw_MUL(p, c->m2Cy, c->tmp, negate(c->inv_w[0]));
+	    brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0));
+	 } 
+	 else {
+	    brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]);
+	    brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0));
+	    brw_MUL(p, c->m2Cy, c->tmp, c->inv_w[0]);
+	    brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0));
+	 }
+      } 
+      else {
+	 brw_MOV(p, c->m1Cx, brw_imm_ud(0));
+	 brw_MOV(p, c->m2Cy, brw_imm_ud(0));
       }
 
       {
 	 brw_set_predicate_control_flag_value(p, pc); 
-	 if (tex->CoordReplace) {
-	     if (c->key.sprite_origin_lower_left) {
-		 brw_MUL(p, c->m3C0, c->inv_w[0], brw_imm_f(1.0));
-		 brw_MOV(p, vec1(suboffset(c->m3C0, 0)), brw_imm_f(0.0));
-	     }
-	     else
-		 brw_MOV(p, c->m3C0, brw_imm_f(0.0));
-	 } else {
-	 	brw_MOV(p, c->m3C0, a0); /* constant value */
+	 if (coord_replace) {
+	    if (c->key.sprite_origin_lower_left) {
+	       brw_MUL(p, c->m3C0, c->inv_w[0], brw_imm_f(1.0));
+	       brw_MOV(p, vec1(suboffset(c->m3C0, 0)), brw_imm_f(0.0));
+	    }
+	    else {
+	       brw_MOV(p, c->m3C0, brw_imm_f(0.0));
+	    }
+	 } 
+	 else {
+	    brw_MOV(p, c->m3C0, a0); /* constant value */
 	 }
 
 	 /* Copy m0..m3 to URB. 
diff --git a/src/gallium/drivers/i965/brw_sf_state.c b/src/gallium/drivers/i965/brw_sf_state.c
index 648a16a038c..fbc9f15eb43 100644
--- a/src/gallium/drivers/i965/brw_sf_state.c
+++ b/src/gallium/drivers/i965/brw_sf_state.c
@@ -29,58 +29,48 @@
   *   Keith Whitwell <[email protected]>
   */
    
+#include "util/u_math.h"
 
+#include "pipe/p_state.h"
 
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
+#include "brw_debug.h"
+#include "brw_pipe_rast.h"
 
-static void upload_sf_vp(struct brw_context *brw)
+static int upload_sf_vp(struct brw_context *brw)
 {
-   const GLfloat depth_scale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   const struct pipe_viewport_state *vp = &brw->curr.vp;
+   const struct pipe_scissor_state *scissor = &brw->curr.scissor;
    struct brw_sf_viewport sfv;
-   GLfloat y_scale, y_bias;
-   const GLfloat *v = ctx->Viewport._WindowMap.m;
 
    memset(&sfv, 0, sizeof(sfv));
 
-   y_scale = 1.0;
-   y_bias = 0;
+   /* PIPE_NEW_VIEWPORT, PIPE_NEW_SCISSOR */
 
-   /* _NEW_VIEWPORT */
+   sfv.viewport.m00 = vp->scale[0];
+   sfv.viewport.m11 = vp->scale[1];
+   sfv.viewport.m22 = vp->scale[2];
+   sfv.viewport.m30 = vp->translate[0];
+   sfv.viewport.m31 = vp->translate[1];
+   sfv.viewport.m32 = vp->translate[2];
 
-   sfv.viewport.m00 = v[MAT_SX];
-   sfv.viewport.m11 = v[MAT_SY] * y_scale;
-   sfv.viewport.m22 = v[MAT_SZ] * depth_scale;
-   sfv.viewport.m30 = v[MAT_TX];
-   sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias;
-   sfv.viewport.m32 = v[MAT_TZ] * depth_scale;
-
-   /* _NEW_SCISSOR | _NEW_BUFFERS | _NEW_VIEWPORT
-    * for DrawBuffer->_[XY]{min,max}
-    */
-
-   /* The scissor only needs to handle the intersection of drawable and
-    * scissor rect.
-    *
-    * Note that the hardware's coordinates are inclusive, while Mesa's min is
-    * inclusive but max is exclusive.
-    */
-   /* Y=0=bottom */
-   sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
-   sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
-   sfv.scissor.ymin = ctx->DrawBuffer->_Ymin;
-   sfv.scissor.ymax = ctx->DrawBuffer->_Ymax - 1;
+   sfv.scissor.xmin = scissor->minx;
+   sfv.scissor.xmax = scissor->maxx; /* -1 ?? */
+   sfv.scissor.ymin = scissor->miny;
+   sfv.scissor.ymax = scissor->maxy; /* -1 ?? */
 
    brw->sws->bo_unreference(brw->sf.vp_bo);
    brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 );
+
+   return 0;
 }
 
 const struct brw_tracked_state brw_sf_vp = {
    .dirty = {
-      .mesa  = (_NEW_VIEWPORT | 
-		_NEW_SCISSOR |
-		_NEW_BUFFERS),
+      .mesa  = (PIPE_NEW_VIEWPORT | 
+		PIPE_NEW_SCISSOR),
       .brw   = 0,
       .cache = 0
    },
@@ -90,15 +80,17 @@ const struct brw_tracked_state brw_sf_vp = {
 struct brw_sf_unit_key {
    unsigned int total_grf;
    unsigned int urb_entry_read_length;
-
    unsigned int nr_urb_entries, urb_size, sfsize;
-
-   GLenum front_face, cull_face, provoking_vertex;
+   
    unsigned scissor:1;
    unsigned line_smooth:1;
    unsigned point_sprite:1;
    unsigned point_attenuated:1;
-   unsigned render_to_fbo:1;
+   unsigned front_face:2;
+   unsigned cull_mode:2;
+   unsigned flatshade_first:1;
+   unsigned gl_rasterization_rules:1;
+   unsigned line_last_pixel_enable:1;
    float line_width;
    float point_size;
 };
@@ -106,6 +98,7 @@ struct brw_sf_unit_key {
 static void
 sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
 {
+   const struct pipe_rasterizer_state *rast = &brw->curr.rast->templ;
    memset(key, 0, sizeof(*key));
 
    /* CACHE_NEW_SF_PROG */
@@ -117,25 +110,22 @@ sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
    key->urb_size = brw->urb.vsize;
    key->sfsize = brw->urb.sfsize;
 
-   key->scissor = ctx->Scissor.Enabled;
-   key->front_face = ctx->Polygon.FrontFace;
-
-   if (ctx->Polygon.CullFlag)
-      key->cull_face = ctx->Polygon.CullFaceMode;
-   else
-      key->cull_face = GL_NONE;
-
-   key->line_width = ctx->Line.Width;
-   key->line_smooth = ctx->Line.SmoothFlag;
-
-   key->point_sprite = ctx->Point.PointSprite;
-   key->point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
-   key->point_attenuated = ctx->Point._Attenuated;
-
-   /* _NEW_LIGHT */
-   key->provoking_vertex = ctx->Light.ProvokingVertex;
-
-   key->render_to_fbo = 1;
+   /* PIPE_NEW_RAST */
+   key->scissor = rast->scissor;
+   key->front_face = rast->front_winding;
+   key->cull_mode = rast->cull_mode;
+   key->line_smooth = rast->line_smooth;
+   key->line_width = rast->line_width;
+   key->flatshade_first = rast->flatshade_first;
+   key->line_last_pixel_enable = rast->line_last_pixel;
+   key->gl_rasterization_rules = rast->gl_rasterization_rules;
+
+   key->point_sprite = rast->point_sprite;
+   key->point_attenuated = rast->point_size_per_vertex;
+
+   key->point_size = CLAMP(rast->point_size, 
+			   rast->point_size_min, 
+			   rast->point_size_max);
 }
 
 static struct brw_winsys_buffer *
@@ -147,7 +137,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    int chipset_max_threads;
    memset(&sf, 0, sizeof(sf));
 
-   sf.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
+   sf.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1;
    sf.thread0.kernel_start_pointer = brw->sf.prog_bo->offset >> 6; /* reloc */
 
    sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
@@ -174,10 +164,10 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 
    sf.thread4.max_threads = MIN2(chipset_max_threads, key->nr_urb_entries) - 1;
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (BRW_DEBUG & DEBUG_SINGLE_THREAD)
       sf.thread4.max_threads = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (BRW_DEBUG & DEBUG_STATS)
       sf.thread4.stats_enable = 1;
 
    /* CACHE_NEW_SF_VP */
@@ -185,31 +175,30 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 
    sf.sf5.viewport_transform = 1;
 
-   /* _NEW_SCISSOR */
    if (key->scissor)
       sf.sf6.scissor = 1;
 
-   /* _NEW_POLYGON */
-   if (key->front_face == GL_CCW)
+   if (key->front_face == PIPE_WINDING_CCW)
       sf.sf5.front_winding = BRW_FRONTWINDING_CCW;
    else
       sf.sf5.front_winding = BRW_FRONTWINDING_CW;
 
-   switch (key->cull_face) {
-   case GL_FRONT:
-      sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
+   switch (key->cull_mode) {
+   case PIPE_WINDING_CCW:
+   case PIPE_WINDING_CW:
+      sf.sf6.cull_mode = (key->front_face == key->cull_mode ?
+			  BRW_CULLMODE_FRONT :
+			  BRW_CULLMODE_BACK);
       break;
-   case GL_BACK:
-      sf.sf6.cull_mode = BRW_CULLMODE_BACK;
-      break;
-   case GL_FRONT_AND_BACK:
+   case PIPE_WINDING_BOTH:
       sf.sf6.cull_mode = BRW_CULLMODE_BOTH;
       break;
-   case GL_NONE:
+   case PIPE_WINDING_NONE:
       sf.sf6.cull_mode = BRW_CULLMODE_NONE;
       break;
    default:
       assert(0);
+      sf.sf6.cull_mode = BRW_CULLMODE_NONE;
       break;
    }
 
@@ -223,9 +212,9 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    else if (sf.sf6.line_width <= 0x2)
        sf.sf6.line_width = 0;
 
-   /* _NEW_BUFFERS */
-   key->render_to_fbo = 1;
-   if (!key->render_to_fbo) {
+   /* XXX: gl_rasterization_rules?  something else?
+    */
+   if (0) {
       /* Rendering to an OpenGL window */
       sf.sf6.point_rast_rule = BRW_RASTRULE_UPPER_RIGHT;
    }
@@ -261,7 +250,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 
    /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
     */
-   if (key->provoking_vertex == GL_LAST_VERTEX_CONVENTION) {
+   if (!key->flatshade_first) {
       sf.sf7.trifan_pv = 2;
       sf.sf7.linestrip_pv = 1;
       sf.sf7.tristrip_pv = 2;
@@ -270,12 +259,19 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
       sf.sf7.linestrip_pv = 0;
       sf.sf7.tristrip_pv = 0;
    }
-   sf.sf7.line_last_pixel_enable = 0;
+
+   sf.sf7.line_last_pixel_enable = key->line_last_pixel_enable;
 
    /* Set bias for OpenGL rasterization rules:
     */
-   sf.sf6.dest_org_vbias = 0x8;
-   sf.sf6.dest_org_hbias = 0x8;
+   if (key->gl_rasterization_rules) {
+      sf.sf6.dest_org_vbias = 0x8;
+      sf.sf6.dest_org_hbias = 0x8;
+   }
+   else {
+      sf.sf6.dest_org_vbias = 0x0;
+      sf.sf6.dest_org_hbias = 0x0;
+   }
 
    bo = brw_upload_cache(&brw->cache, BRW_SF_UNIT,
 			 key, sizeof(*key),
@@ -287,23 +283,23 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
     * something loaded through the GPE (L2 ISC), so it's INSTRUCTION domain.
     */
    /* Emit SF program relocation */
-   dri_bo_emit_reloc(bo,
-		     I915_GEM_DOMAIN_INSTRUCTION, 0,
-		     sf.thread0.grf_reg_count << 1,
-		     offsetof(struct brw_sf_unit_state, thread0),
-		     brw->sf.prog_bo);
+   brw->sws->bo_emit_reloc(bo,
+			   I915_GEM_DOMAIN_INSTRUCTION, 0,
+			   sf.thread0.grf_reg_count << 1,
+			   offsetof(struct brw_sf_unit_state, thread0),
+			   brw->sf.prog_bo);
 
    /* Emit SF viewport relocation */
-   dri_bo_emit_reloc(bo,
-		     I915_GEM_DOMAIN_INSTRUCTION, 0,
-		     sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
-		     offsetof(struct brw_sf_unit_state, sf5),
-		     brw->sf.vp_bo);
+   brw->sws->bo_emit_reloc(bo,
+			   I915_GEM_DOMAIN_INSTRUCTION, 0,
+			   sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
+			   offsetof(struct brw_sf_unit_state, sf5),
+			   brw->sf.vp_bo);
 
    return bo;
 }
 
-static void upload_sf_unit( struct brw_context *brw )
+static int upload_sf_unit( struct brw_context *brw )
 {
    struct brw_sf_unit_key key;
    struct brw_winsys_buffer *reloc_bufs[2];
@@ -321,16 +317,12 @@ static void upload_sf_unit( struct brw_context *brw )
    if (brw->sf.state_bo == NULL) {
       brw->sf.state_bo = sf_unit_create_from_key(brw, &key, reloc_bufs);
    }
+   return 0;
 }
 
 const struct brw_tracked_state brw_sf_unit = {
    .dirty = {
-      .mesa  = (_NEW_POLYGON | 
-		_NEW_LIGHT |
-		_NEW_LINE | 
-		_NEW_POINT | 
-		_NEW_SCISSOR |
-		_NEW_BUFFERS),
+      .mesa  = (PIPE_NEW_RAST),
       .brw   = BRW_NEW_URB_FENCE,
       .cache = (CACHE_NEW_SF_VP |
 		CACHE_NEW_SF_PROG)
diff --git a/src/gallium/drivers/i965/brw_state.h b/src/gallium/drivers/i965/brw_state.h
index 663fc839dfe..2275e9ad690 100644
--- a/src/gallium/drivers/i965/brw_state.h
+++ b/src/gallium/drivers/i965/brw_state.h
@@ -168,9 +168,20 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
 void brw_destroy_batch_cache( struct brw_context *brw );
 void brw_clear_batch_cache( struct brw_context *brw );
 
-/* brw_wm_surface_state.c */
+/***********************************************************************
+ * brw_wm_surface_state.c 
+ */
 struct brw_winsys_buffer *
 brw_create_constant_surface( struct brw_context *brw,
                              struct brw_surface_key *key );
 
+/***********************************************************************
+ * brw_state_debug.c
+ */
+void brw_update_dirty_counts( unsigned mesa,
+			      unsigned brw,
+			      unsigned cache );
+
+
+
 #endif
diff --git a/src/gallium/drivers/i965/brw_state_batch.c b/src/gallium/drivers/i965/brw_state_batch.c
index 324fce51636..7d212e5c247 100644
--- a/src/gallium/drivers/i965/brw_state_batch.c
+++ b/src/gallium/drivers/i965/brw_state_batch.c
@@ -46,7 +46,7 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
    struct brw_cached_batch_item *item = brw->cached_batch_items;
    struct header *newheader = (struct header *)data;
 
-   if (brw->emit_state_always) {
+   if (brw->flags.always_emit_state) {
       brw_batchbuffer_data(brw->batch, data, sz, IGNORE_CLIPRECTS);
       return GL_TRUE;
    }
@@ -56,8 +56,8 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
 	 if (item->sz == sz && memcmp(item->header, newheader, sz) == 0)
 	    return GL_FALSE;
 	 if (item->sz != sz) {
-	    _mesa_free(item->header);
-	    item->header = _mesa_malloc(sz);
+	    FREE(item->header);
+	    item->header = MALLOC(sz);
 	    item->sz = sz;
 	 }
 	 goto emit;
@@ -67,7 +67,7 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
 
    assert(!item);
    item = CALLOC_STRUCT(brw_cached_batch_item);
-   item->header = _mesa_malloc(sz);
+   item->header = MALLOC(sz);
    item->sz = sz;
    item->next = brw->cached_batch_items;
    brw->cached_batch_items = item;
diff --git a/src/gallium/drivers/i965/brw_state_cache.c b/src/gallium/drivers/i965/brw_state_cache.c
index 97f88b3ab39..4310d01ba2e 100644
--- a/src/gallium/drivers/i965/brw_state_cache.c
+++ b/src/gallium/drivers/i965/brw_state_cache.c
@@ -55,7 +55,9 @@
  * only one of the two buffers referenced gets put into the offset, and the
  * incorrect program is run for the other instance.
  */
+#include "util/u_memory.h"
 
+#include "brw_debug.h"
 #include "brw_state.h"
 #include "brw_batchbuffer.h"
 
@@ -107,9 +109,9 @@ update_cache_last(struct brw_cache *cache, enum brw_cache_id cache_id,
    if (bo == cache->last_bo[cache_id])
       return; /* no change */
 
-   brw->sws->bo_unreference(cache->last_bo[cache_id]);
+   cache->sws->bo_unreference(cache->last_bo[cache_id]);
    cache->last_bo[cache_id] = bo;
-   brw->sws->bo_reference(cache->last_bo[cache_id]);
+   cache->sws->bo_reference(cache->last_bo[cache_id]);
    cache->brw->state.dirty.cache |= 1 << cache_id;
 }
 
@@ -127,7 +129,7 @@ search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
    for (c = cache->items[hash % cache->size]; c; c = c->next)
       bucketcount++;
 
-   fprintf(stderr, "bucket %d/%d = %d/%d items\n", hash % cache->size,
+   debug_printf("bucket %d/%d = %d/%d items\n", hash % cache->size,
 	   cache->size, bucketcount, cache->n_items);
 #endif
 
@@ -154,7 +156,7 @@ rehash(struct brw_cache *cache)
    GLuint size, i;
 
    size = cache->size * 3;
-   items = (struct brw_cache_item**) _mesa_calloc(size * sizeof(*items));
+   items = (struct brw_cache_item**) CALLOC(size, sizeof(*items));
 
    for (i = 0; i < cache->size; i++)
       for (c = cache->items[i]; c; c = next) {
@@ -194,7 +196,7 @@ brw_search_cache(struct brw_cache *cache,
 
    update_cache_last(cache, cache_id, item->bo);
 
-   brw->sws->bo_reference(item->bo);
+   cache->sws->bo_reference(item->bo);
    return item->bo;
 }
 
@@ -219,20 +221,25 @@ brw_upload_cache( struct brw_cache *cache,
    struct brw_winsys_buffer *bo;
    int i;
 
-   /* Create the buffer object to contain the data */
-   bo = brw->sws->bo_alloc(cache->sws,
-			   cache->buffer_type[cache_id], data_size, 1 << 6);
+   /* Create the buffer object to contain the data.  For now, use a
+    * single buffer type to describe all cached state atoms.  Later,
+    * may want to take advantage of hardware distinctions between
+    * these various entities.
+    */
+   bo = cache->sws->bo_alloc(cache->sws,
+			     BRW_BUFFER_TYPE_STATE_CACHE, 
+			     data_size, 1 << 6);
 
 
    /* Set up the memory containing the key, aux_data, and reloc_bufs */
-   tmp = _mesa_malloc(key_size + aux_size + relocs_size);
+   tmp = MALLOC(key_size + aux_size + relocs_size);
 
    memcpy(tmp, key, key_size);
    memcpy(tmp + key_size, aux, cache->aux_size[cache_id]);
    memcpy(tmp + key_size + aux_size, reloc_bufs, relocs_size);
    for (i = 0; i < nr_reloc_bufs; i++) {
       if (reloc_bufs[i] != NULL)
-	 brw->sws->bo_reference(reloc_bufs[i]);
+	 cache->sws->bo_reference(reloc_bufs[i]);
    }
 
    item->cache_id = cache_id;
@@ -243,7 +250,7 @@ brw_upload_cache( struct brw_cache *cache,
    item->nr_reloc_bufs = nr_reloc_bufs;
 
    item->bo = bo;
-   brw->sws->bo_reference(bo);
+   cache->sws->bo_reference(bo);
    item->data_size = data_size;
 
    if (cache->n_items > cache->size * 1.5)
@@ -259,13 +266,13 @@ brw_upload_cache( struct brw_cache *cache,
       *(void **)aux_return = (void *)((char *)item->key + item->key_size);
    }
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      _mesa_printf("upload %s: %d bytes to cache id %d\n",
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("upload %s: %d bytes to cache id %d\n",
 		   cache->name[cache_id],
 		   data_size, cache_id);
 
    /* Copy data to the buffer */
-   dri_bo_subdata(bo, 0, data_size, data);
+   cache->sws->bo_subdata(bo, 0, data_size, data);
 
    update_cache_last(cache, cache_id, bo);
 
@@ -292,7 +299,7 @@ brw_cache_data_sz(struct brw_cache *cache,
 		       reloc_bufs, nr_reloc_bufs);
    if (item) {
       update_cache_last(cache, cache_id, item->bo);
-      brw->sws->bo_reference(item->bo);
+      cache->sws->bo_reference(item->bo);
       return item->bo;
    }
 
@@ -349,11 +356,12 @@ brw_init_non_surface_cache(struct brw_context *brw)
    struct brw_cache *cache = &brw->cache;
 
    cache->brw = brw;
+   cache->sws = brw->sws;
 
    cache->size = 7;
    cache->n_items = 0;
    cache->items = (struct brw_cache_item **)
-      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
+      CALLOC(cache->size, sizeof(struct brw_cache_item));
 
    brw_init_cache_id(cache,
 		     "CC_VP",
@@ -457,7 +465,7 @@ brw_init_surface_cache(struct brw_context *brw)
    cache->size = 7;
    cache->n_items = 0;
    cache->items = (struct brw_cache_item **)
-      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
+      CALLOC(cache->size, sizeof(struct brw_cache_item));
 
    brw_init_cache_id(cache,
 		     "SS_SURFACE",
@@ -487,8 +495,8 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
    struct brw_cache_item *c, *next;
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      _mesa_printf("%s\n", __FUNCTION__);
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("%s\n", __FUNCTION__);
 
    for (i = 0; i < cache->size; i++) {
       for (c = cache->items[i]; c; c = next) {
@@ -507,7 +515,7 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
    cache->n_items = 0;
 
    if (brw->curbe.last_buf) {
-      _mesa_free(brw->curbe.last_buf);
+      FREE(brw->curbe.last_buf);
       brw->curbe.last_buf = NULL;
    }
 
@@ -527,8 +535,8 @@ brw_state_cache_bo_delete(struct brw_cache *cache, struct brw_winsys_buffer *bo)
    struct brw_cache_item **prev;
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      _mesa_printf("%s\n", __FUNCTION__);
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("%s\n", __FUNCTION__);
 
    for (i = 0; i < cache->size; i++) {
       for (prev = &cache->items[i]; *prev;) {
@@ -540,8 +548,8 @@ brw_state_cache_bo_delete(struct brw_cache *cache, struct brw_winsys_buffer *bo)
 	    *prev = c->next;
 
 	    for (j = 0; j < c->nr_reloc_bufs; j++)
-	       brw->sws->bo_unreference(c->reloc_bufs[j]);
-	    brw->sws->bo_unreference(c->bo);
+	       cache->sws->bo_unreference(c->reloc_bufs[j]);
+	    cache->sws->bo_unreference(c->bo);
 	    free((void *)c->key);
 	    free(c);
 	    cache->n_items--;
@@ -555,8 +563,8 @@ brw_state_cache_bo_delete(struct brw_cache *cache, struct brw_winsys_buffer *bo)
 void
 brw_state_cache_check_size(struct brw_context *brw)
 {
-   if (INTEL_DEBUG & DEBUG_STATE)
-      _mesa_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
 
    /* un-tuned guess.  We've got around 20 state objects for a total of around
     * 32k, so 1000 of them is around 1.5MB.
@@ -574,8 +582,8 @@ brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      _mesa_printf("%s\n", __FUNCTION__);
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("%s\n", __FUNCTION__);
 
    brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
diff --git a/src/gallium/drivers/i965/brw_state_debug.c b/src/gallium/drivers/i965/brw_state_debug.c
index 22cea4b7d85..cc4744dc166 100644
--- a/src/gallium/drivers/i965/brw_state_debug.c
+++ b/src/gallium/drivers/i965/brw_state_debug.c
@@ -109,8 +109,25 @@ brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
       if (bit_map[i].bit == 0)
 	 return;
 
-      fprintf(stderr, "0x%08x: %12d (%s)\n",
+      debug_printf("0x%08x: %12d (%s)\n",
 	      bit_map[i].bit, bit_map[i].count, bit_map[i].name);
    }
 }
 
+void
+brw_update_dirty_counts( unsigned mesa,
+			 unsigned brw,
+			 unsigned cache )
+{
+   static int dirty_count = 0;
+
+   brw_update_dirty_count(mesa_bits, mesa);
+   brw_update_dirty_count(brw_bits, brw);
+   brw_update_dirty_count(cache_bits, cache);
+      if (dirty_count++ % 1000 == 0) {
+	 brw_print_dirty_count(mesa_bits, mesa);
+	 brw_print_dirty_count(brw_bits, brw);
+	 brw_print_dirty_count(cache_bits, cache);
+	 debug_printf("\n");
+      }
+}
diff --git a/src/gallium/drivers/i965/brw_state_dump.c b/src/gallium/drivers/i965/brw_state_dump.c
index 1bc83fb9c15..72604304d41 100644
--- a/src/gallium/drivers/i965/brw_state_dump.c
+++ b/src/gallium/drivers/i965/brw_state_dump.c
@@ -28,6 +28,7 @@
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
+#include "brw_winsys.h"
 
 /**
  * Prints out a header, the contents, and the message associated with
@@ -44,28 +45,32 @@ state_out(const char *name, void *data, uint32_t hw_offset, int index,
 {
     va_list va;
 
-    fprintf(stderr, "%8s: 0x%08x: 0x%08x: ",
-	    name, hw_offset + index * 4, ((uint32_t *)data)[index]);
+    debug_printf("%8s: 0x%08x: 0x%08x: ",
+		 name, hw_offset + index * 4, ((uint32_t *)data)[index]);
     va_start(va, fmt);
-    vfprintf(stderr, fmt, va);
+    debug_vprintf(fmt, va);
     va_end(va);
 }
 
 /** Generic, undecoded state buffer debug printout */
 static void
-state_struct_out(const char *name, struct brw_winsys_buffer *buffer, unsigned int state_size)
+state_struct_out(struct brw_winsys_screen *sws,
+		 const char *name,
+		 struct brw_winsys_buffer *buffer,
+		 unsigned int state_size)
 {
    int i;
+   void *data;
 
    if (buffer == NULL)
       return;
 
-   dri_bo_map(buffer, GL_FALSE);
+   data = sws->bo_map(buffer, GL_FALSE);
    for (i = 0; i < state_size / 4; i++) {
-      state_out(name, buffer->virtual, buffer->offset, i,
+      state_out(name, data, buffer->offset, i,
 		"dword %d\n", i);
    }
-   dri_bo_unmap(buffer);
+   sws->bo_unmap(buffer);
 }
 
 static const char *
@@ -106,12 +111,11 @@ static void dump_wm_surface_state(struct brw_context *brw)
       char name[20];
 
       if (surf_bo == NULL) {
-	 fprintf(stderr, "  WM SS%d: NULL\n", i);
+	 debug_printf("  WM SS%d: NULL\n", i);
 	 continue;
       }
-      dri_bo_map(surf_bo, GL_FALSE);
+      surf = (struct brw_surface_state *)brw->sws->bo_map(surf_bo, GL_FALSE);
       surfoff = surf_bo->offset;
-      surf = (struct brw_surface_state *)(surf_bo->virtual);
 
       sprintf(name, "WM SS%d", i);
       state_out(name, surf, surfoff, 0, "%s %s\n",
@@ -127,7 +131,7 @@ static void dump_wm_surface_state(struct brw_context *brw)
       state_out(name, surf, surfoff, 5, "x,y offset: %d,%d\n",
 		surf->ss5.x_offset, surf->ss5.y_offset);
 
-      dri_bo_unmap(surf_bo);
+      brw->sws->bo_unmap(surf_bo);
    }
 }
 
@@ -140,9 +144,7 @@ static void dump_sf_viewport_state(struct brw_context *brw)
    if (brw->sf.vp_bo == NULL)
       return;
 
-   dri_bo_map(brw->sf.vp_bo, GL_FALSE);
-
-   vp = brw->sf.vp_bo->virtual;
+   vp = (struct brw_sf_viewport *)brw->sws->bo_map(brw->sf.vp_bo, GL_FALSE);
    vp_off = brw->sf.vp_bo->offset;
 
    state_out(name, vp, vp_off, 0, "m00 = %f\n", vp->viewport.m00);
@@ -157,10 +159,12 @@ static void dump_sf_viewport_state(struct brw_context *brw)
    state_out(name, vp, vp_off, 7, "bottom right = %d,%d\n",
 	     vp->scissor.xmax, vp->scissor.ymax);
 
-   dri_bo_unmap(brw->sf.vp_bo);
+   brw->sws->bo_unmap(brw->sf.vp_bo);
 }
 
-static void brw_debug_prog(const char *name, struct brw_winsys_buffer *prog)
+static void brw_debug_prog(struct brw_winsys_screen *sws,
+			   const char *name,
+			   struct brw_winsys_buffer *prog)
 {
    unsigned int i;
    uint32_t *data;
@@ -168,12 +172,10 @@ static void brw_debug_prog(const char *name, struct brw_winsys_buffer *prog)
    if (prog == NULL)
       return;
 
-   dri_bo_map(prog, GL_FALSE);
-
-   data = prog->virtual;
+   data = (uint32_t *)sws->bo_map(prog, GL_FALSE);
 
    for (i = 0; i < prog->size / 4 / 4; i++) {
-      fprintf(stderr, "%8s: 0x%08x: 0x%08x 0x%08x 0x%08x 0x%08x\n",
+      debug_printf("%8s: 0x%08x: 0x%08x 0x%08x 0x%08x 0x%08x\n",
 	      name, (unsigned int)prog->offset + i * 4 * 4,
 	      data[i * 4], data[i * 4 + 1], data[i * 4 + 2], data[i * 4 + 3]);
       /* Stop at the end of the program.  It'd be nice to keep track of the actual
@@ -186,7 +188,7 @@ static void brw_debug_prog(const char *name, struct brw_winsys_buffer *prog)
 	 break;
    }
 
-   dri_bo_unmap(prog);
+   sws->bo_unmap(prog);
 }
 
 
@@ -202,19 +204,21 @@ static void brw_debug_prog(const char *name, struct brw_winsys_buffer *prog)
  */
 void brw_debug_batch(struct brw_context *brw)
 {
-   state_struct_out("WM bind", brw->wm.bind_bo, 4 * brw->wm.nr_surfaces);
+   struct brw_winsys_screen *sws = brw->sws;
+
+   state_struct_out(sws, "WM bind", brw->wm.bind_bo, 4 * brw->wm.nr_surfaces);
    dump_wm_surface_state(brw);
 
-   state_struct_out("VS", brw->vs.state_bo, sizeof(struct brw_vs_unit_state));
-   brw_debug_prog("VS prog", brw->vs.prog_bo);
+   state_struct_out(sws, "VS", brw->vs.state_bo, sizeof(struct brw_vs_unit_state));
+   brw_debug_prog(sws, "VS prog", brw->vs.prog_bo);
 
-   state_struct_out("GS", brw->gs.state_bo, sizeof(struct brw_gs_unit_state));
-   brw_debug_prog("GS prog", brw->gs.prog_bo);
+   state_struct_out(sws, "GS", brw->gs.state_bo, sizeof(struct brw_gs_unit_state));
+   brw_debug_prog(sws, "GS prog", brw->gs.prog_bo);
 
-   state_struct_out("SF", brw->sf.state_bo, sizeof(struct brw_sf_unit_state));
+   state_struct_out(sws, "SF", brw->sf.state_bo, sizeof(struct brw_sf_unit_state));
    dump_sf_viewport_state(brw);
-   brw_debug_prog("SF prog", brw->sf.prog_bo);
+   brw_debug_prog(sws, "SF prog", brw->sf.prog_bo);
 
-   state_struct_out("WM", brw->wm.state_bo, sizeof(struct brw_wm_unit_state));
-   brw_debug_prog("WM prog", brw->wm.prog_bo);
+   state_struct_out(sws, "WM", brw->wm.state_bo, sizeof(struct brw_wm_unit_state));
+   brw_debug_prog(sws, "WM prog", brw->wm.prog_bo);
 }
diff --git a/src/gallium/drivers/i965/brw_state_upload.c b/src/gallium/drivers/i965/brw_state_upload.c
index 8659e35289f..eff3a40a464 100644
--- a/src/gallium/drivers/i965/brw_state_upload.c
+++ b/src/gallium/drivers/i965/brw_state_upload.c
@@ -34,6 +34,7 @@
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_batchbuffer.h"
+#include "brw_debug.h"
 
 /* This is used to initialize brw->state.atoms[].  We could use this
  * list directly except for a single atom, brw_constant_buffer, which
@@ -83,12 +84,8 @@ const struct brw_tracked_state *atoms[] =
    &brw_blend_constant_color,
 
    &brw_depthbuffer,
-
    &brw_polygon_stipple,
-   &brw_polygon_stipple_offset,
-
    &brw_line_stipple,
-   &brw_aa_line_parameters,
 
    &brw_psp_urb_cbs,
 
@@ -163,11 +160,12 @@ enum pipe_error brw_validate_state( struct brw_context *brw )
 {
    struct brw_state_flags *state = &brw->state.dirty;
    GLuint i;
+   int ret;
 
    brw_clear_validated_bos(brw);
-   brw_add_validated_bo(brw, intel->batch->buf);
+   brw_add_validated_bo(brw, brw->batch->buf);
 
-   if (brw->emit_state_always) {
+   if (brw->flags.always_emit_state) {
       state->mesa |= ~0;
       state->brw |= ~0;
       state->cache |= ~0;
@@ -199,10 +197,10 @@ enum pipe_error brw_validate_state( struct brw_context *brw )
     * If this fails, we can experience GPU lock-ups.
     */
    {
-      const struct brw_fragment_program *fp = brw->fragment_program;
+      const struct brw_fragment_shader *fp = brw->curr.fragment_shader;
       if (fp) {
-         assert(fp->info.max_sampler <= brw->nr_samplers &&
-		fp->info.max_texture <= brw->nr_textures);
+         assert(fp->info.file_max[TGSI_FILE_SAMPLER] < brw->curr.num_samplers &&
+		fp->info.texture_max < brw->curr.num_textures);
       }
    }
 
@@ -213,18 +211,18 @@ enum pipe_error brw_validate_state( struct brw_context *brw )
 enum pipe_error brw_upload_state(struct brw_context *brw)
 {
    struct brw_state_flags *state = &brw->state.dirty;
+   int ret;
    int i;
-   static int dirty_count = 0;
 
    brw_clear_validated_bos(brw);
 
-   if (INTEL_DEBUG) {
+   if (BRW_DEBUG) {
       /* Debug version which enforces various sanity checks on the
        * state flags which are generated and checked to help ensure
        * state atoms are ordered correctly in the list.
        */
       struct brw_state_flags examined, prev;      
-      _mesa_memset(&examined, 0, sizeof(examined));
+      memset(&examined, 0, sizeof(examined));
       prev = *state;
 
       for (i = 0; i < Elements(atoms); i++) {
@@ -268,19 +266,14 @@ enum pipe_error brw_upload_state(struct brw_context *brw)
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_STATE) {
-      brw_update_dirty_count(mesa_bits, state->mesa);
-      brw_update_dirty_count(brw_bits, state->brw);
-      brw_update_dirty_count(cache_bits, state->cache);
-      if (dirty_count++ % 1000 == 0) {
-	 brw_print_dirty_count(mesa_bits, state->mesa);
-	 brw_print_dirty_count(brw_bits, state->brw);
-	 brw_print_dirty_count(cache_bits, state->cache);
-	 debug_printf("\n");
-      }
+   if (BRW_DEBUG & DEBUG_STATE) {
+      brw_update_dirty_counts( state->mesa, 
+			       state->brw,
+			       state->cache );
    }
    
    /* Clear dirty flags:
     */
    memset(state, 0, sizeof(*state));
+   return 0;
 }
diff --git a/src/gallium/drivers/i965/brw_tex.c b/src/gallium/drivers/i965/brw_tex.c
deleted file mode 100644
index 6f7adb6393c..00000000000
--- a/src/gallium/drivers/i965/brw_tex.c
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- Copyright (C) Intel Corp.  2006.  All Rights Reserved.
- Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
- develop this 3D driver.
- 
- 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, sublicense, 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 NONINFRINGEMENT.
- IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- 
- **********************************************************************/
- /*
-  * Authors:
-  *   Keith Whitwell <[email protected]>
-  */
-        
-
-#include "brw_context.h"
-
-/**
- * Finalizes all textures, completing any rendering that needs to be done
- * to prepare them.
- */
-void brw_validate_textures( struct brw_context *brw )
-{
-   int i;
-
-   for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
-      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
-
-      if (texUnit->_ReallyEnabled) {
-	 intel_finalize_mipmap_tree(intel, i);
-      }
-   }
-}
diff --git a/src/gallium/drivers/i965/brw_urb.c b/src/gallium/drivers/i965/brw_urb.c
index a2277519ad7..ff2466528df 100644
--- a/src/gallium/drivers/i965/brw_urb.c
+++ b/src/gallium/drivers/i965/brw_urb.c
@@ -184,17 +184,17 @@ static void recalculate_urb_fence( struct brw_context *brw )
 	     * entries and the values for minimum nr of entries
 	     * provided above.
 	     */
-	    _mesa_printf("couldn't calculate URB layout!\n");
+	    debug_printf("couldn't calculate URB layout!\n");
 	    exit(1);
 	 }
 	 
-	 if (INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
-	    _mesa_printf("URB CONSTRAINED\n");
+	 if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
+	    debug_printf("URB CONSTRAINED\n");
       }
 
 done:
-      if (INTEL_DEBUG & DEBUG_URB)
-	 _mesa_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
+      if (BRW_DEBUG & DEBUG_URB)
+	 debug_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
 		      brw->urb.vs_start,
 		      brw->urb.gs_start,
 		      brw->urb.clip_start,
diff --git a/src/gallium/drivers/i965/brw_vs.h b/src/gallium/drivers/i965/brw_vs.h
index 54f7d7d7c49..e33fa2f0aad 100644
--- a/src/gallium/drivers/i965/brw_vs.h
+++ b/src/gallium/drivers/i965/brw_vs.h
@@ -64,7 +64,7 @@ struct brw_vs_compile {
 
    struct brw_reg r0;
    struct brw_reg r1;
-   struct brw_reg regs[PROGRAM_ADDRESS+1][128];
+   struct brw_reg regs[TGSI_FILE_COUNT][128];
    struct brw_reg tmp;
    struct brw_reg stack;
 
diff --git a/src/gallium/drivers/i965/brw_vs_emit.c b/src/gallium/drivers/i965/brw_vs_emit.c
index 086f54799e6..04132a167bc 100644
--- a/src/gallium/drivers/i965/brw_vs_emit.c
+++ b/src/gallium/drivers/i965/brw_vs_emit.c
@@ -242,10 +242,10 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 
    c->prog_data.total_grf = reg;
 
-   if (INTEL_DEBUG & DEBUG_VS) {
-      _mesa_printf("%s NumAddrRegs %d\n", __FUNCTION__, c->vp->program.Base.NumAddressRegs);
-      _mesa_printf("%s NumTemps %d\n", __FUNCTION__, c->vp->program.Base.NumTemporaries);
-      _mesa_printf("%s reg = %d\n", __FUNCTION__, reg);
+   if (BRW_DEBUG & DEBUG_VS) {
+      debug_printf("%s NumAddrRegs %d\n", __FUNCTION__, c->vp->program.Base.NumAddressRegs);
+      debug_printf("%s NumTemps %d\n", __FUNCTION__, c->vp->program.Base.NumTemporaries);
+      debug_printf("%s reg = %d\n", __FUNCTION__, reg);
    }
 }
 
@@ -1248,10 +1248,10 @@ void brw_vs_emit(struct brw_vs_compile *c )
    GLuint index;
    GLuint file;
 
-   if (INTEL_DEBUG & DEBUG_VS) {
-      _mesa_printf("vs-mesa:\n");
+   if (BRW_DEBUG & DEBUG_VS) {
+      debug_printf("vs-mesa:\n");
       _mesa_print_program(&c->vp->program.Base); 
-      _mesa_printf("\n");
+      debug_printf("\n");
    }
 
    brw_set_compression_control(p, BRW_COMPRESSION_NONE);
@@ -1526,12 +1526,12 @@ void brw_vs_emit(struct brw_vs_compile *c )
 
    post_vs_emit(c, end_inst, last_inst);
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (BRW_DEBUG & DEBUG_VS) {
       int i;
 
-      _mesa_printf("vs-native:\n");
+      debug_printf("vs-native:\n");
       for (i = 0; i < p->nr_insn; i++)
 	 brw_disasm(stderr, &p->store[i]);
-      _mesa_printf("\n");
+      debug_printf("\n");
    }
 }
diff --git a/src/gallium/drivers/i965/brw_vs_state.c b/src/gallium/drivers/i965/brw_vs_state.c
index 1717223e49c..05a91f2de44 100644
--- a/src/gallium/drivers/i965/brw_vs_state.c
+++ b/src/gallium/drivers/i965/brw_vs_state.c
@@ -122,7 +122,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
    vs.thread4.max_threads = CLAMP(key->nr_urb_entries / 2,
 				  1, chipset_max_threads) - 1;
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (BRW_DEBUG & DEBUG_SINGLE_THREAD)
       vs.thread4.max_threads = 0;
 
    /* No samplers for ARB_vp programs:
@@ -131,7 +131,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
     */
    vs.vs5.sampler_count = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (BRW_DEBUG & DEBUG_STATS)
       vs.thread4.stats_enable = 1;
 
    /* Vertex program always enabled:
diff --git a/src/gallium/drivers/i965/brw_winsys.h b/src/gallium/drivers/i965/brw_winsys.h
index 51e23b96404..33032276bc1 100644
--- a/src/gallium/drivers/i965/brw_winsys.h
+++ b/src/gallium/drivers/i965/brw_winsys.h
@@ -69,6 +69,7 @@ enum brw_buffer_type
    BRW_BUFFER_TYPE_SHADER_CONSTANTS,
    BRW_BUFFER_TYPE_WM_SCRATCH,
    BRW_BUFFER_TYPE_BATCH,
+   BRW_BUFFER_TYPE_STATE_CACHE,
 };
 
 
@@ -156,11 +157,15 @@ struct brw_winsys_screen {
 			  unsigned offset,
 			  struct brw_winsys_buffer *b2);
 
-   void (*bo_subdata)(struct brw_winsys_buffer *dst,
+   void (*bo_subdata)(struct brw_winsys_buffer *buffer,
 		      size_t offset,
 		      size_t size,
 		      const void *data);
 
+   boolean (*bo_is_busy)(struct brw_winsys_buffer *buffer);
+   boolean (*bo_references)(struct brw_winsys_buffer *a,
+			    struct brw_winsys_buffer *b);
+
    /* XXX: couldn't this be handled by returning true/false on
     * bo_emit_reloc?
     */
@@ -171,18 +176,13 @@ struct brw_winsys_screen {
    /**
     * Map a buffer.
     */
-   void *(*buffer_map)(struct brw_winsys *iws,
-                       struct brw_winsys_buffer *buffer,
-                       boolean write);
+   void *(*bo_map)(struct brw_winsys_buffer *buffer,
+		   boolean write);
 
    /**
     * Unmap a buffer.
     */
-   void (*buffer_unmap)(struct brw_winsys *iws,
-                        struct brw_winsys_buffer *buffer);
-
-   void (*buffer_destroy)(struct brw_winsys *iws,
-                          struct brw_winsys_buffer *buffer);
+   void (*bo_unmap)(struct brw_winsys_buffer *buffer);
    /*@}*/
 
 
diff --git a/src/gallium/drivers/i965/brw_wm.c b/src/gallium/drivers/i965/brw_wm.c
index 764708f7df9..3d889699f89 100644
--- a/src/gallium/drivers/i965/brw_wm.c
+++ b/src/gallium/drivers/i965/brw_wm.c
@@ -178,8 +178,8 @@ static void do_wm_prog( struct brw_context *brw,
       brw_wm_non_glsl_emit(brw, c);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM)
-      fprintf(stderr, "\n");
+   if (BRW_DEBUG & DEBUG_WM)
+      debug_printf("\n");
 
    /* get the program
     */
diff --git a/src/gallium/drivers/i965/brw_wm.h b/src/gallium/drivers/i965/brw_wm.h
index bf241f5fa4d..5bc2a49c1f4 100644
--- a/src/gallium/drivers/i965/brw_wm.h
+++ b/src/gallium/drivers/i965/brw_wm.h
@@ -33,6 +33,7 @@
 #ifndef BRW_WM_H
 #define BRW_WM_H
 
+#include "tgsi/tgsi_ureg.h"
 
 #include "brw_context.h"
 #include "brw_eu.h"
@@ -57,17 +58,18 @@
 #define AA_ALWAYS    2
 
 struct brw_wm_prog_key {
+   unsigned proj_attrib_mask; /**< one bit per fragment program attribute */
+   unsigned linear_attrib_mask:1;  /**< linear interpolation vs perspective interp */
+
    GLuint source_depth_reg:3;
    GLuint aa_dest_stencil_reg:3;
    GLuint dest_depth_reg:3;
    GLuint nr_depth_regs:3;
-   GLuint computes_depth:1;	/* could be derived from program string */
+   GLuint computes_depth:1;
    GLuint source_depth_to_render_target:1;
    GLuint flat_shade:1;
-   GLuint linear_color:1;  /**< linear interpolation vs perspective interp */
    GLuint runtime_check_aads_emit:1;
-   
-   GLbitfield proj_attrib_mask; /**< one bit per fragment program attribute */
+
    GLuint shadowtex_mask:16;
    GLuint yuvtex_mask:16;
    GLuint yuvtex_swap_mask:16;	/* UV swaped */
@@ -75,7 +77,7 @@ struct brw_wm_prog_key {
    GLuint tex_swizzles[BRW_MAX_TEX_UNIT];
 
    GLuint program_string_id:32;
-   GLuint drawable_height;
+
    GLuint vp_nr_outputs_written;
 };
 
@@ -151,7 +153,7 @@ struct brw_wm_instruction {
 };
 
 
-#define BRW_WM_MAX_INSN  (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + FRAG_ATTRIB_MAX + 3)
+#define BRW_WM_MAX_INSN  2048
 #define BRW_WM_MAX_GRF   128		/* hardware limit */
 #define BRW_WM_MAX_VREG  (BRW_WM_MAX_INSN * 4)
 #define BRW_WM_MAX_REF   (BRW_WM_MAX_INSN * 12)
@@ -161,11 +163,19 @@ struct brw_wm_instruction {
 #define BRW_WM_MAX_SUBROUTINE 16
 
 
+struct ureg_instruction {
+   unsigned opcode:8;
+   unsigned tex_target:3;
+   struct ureg_dst dst;
+   struct ureg_src src[3];
+};
+
 
 /* New opcodes to track internal operations required for WM unit.
  * These are added early so that the registers used can be tracked,
  * freed and reused like those of other instructions.
  */
+#define MAX_OPCODE        TGSI_OPCODE_LAST
 #define WM_PIXELXY        (MAX_OPCODE)
 #define WM_DELTAXY        (MAX_OPCODE + 1)
 #define WM_PIXELW         (MAX_OPCODE + 2)
@@ -177,7 +187,7 @@ struct brw_wm_instruction {
 #define WM_FRONTFACING    (MAX_OPCODE + 8)
 #define MAX_WM_OPCODE     (MAX_OPCODE + 9)
 
-#define PROGRAM_PAYLOAD   (PROGRAM_FILE_MAX)
+#define PROGRAM_PAYLOAD   (TGSI_FILE_COUNT)
 #define PAYLOAD_DEPTH     (FRAG_ATTRIB_MAX)
 
 struct brw_wm_compile {
@@ -198,15 +208,15 @@ struct brw_wm_compile {
     * simplifying and adding instructions for interpolation and
     * framebuffer writes.
     */
-   struct prog_instruction prog_instructions[BRW_WM_MAX_INSN];
+   struct ureg_instruction prog_instructions[BRW_WM_MAX_INSN];
    GLuint nr_fp_insns;
    GLuint fp_temp;
    GLuint fp_interp_emitted;
    GLuint fp_fragcolor_emitted;
 
-   struct prog_src_register pixel_xy;
-   struct prog_src_register delta_xy;
-   struct prog_src_register pixel_w;
+   struct ureg_src pixel_xy;
+   struct ureg_src delta_xy;
+   struct ureg_src pixel_w;
 
 
    struct brw_wm_value vreg[BRW_WM_MAX_VREG];
@@ -217,7 +227,7 @@ struct brw_wm_compile {
 
    struct {
       struct brw_wm_value depth[4]; /* includes r0/r1 */
-      struct brw_wm_value input_interp[FRAG_ATTRIB_MAX];
+      struct brw_wm_value input_interp[PIPE_MAX_SHADER_INPUTS];
    } payload;
 
 
@@ -295,7 +305,7 @@ void brw_wm_lookup_iz( GLuint line_aa,
 		       GLboolean ps_uses_depth,
 		       struct brw_wm_prog_key *key );
 
-GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp);
+//GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp);
 void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c);
 
 void emit_ddxy(struct brw_compile *p,
diff --git a/src/gallium/drivers/i965/brw_wm_debug.c b/src/gallium/drivers/i965/brw_wm_debug.c
index c6659646f2c..04dec5ba392 100644
--- a/src/gallium/drivers/i965/brw_wm_debug.c
+++ b/src/gallium/drivers/i965/brw_wm_debug.c
@@ -41,21 +41,21 @@ void brw_wm_print_value( struct brw_wm_compile *c,
    if (c->state >= PASS2_DONE) 
       brw_print_reg(value->hw_reg);
    else if( value == &c->undef_value )
-      _mesa_printf("undef");
+      debug_printf("undef");
    else if( value - c->vreg >= 0 &&
 	    value - c->vreg < BRW_WM_MAX_VREG)
-      _mesa_printf("r%d", value - c->vreg);
+      debug_printf("r%d", value - c->vreg);
    else if (value - c->creg >= 0 &&
 	    value - c->creg < BRW_WM_MAX_PARAM)
-      _mesa_printf("c%d", value - c->creg);
+      debug_printf("c%d", value - c->creg);
    else if (value - c->payload.input_interp >= 0 &&
 	    value - c->payload.input_interp < FRAG_ATTRIB_MAX)
-      _mesa_printf("i%d", value - c->payload.input_interp);
+      debug_printf("i%d", value - c->payload.input_interp);
    else if (value - c->payload.depth >= 0 &&
 	    value - c->payload.depth < FRAG_ATTRIB_MAX)
-      _mesa_printf("d%d", value - c->payload.depth);
+      debug_printf("d%d", value - c->payload.depth);
    else 
-      _mesa_printf("?");
+      debug_printf("?");
 }
 
 void brw_wm_print_ref( struct brw_wm_compile *c,
@@ -64,16 +64,16 @@ void brw_wm_print_ref( struct brw_wm_compile *c,
    struct brw_reg hw_reg = ref->hw_reg;
 
    if (ref->unspill_reg)
-      _mesa_printf("UNSPILL(%x)/", ref->value->spill_slot);
+      debug_printf("UNSPILL(%x)/", ref->value->spill_slot);
 
    if (c->state >= PASS2_DONE)
       brw_print_reg(ref->hw_reg);
    else {
-      _mesa_printf("%s", hw_reg.negate ? "-" : "");
-      _mesa_printf("%s", hw_reg.abs ? "abs/" : "");
+      debug_printf("%s", hw_reg.negate ? "-" : "");
+      debug_printf("%s", hw_reg.abs ? "abs/" : "");
       brw_wm_print_value(c, ref->value);
       if ((hw_reg.nr&1) || hw_reg.subnr) {
-	 _mesa_printf("->%d.%d", (hw_reg.nr&1), hw_reg.subnr);
+	 debug_printf("->%d.%d", (hw_reg.nr&1), hw_reg.subnr);
       }
    }
 }
@@ -84,22 +84,22 @@ void brw_wm_print_insn( struct brw_wm_compile *c,
    GLuint i, arg;
    GLuint nr_args = brw_wm_nr_args(inst->opcode);
 
-   _mesa_printf("[");
+   debug_printf("[");
    for (i = 0; i < 4; i++) {
       if (inst->dst[i]) {
 	 brw_wm_print_value(c, inst->dst[i]);
 	 if (inst->dst[i]->spill_slot)
-	    _mesa_printf("/SPILL(%x)",inst->dst[i]->spill_slot);
+	    debug_printf("/SPILL(%x)",inst->dst[i]->spill_slot);
       }
       else
-	 _mesa_printf("#");
+	 debug_printf("#");
       if (i < 3)      
-	 _mesa_printf(",");
+	 debug_printf(",");
    }
-   _mesa_printf("]");
+   debug_printf("]");
 
    if (inst->writemask != BRW_WRITEMASK_XYZW)
-      _mesa_printf(".%s%s%s%s", 
+      debug_printf(".%s%s%s%s", 
 		   GET_BIT(inst->writemask, 0) ? "x" : "",
 		   GET_BIT(inst->writemask, 1) ? "y" : "",
 		   GET_BIT(inst->writemask, 2) ? "z" : "",
@@ -107,58 +107,58 @@ void brw_wm_print_insn( struct brw_wm_compile *c,
 
    switch (inst->opcode) {
    case WM_PIXELXY:
-      _mesa_printf(" = PIXELXY");
+      debug_printf(" = PIXELXY");
       break;
    case WM_DELTAXY:
-      _mesa_printf(" = DELTAXY");
+      debug_printf(" = DELTAXY");
       break;
    case WM_PIXELW:
-      _mesa_printf(" = PIXELW");
+      debug_printf(" = PIXELW");
       break;
    case WM_WPOSXY:
-      _mesa_printf(" = WPOSXY");
+      debug_printf(" = WPOSXY");
       break;
    case WM_PINTERP:
-      _mesa_printf(" = PINTERP");
+      debug_printf(" = PINTERP");
       break;
    case WM_LINTERP:
-      _mesa_printf(" = LINTERP");
+      debug_printf(" = LINTERP");
       break;
    case WM_CINTERP:
-      _mesa_printf(" = CINTERP");
+      debug_printf(" = CINTERP");
       break;
    case WM_FB_WRITE:
-      _mesa_printf(" = FB_WRITE");
+      debug_printf(" = FB_WRITE");
       break;
    case WM_FRONTFACING:
-      _mesa_printf(" = FRONTFACING");
+      debug_printf(" = FRONTFACING");
       break;
    default:
-      _mesa_printf(" = %s", _mesa_opcode_string(inst->opcode));
+      debug_printf(" = %s", _mesa_opcode_string(inst->opcode));
       break;
    }
 
    if (inst->saturate)
-      _mesa_printf("_SAT");
+      debug_printf("_SAT");
 
    for (arg = 0; arg < nr_args; arg++) {
 
-      _mesa_printf(" [");
+      debug_printf(" [");
 
       for (i = 0; i < 4; i++) {
 	 if (inst->src[arg][i]) {
 	    brw_wm_print_ref(c, inst->src[arg][i]);
 	 }
 	 else
-	    _mesa_printf("%%");
+	    debug_printf("%%");
 
 	 if (i < 3) 
-	    _mesa_printf(",");
+	    debug_printf(",");
 	 else
-	    _mesa_printf("]");
+	    debug_printf("]");
       }
    }
-   _mesa_printf("\n");
+   debug_printf("\n");
 }
 
 void brw_wm_print_program( struct brw_wm_compile *c,
@@ -166,9 +166,9 @@ void brw_wm_print_program( struct brw_wm_compile *c,
 {
    GLuint insn;
 
-   _mesa_printf("%s:\n", stage);
+   debug_printf("%s:\n", stage);
    for (insn = 0; insn < c->nr_insns; insn++)
       brw_wm_print_insn(c, &c->instruction[insn]);
-   _mesa_printf("\n");
+   debug_printf("\n");
 }
 
diff --git a/src/gallium/drivers/i965/brw_wm_emit.c b/src/gallium/drivers/i965/brw_wm_emit.c
index 7df9b79d7a2..5f7ae6592c6 100644
--- a/src/gallium/drivers/i965/brw_wm_emit.c
+++ b/src/gallium/drivers/i965/brw_wm_emit.c
@@ -1481,7 +1481,7 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 break;
 
       default:
-	 _mesa_printf("Unsupported opcode %i (%s) in fragment shader\n",
+	 debug_printf("Unsupported opcode %i (%s) in fragment shader\n",
 		      inst->opcode, inst->opcode < MAX_OPCODE ?
 				    _mesa_opcode_string(inst->opcode) :
 				    "unknown");
@@ -1494,12 +1494,12 @@ void brw_wm_emit( struct brw_wm_compile *c )
 		      inst->dst[i]->spill_slot);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (BRW_DEBUG & DEBUG_WM) {
       int i;
 
-      _mesa_printf("wm-native:\n");
+      debug_printf("wm-native:\n");
       for (i = 0; i < p->nr_insn; i++)
 	 brw_disasm(stderr, &p->store[i]);
-      _mesa_printf("\n");
+      debug_printf("\n");
    }
 }
diff --git a/src/gallium/drivers/i965/brw_wm_fp.c b/src/gallium/drivers/i965/brw_wm_fp.c
index be240031c7c..d5947307301 100644
--- a/src/gallium/drivers/i965/brw_wm_fp.c
+++ b/src/gallium/drivers/i965/brw_wm_fp.c
@@ -142,7 +142,7 @@ static struct prog_dst_register get_temp( struct brw_wm_compile *c )
    int bit = _mesa_ffs( ~c->fp_temp );
 
    if (!bit) {
-      _mesa_printf("%s: out of temporaries\n", __FILE__);
+      debug_printf("%s: out of temporaries\n", __FILE__);
       exit(1);
    }
 
@@ -977,7 +977,7 @@ static void print_insns( const struct prog_instruction *insn,
 {
    GLuint i;
    for (i = 0; i < nr; i++, insn++) {
-      _mesa_printf("%3d: ", i);
+      debug_printf("%3d: ", i);
       if (insn->Opcode < MAX_OPCODE)
 	 _mesa_print_instruction(insn);
       else if (insn->Opcode < MAX_WM_OPCODE) {
@@ -988,7 +988,7 @@ static void print_insns( const struct prog_instruction *insn,
 				     3);
       }
       else 
-	 _mesa_printf("965 Opcode %d\n", insn->Opcode);
+	 debug_printf("965 Opcode %d\n", insn->Opcode);
    }
 }
 
@@ -1002,10 +1002,10 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
    struct brw_fragment_program *fp = c->fp;
    GLuint insn;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
-      _mesa_printf("pre-fp:\n");
+   if (BRW_DEBUG & DEBUG_WM) {
+      debug_printf("pre-fp:\n");
       _mesa_print_program(&fp->program.Base); 
-      _mesa_printf("\n");
+      debug_printf("\n");
    }
 
    c->pixel_xy = src_undef();
@@ -1103,10 +1103,10 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
-      _mesa_printf("pass_fp:\n");
+   if (BRW_DEBUG & DEBUG_WM) {
+      debug_printf("pass_fp:\n");
       print_insns( c->prog_instructions, c->nr_fp_insns );
-      _mesa_printf("\n");
+      debug_printf("\n");
    }
 }
 
diff --git a/src/gallium/drivers/i965/brw_wm_glsl.c b/src/gallium/drivers/i965/brw_wm_glsl.c
index a8de5fdd0b9..3118e615f91 100644
--- a/src/gallium/drivers/i965/brw_wm_glsl.c
+++ b/src/gallium/drivers/i965/brw_wm_glsl.c
@@ -1694,7 +1694,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
         c->cur_inst = i;
 
 #if 0
-        _mesa_printf("Inst %d: ", i);
+        debug_printf("Inst %d: ", i);
         _mesa_print_instruction(inst);
 #endif
 
@@ -1920,7 +1920,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
                }
                break;
 	    default:
-		_mesa_printf("unsupported IR in fragment shader %d\n",
+		debug_printf("unsupported IR in fragment shader %d\n",
 			inst->Opcode);
 	}
 
@@ -1931,11 +1931,11 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     }
     post_wm_emit(c);
 
-    if (INTEL_DEBUG & DEBUG_WM) {
-      _mesa_printf("wm-native:\n");
+    if (BRW_DEBUG & DEBUG_WM) {
+      debug_printf("wm-native:\n");
       for (i = 0; i < p->nr_insn; i++)
 	 brw_disasm(stderr, &p->store[i]);
-      _mesa_printf("\n");
+      debug_printf("\n");
     }
 }
 
@@ -1945,8 +1945,8 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
  */
 void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
 {
-    if (INTEL_DEBUG & DEBUG_WM) {
-        _mesa_printf("brw_wm_glsl_emit:\n");
+    if (BRW_DEBUG & DEBUG_WM) {
+        debug_printf("brw_wm_glsl_emit:\n");
     }
 
     /* initial instruction translation/simplification */
@@ -1955,7 +1955,7 @@ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
     /* actual code generation */
     brw_wm_emit_glsl(brw, c);
 
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (BRW_DEBUG & DEBUG_WM) {
         brw_wm_print_program(c, "brw_wm_glsl_emit done");
     }
 
diff --git a/src/gallium/drivers/i965/brw_wm_pass0.c b/src/gallium/drivers/i965/brw_wm_pass0.c
index 31b0270e843..71e4c568356 100644
--- a/src/gallium/drivers/i965/brw_wm_pass0.c
+++ b/src/gallium/drivers/i965/brw_wm_pass0.c
@@ -101,7 +101,7 @@ static const struct brw_wm_ref *get_param_ref( struct brw_wm_compile *c,
    GLuint i = c->prog_data.nr_params++;
    
    if (i >= BRW_WM_MAX_PARAM) {
-      _mesa_printf("%s: out of params\n", __FUNCTION__);
+      debug_printf("%s: out of params\n", __FUNCTION__);
       c->prog_data.error = 1;
       return NULL;
    }
@@ -150,7 +150,7 @@ static const struct brw_wm_ref *get_imm_ref( struct brw_wm_compile *c,
       return c->imm_ref[i].ref;
    }
    else {
-      _mesa_printf("%s: out of imm_refs\n", __FUNCTION__);
+      debug_printf("%s: out of imm_refs\n", __FUNCTION__);
       c->prog_data.error = 1;
       return NULL;
    }
@@ -434,7 +434,7 @@ void brw_wm_pass0( struct brw_wm_compile *c )
       }
    }
  
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (BRW_DEBUG & DEBUG_WM) {
       brw_wm_print_program(c, "pass0");
    }
 }
diff --git a/src/gallium/drivers/i965/brw_wm_pass1.c b/src/gallium/drivers/i965/brw_wm_pass1.c
index f2ae3a958f9..85a3a55ca4f 100644
--- a/src/gallium/drivers/i965/brw_wm_pass1.c
+++ b/src/gallium/drivers/i965/brw_wm_pass1.c
@@ -284,7 +284,7 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       track_arg(c, inst, 2, read2);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (BRW_DEBUG & DEBUG_WM) {
       brw_wm_print_program(c, "pass1");
    }
 }
diff --git a/src/gallium/drivers/i965/brw_wm_pass2.c b/src/gallium/drivers/i965/brw_wm_pass2.c
index 6faea018fbc..a19ca623287 100644
--- a/src/gallium/drivers/i965/brw_wm_pass2.c
+++ b/src/gallium/drivers/i965/brw_wm_pass2.c
@@ -331,13 +331,13 @@ void brw_wm_pass2( struct brw_wm_compile *c )
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (BRW_DEBUG & DEBUG_WM) {
       brw_wm_print_program(c, "pass2");
    }
 
    c->state = PASS2_DONE;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (BRW_DEBUG & DEBUG_WM) {
        brw_wm_print_program(c, "pass2/done");
    }
 }
diff --git a/src/gallium/drivers/i965/brw_wm_sampler_state.c b/src/gallium/drivers/i965/brw_wm_sampler_state.c
index a8993f93120..32692d533c4 100644
--- a/src/gallium/drivers/i965/brw_wm_sampler_state.c
+++ b/src/gallium/drivers/i965/brw_wm_sampler_state.c
@@ -76,8 +76,9 @@ static GLint S_FIXED(GLfloat value, GLuint frac_bits)
 }
 
 
-static struct brw_winsys_buffer *upload_default_color( struct brw_context *brw,
-				     const GLfloat *color )
+static struct brw_winsys_buffer *
+upload_default_color( struct brw_context *brw,
+		      const GLfloat *color )
 {
    struct brw_sampler_default_color sdc;
 
@@ -117,63 +118,6 @@ static void brw_update_sampler_state(struct wm_sampler_entry *key,
 {
    _mesa_memset(sampler, 0, sizeof(*sampler));
 
-   switch (key->minfilter) {
-   case GL_NEAREST:
-      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
-      sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
-      break;
-   case GL_LINEAR:
-      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
-      sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
-      break;
-   case GL_NEAREST_MIPMAP_NEAREST:
-      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
-      sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
-      break;
-   case GL_LINEAR_MIPMAP_NEAREST:
-      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
-      sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
-      break;
-   case GL_NEAREST_MIPMAP_LINEAR:
-      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
-      sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
-      break;
-   case GL_LINEAR_MIPMAP_LINEAR:
-      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
-      sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
-      break;
-   default:
-      break;
-   }
-
-   /* Set Anisotropy: 
-    */
-   if (key->max_aniso > 1.0) {
-      sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; 
-      sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC;
-
-      if (key->max_aniso > 2.0) {
-	 sampler->ss3.max_aniso = MIN2((key->max_aniso - 2) / 2,
-				       BRW_ANISORATIO_16);
-      }
-   }
-   else {
-      switch (key->magfilter) {
-      case GL_NEAREST:
-	 sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
-	 break;
-      case GL_LINEAR:
-	 sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
-	 break;
-      default:
-	 break;
-      }  
-   }
-
-   sampler->ss1.r_wrap_mode = translate_wrap_mode(key->wrap_r);
-   sampler->ss1.s_wrap_mode = translate_wrap_mode(key->wrap_s);
-   sampler->ss1.t_wrap_mode = translate_wrap_mode(key->wrap_t);
-
    /* Cube-maps on 965 and later must use the same wrap mode for all 3
     * coordinate dimensions.  Futher, only CUBE and CLAMP are valid.
     */
@@ -198,36 +142,7 @@ static void brw_update_sampler_state(struct wm_sampler_entry *key,
    }
 
 
-   /* Set shadow function: 
-    */
-   if (key->comparemode == GL_COMPARE_R_TO_TEXTURE_ARB) {
-      /* Shadowing is "enabled" by emitting a particular sampler
-       * message (sample_c).  So need to recompile WM program when
-       * shadow comparison is enabled on each/any texture unit.
-       */
-      sampler->ss0.shadow_function =
-	 intel_translate_shadow_compare_func(key->comparefunc);
-   }
-
-   /* Set LOD bias: 
-    */
-   sampler->ss0.lod_bias = S_FIXED(CLAMP(key->lod_bias, -16, 15), 6);
-
-   sampler->ss0.lod_preclamp = 1; /* OpenGL mode */
-   sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */
-
-   /* Set BaseMipLevel, MaxLOD, MinLOD: 
-    *
-    * XXX: I don't think that using firstLevel, lastLevel works,
-    * because we always setup the surface state as if firstLevel ==
-    * level zero.  Probably have to subtract firstLevel from each of
-    * these:
-    */
-   sampler->ss0.base_level = U_FIXED(0, 1);
 
-   sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(key->maxlod, 0), 13), 6);
-   sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(key->minlod, 0), 13), 6);
-   
    sampler->ss2.default_color_pointer = sdc_bo->offset >> 5; /* reloc */
 }
 
@@ -237,57 +152,42 @@ static void
 brw_wm_sampler_populate_key(struct brw_context *brw,
 			    struct wm_sampler_key *key)
 {
-   int unit;
+   int nr = MIN2(brw->curr.number_textures,
+		 brw->curr.number_samplers);
+   int i;
 
    memset(key, 0, sizeof(*key));
 
-   for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) {
-      if (ctx->Texture.Unit[unit]._ReallyEnabled) {
-	 struct wm_sampler_entry *entry = &key->sampler[unit];
-	 struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	 struct gl_texture_object *texObj = texUnit->_Current;
-	 struct intel_texture_object *intelObj = intel_texture_object(texObj);
-	 struct gl_texture_image *firstImage =
-	    texObj->Image[0][intelObj->firstLevel];
-
-         entry->tex_target = texObj->Target;
-
-	 entry->seamless_cube_map = (texObj->Target == GL_TEXTURE_CUBE_MAP)
-	    ? ctx->Texture.CubeMapSeamless : GL_FALSE;
-
-	 entry->wrap_r = texObj->WrapR;
-	 entry->wrap_s = texObj->WrapS;
-	 entry->wrap_t = texObj->WrapT;
-
-	 entry->maxlod = texObj->MaxLod;
-	 entry->minlod = texObj->MinLod;
-	 entry->lod_bias = texUnit->LodBias + texObj->LodBias;
-	 entry->max_aniso = texObj->MaxAnisotropy;
-	 entry->minfilter = texObj->MinFilter;
-	 entry->magfilter = texObj->MagFilter;
-	 entry->comparemode = texObj->CompareMode;
-         entry->comparefunc = texObj->CompareFunc;
-
-	 brw->sws->bo_unreference(brw->wm.sdc_bo[unit]);
-	 if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
-	    float bordercolor[4] = {
-	       texObj->BorderColor[0],
-	       texObj->BorderColor[0],
-	       texObj->BorderColor[0],
-	       texObj->BorderColor[0]
-	    };
-	    /* GL specs that border color for depth textures is taken from the
-	     * R channel, while the hardware uses A.  Spam R into all the
-	     * channels for safety.
-	     */
-	    brw->wm.sdc_bo[unit] = upload_default_color(brw, bordercolor);
-	 } else {
-	    brw->wm.sdc_bo[unit] = upload_default_color(brw,
-							texObj->BorderColor);
-	 }
-	 key->sampler_count = unit + 1;
+   for (i = 0; i < nr; i++) {
+      const struct brw_texture *tex = brw->curr.texture[i];
+      const struct brw_sampler *sampler = brw->curr.sampler[i];
+      struct wm_sampler_entry *entry = &key->sampler[i];
+
+      entry->tex_target = texObj->Target;
+      entry->seamless_cube_map = FALSE; /* XXX: add this to gallium */
+      entry->ss0 = sampler->ss0;
+      entry->ss1 = sampler->ss1;
+      entry->ss3 = sampler->ss3;
+
+      brw->sws->bo_unreference(brw->wm.sdc_bo[i]);
+      if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
+	 float bordercolor[4] = {
+	    texObj->BorderColor[0],
+	    texObj->BorderColor[0],
+	    texObj->BorderColor[0],
+	    texObj->BorderColor[0]
+	 };
+	 /* GL specs that border color for depth textures is taken from the
+	  * R channel, while the hardware uses A.  Spam R into all the
+	  * channels for safety.
+	  */
+	 brw->wm.sdc_bo[i] = upload_default_color(brw, bordercolor);
+      } else {
+	 brw->wm.sdc_bo[i] = upload_default_color(brw, texObj->BorderColor);
       }
    }
+
+   key->sampler_count = nr;
 }
 
 /* All samplers must be uploaded in a single contiguous array, which
@@ -354,7 +254,7 @@ static void upload_wm_samplers( struct brw_context *brw )
 
 const struct brw_tracked_state brw_wm_samplers = {
    .dirty = {
-      .mesa = _NEW_TEXTURE,
+      .mesa = PIPE_NEW_BOUND_TEXTURES | PIPE_NEW_SAMPLER,
       .brw = 0,
       .cache = 0
    },
diff --git a/src/gallium/drivers/i965/brw_wm_state.c b/src/gallium/drivers/i965/brw_wm_state.c
index 4989aae830c..edabf6ceb6a 100644
--- a/src/gallium/drivers/i965/brw_wm_state.c
+++ b/src/gallium/drivers/i965/brw_wm_state.c
@@ -65,7 +65,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
 
    memset(key, 0, sizeof(*key));
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (BRW_DEBUG & DEBUG_SINGLE_THREAD)
       key->max_threads = 1;
    else {
       /* WM maximum threads is number of EUs times number of threads per EU. */
@@ -120,7 +120,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
    ASSERT(bfp->isGLSL == brw_wm_is_glsl(fp));
 
    /* _NEW_QUERY */
-   key->stats_wm = intel->stats_wm;
+   key->stats_wm = (brw->query.stats_wm != 0);
 
    /* _NEW_LINE */
    key->line_stipple = ctx->Line.StippleFlag;
@@ -215,7 +215,7 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
 
    wm.wm5.line_stipple = key->line_stipple;
 
-   if (INTEL_DEBUG & DEBUG_STATS || key->stats_wm)
+   if (BRW_DEBUG & DEBUG_STATS || key->stats_wm)
       wm.wm4.stats_enable = 1;
 
    bo = brw_upload_cache(&brw->cache, BRW_WM_UNIT,