i965g: more compiling wip

16 files changed, 243 insertions, 200 deletions

diff --git a/src/gallium/drivers/i965/brw_context.h b/src/gallium/drivers/i965/brw_context.h
index 7ead641811..2e17e150bb 100644
--- a/src/gallium/drivers/i965/brw_context.h
+++ b/src/gallium/drivers/i965/brw_context.h
@@ -184,6 +184,8 @@ struct brw_fragment_shader {
 #define PIPE_NEW_CLIP                   0x2
 #define PIPE_NEW_INDEX_BUFFER           0x2
 #define PIPE_NEW_INDEX_RANGE            0x2
+#define PIPE_NEW_BLEND_COLOR            0x2
+#define PIPE_NEW_POLYGON_STIPPLE        0x2
 
 
 #define BRW_NEW_URB_FENCE               0x1
@@ -202,7 +204,9 @@ struct brw_fragment_shader {
 #define BRW_NEW_VERTICES		0x8000
 /**
  * Used for any batch entry with a relocated pointer that will be used
- * by any 3D rendering.
+ * by any 3D rendering.  Need to re-emit these fresh in each
+ * batchbuffer as the referenced buffers may be relocated in the
+ * meantime.
  */
 #define BRW_NEW_BATCH			0x10000
 /** brw->depth_region updated */
@@ -271,7 +275,7 @@ struct brw_vs_prog_data {
    GLuint curb_read_length;
    GLuint urb_read_length;
    GLuint total_grf;
-   GLuint outputs_written;
+   GLuint nr_outputs_written;
    GLuint nr_params;       /**< number of float params/constants */
 
    GLuint inputs_read;
@@ -487,6 +491,9 @@ struct brw_context
       struct pipe_buffer *vertex_constants;
       struct pipe_buffer *fragment_constants;
 
+      struct brw_blend_constant_color bcc;
+      struct brw_polygon_stipple bps;
+
       /**
        * Index buffer for this draw_prims call.
        *
@@ -726,11 +733,11 @@ void brw_init_shader_funcs( struct brw_context *brw );
 
 /* brw_urb.c
  */
-void brw_upload_urb_fence(struct brw_context *brw);
+int brw_upload_urb_fence(struct brw_context *brw);
 
 /* brw_curbe.c
  */
-void brw_upload_cs_urb_state(struct brw_context *brw);
+int brw_upload_cs_urb_state(struct brw_context *brw);
 
 /* brw_disasm.c */
 int brw_disasm (FILE *file, struct brw_instruction *inst);
diff --git a/src/gallium/drivers/i965/brw_curbe.c b/src/gallium/drivers/i965/brw_curbe.c
index 278ffa4ca2..3dd08f6eeb 100644
--- a/src/gallium/drivers/i965/brw_curbe.c
+++ b/src/gallium/drivers/i965/brw_curbe.c
@@ -126,7 +126,7 @@ const struct brw_tracked_state brw_curbe_offsets = {
  * fixed-function hardware in a double-buffering scheme to avoid a
  * pipeline stall each time the contents of the curbe is changed.
  */
-void brw_upload_cs_urb_state(struct brw_context *brw)
+int brw_upload_cs_urb_state(struct brw_context *brw)
 {
    struct brw_cs_urb_state cs_urb;
    memset(&cs_urb, 0, sizeof(cs_urb));
@@ -144,6 +144,7 @@ void brw_upload_cs_urb_state(struct brw_context *brw)
 
    assert(brw->urb.nr_cs_entries);
    BRW_CACHED_BATCH_STRUCT(brw, &cs_urb);
+   return 0;
 }
 
 static GLfloat fixed_plane[6][4] = {
diff --git a/src/gallium/drivers/i965/brw_gs.c b/src/gallium/drivers/i965/brw_gs.c
index 692ce46679..3ecaa74e4f 100644
--- a/src/gallium/drivers/i965/brw_gs.c
+++ b/src/gallium/drivers/i965/brw_gs.c
@@ -54,7 +54,7 @@ static void compile_gs_prog( struct brw_context *brw,
    /* Need to locate the two positions present in vertex + header.
     * These are currently hardcoded:
     */
-   c.nr_attrs = util_count_bits(c.key.attrs);
+   c.nr_attrs = c.key.nr_attrs;
 
    if (BRW_IS_IGDNG(brw))
        c.nr_regs = (c.nr_attrs + 1) / 2 + 3;  /* are vertices packed, or reg-aligned? */
@@ -80,30 +80,30 @@ static void compile_gs_prog( struct brw_context *brw,
     * already been weeded out by this stage:
     */
    switch (key->primitive) {
-   case GL_QUADS:
+   case PIPE_PRIM_QUADS:
       brw_gs_quads( &c ); 
       break;
-   case GL_QUAD_STRIP:
+   case PIPE_PRIM_QUAD_STRIP:
       brw_gs_quad_strip( &c );
       break;
-   case GL_LINE_LOOP:
+   case PIPE_PRIM_LINE_LOOP:
       brw_gs_lines( &c );
       break;
-   case GL_LINES:
+   case PIPE_PRIM_LINES:
       if (key->hint_gs_always)
 	 brw_gs_lines( &c );
       else {
 	 return;
       }
       break;
-   case GL_TRIANGLES:
+   case PIPE_PRIM_TRIANGLES:
       if (key->hint_gs_always)
 	 brw_gs_tris( &c );
       else {
 	 return;
       }
       break;
-   case GL_POINTS:
+   case PIPE_PRIM_POINTS:
       if (key->hint_gs_always)
 	 brw_gs_points( &c );
       else {
@@ -129,17 +129,17 @@ static void compile_gs_prog( struct brw_context *brw,
 				       &brw->gs.prog_data );
 }
 
-static const GLenum gs_prim[GL_POLYGON+1] = {  
-   GL_POINTS,
-   GL_LINES,
-   GL_LINE_LOOP,
-   GL_LINES,
-   GL_TRIANGLES,
-   GL_TRIANGLES,
-   GL_TRIANGLES,
-   GL_QUADS,
-   GL_QUAD_STRIP,
-   GL_TRIANGLES
+static const unsigned gs_prim[PIPE_PRIM_MAX] = {  
+   PIPE_PRIM_POINTS,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINE_LOOP,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_QUADS,
+   PIPE_PRIM_QUAD_STRIP,
+   PIPE_PRIM_TRIANGLES
 };
 
 static void populate_key( struct brw_context *brw,
@@ -148,7 +148,7 @@ static void populate_key( struct brw_context *brw,
    memset(key, 0, sizeof(*key));
 
    /* CACHE_NEW_VS_PROG */
-   key->attrs = brw->vs.prog_data->outputs_written;
+   key->nr_attrs = brw->vs.prog_data->nr_outputs_written;
 
    /* BRW_NEW_PRIMITIVE */
    key->primitive = gs_prim[brw->primitive];
@@ -156,14 +156,14 @@ static void populate_key( struct brw_context *brw,
    key->hint_gs_always = 0;	/* debug code? */
 
    key->need_gs_prog = (key->hint_gs_always ||
-			brw->primitive == GL_QUADS ||
-			brw->primitive == GL_QUAD_STRIP ||
-			brw->primitive == GL_LINE_LOOP);
+			brw->primitive == PIPE_PRIM_QUADS ||
+			brw->primitive == PIPE_PRIM_QUAD_STRIP ||
+			brw->primitive == PIPE_PRIM_LINE_LOOP);
 }
 
 /* Calculate interpolants for triangle and line rasterization.
  */
-static void prepare_gs_prog(struct brw_context *brw)
+static int prepare_gs_prog(struct brw_context *brw)
 {
    struct brw_gs_prog_key key;
    /* Populate the key:
@@ -184,6 +184,8 @@ static void prepare_gs_prog(struct brw_context *brw)
       if (brw->gs.prog_bo == NULL)
 	 compile_gs_prog( brw, &key );
    }
+
+   return 0;
 }
 
 
diff --git a/src/gallium/drivers/i965/brw_gs.h b/src/gallium/drivers/i965/brw_gs.h
index bbb991ea2e..6e616dcb87 100644
--- a/src/gallium/drivers/i965/brw_gs.h
+++ b/src/gallium/drivers/i965/brw_gs.h
@@ -40,11 +40,11 @@
 #define MAX_GS_VERTS (4)	     
 
 struct brw_gs_prog_key {
-   GLuint attrs:32;
+   GLuint nr_attrs:8;
    GLuint primitive:4;
    GLuint hint_gs_always:1;
    GLuint need_gs_prog:1;
-   GLuint pad:26;
+   GLuint pad:18;
 };
 
 struct brw_gs_compile {
diff --git a/src/gallium/drivers/i965/brw_gs_state.c b/src/gallium/drivers/i965/brw_gs_state.c
index 6d03d72d96..15a66c9741 100644
--- a/src/gallium/drivers/i965/brw_gs_state.c
+++ b/src/gallium/drivers/i965/brw_gs_state.c
@@ -29,11 +29,12 @@
   *   Keith Whitwell <keith@tungstengraphics.com>
   */
  
-
+#include "util/u_math.h"
 
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
+#include "brw_debug.h"
 
 struct brw_gs_unit_key {
    unsigned int total_grf;
@@ -76,7 +77,7 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
 
    memset(&gs, 0, sizeof(gs));
 
-   gs.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
+   gs.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1;
    if (key->prog_active) /* reloc */
       gs.thread0.kernel_start_pointer = brw->gs.prog_bo->offset >> 6;
 
@@ -100,7 +101,7 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
    if (BRW_IS_IGDNG(brw))
       gs.thread4.rendering_enable = 1;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (BRW_DEBUG & DEBUG_STATS)
       gs.thread4.stats_enable = 1;
 
    bo = brw_upload_cache(&brw->cache, BRW_GS_UNIT,
@@ -111,17 +112,17 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
 
    if (key->prog_active) {
       /* Emit GS program relocation */
-      dri_bo_emit_reloc(bo,
-			I915_GEM_DOMAIN_INSTRUCTION, 0,
-			gs.thread0.grf_reg_count << 1,
-			offsetof(struct brw_gs_unit_state, thread0),
-			brw->gs.prog_bo);
+      brw->sws->bo_emit_reloc(bo,
+			      I915_GEM_DOMAIN_INSTRUCTION, 0,
+			      gs.thread0.grf_reg_count << 1,
+			      offsetof(struct brw_gs_unit_state, thread0),
+			      brw->gs.prog_bo);
    }
 
    return bo;
 }
 
-static void prepare_gs_unit(struct brw_context *brw)
+static int prepare_gs_unit(struct brw_context *brw)
 {
    struct brw_gs_unit_key key;
 
@@ -135,6 +136,8 @@ static void prepare_gs_unit(struct brw_context *brw)
    if (brw->gs.state_bo == NULL) {
       brw->gs.state_bo = gs_unit_create_from_key(brw, &key);
    }
+
+   return 0;
 }
 
 const struct brw_tracked_state brw_gs_unit = {
diff --git a/src/gallium/drivers/i965/brw_misc_state.c b/src/gallium/drivers/i965/brw_misc_state.c
index 98fec85c1d..ccebe08b4f 100644
--- a/src/gallium/drivers/i965/brw_misc_state.c
+++ b/src/gallium/drivers/i965/brw_misc_state.c
@@ -31,10 +31,12 @@
  
 
 
+#include "brw_debug.h"
 #include "brw_batchbuffer.h"
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
+#include "brw_screen.h"
 
 
 
@@ -44,25 +46,16 @@
  * Blend color
  */
 
-static void upload_blend_constant_color(struct brw_context *brw)
+static int upload_blend_constant_color(struct brw_context *brw)
 {
-   struct brw_blend_constant_color bcc;
-
-   memset(&bcc, 0, sizeof(bcc));      
-   bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR;
-   bcc.header.length = sizeof(bcc)/4-2;
-   bcc.blend_constant_color[0] = ctx->Color.BlendColor[0];
-   bcc.blend_constant_color[1] = ctx->Color.BlendColor[1];
-   bcc.blend_constant_color[2] = ctx->Color.BlendColor[2];
-   bcc.blend_constant_color[3] = ctx->Color.BlendColor[3];
-
-   BRW_CACHED_BATCH_STRUCT(brw, &bcc);
+   BRW_CACHED_BATCH_STRUCT(brw, &brw->curr.bcc);
+   return 0;
 }
 
 
 const struct brw_tracked_state brw_blend_constant_color = {
    .dirty = {
-      .mesa = _NEW_COLOR,
+      .mesa = PIPE_NEW_BLEND_COLOR,
       .brw = 0,
       .cache = 0
    },
@@ -70,30 +63,32 @@ const struct brw_tracked_state brw_blend_constant_color = {
 };
 
 /* Constant single cliprect for framebuffer object or DRI2 drawing */
-static void upload_drawing_rect(struct brw_context *brw)
+static int upload_drawing_rect(struct brw_context *brw)
 {
    BEGIN_BATCH(4, NO_LOOP_CLIPRECTS);
    OUT_BATCH(_3DSTATE_DRAWRECT_INFO_I965);
    OUT_BATCH(0);
-   OUT_BATCH(((brw->fb.width - 1) & 0xffff) |
-	    ((brw->fb.height - 1) << 16));
+   OUT_BATCH(((brw->curr.fb.width - 1) & 0xffff) |
+	    ((brw->curr.fb.height - 1) << 16));
    OUT_BATCH(0);
    ADVANCE_BATCH();
+   return 0;
 }
 
 const struct brw_tracked_state brw_drawing_rect = {
    .dirty = {
-      .mesa = _NEW_BUFFERS,
+      .mesa = PIPE_NEW_FRAMEBUFFER,
       .brw = 0,
       .cache = 0
    },
    .emit = upload_drawing_rect
 };
 
-static void prepare_binding_table_pointers(struct brw_context *brw)
+static int prepare_binding_table_pointers(struct brw_context *brw)
 {
    brw_add_validated_bo(brw, brw->vs.bind_bo);
    brw_add_validated_bo(brw, brw->wm.bind_bo);
+   return 0;
 }
 
 /**
@@ -103,7 +98,7 @@ static void prepare_binding_table_pointers(struct brw_context *brw)
  * The binding table pointers are relative to the surface state base address,
  * which is 0.
  */
-static void upload_binding_table_pointers(struct brw_context *brw)
+static int upload_binding_table_pointers(struct brw_context *brw)
 {
    BEGIN_BATCH(6, IGNORE_CLIPRECTS);
    OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2));
@@ -116,6 +111,7 @@ static void upload_binding_table_pointers(struct brw_context *brw)
    OUT_BATCH(0); /* sf */
    OUT_RELOC(brw->wm.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* wm/ps */
    ADVANCE_BATCH();
+   return 0;
 }
 
 const struct brw_tracked_state brw_binding_table_pointers = {
@@ -135,7 +131,7 @@ const struct brw_tracked_state brw_binding_table_pointers = {
  * The state pointers in this packet are all relative to the general state
  * base address set by CMD_STATE_BASE_ADDRESS, which is 0.
  */
-static void upload_pipelined_state_pointers(struct brw_context *brw )
+static int upload_pipelined_state_pointers(struct brw_context *brw )
 {
    BEGIN_BATCH(7, IGNORE_CLIPRECTS);
    OUT_BATCH(CMD_PIPELINED_STATE_POINTERS << 16 | (7 - 2));
@@ -151,10 +147,11 @@ static void upload_pipelined_state_pointers(struct brw_context *brw )
    ADVANCE_BATCH();
 
    brw->state.dirty.brw |= BRW_NEW_PSP;
+   return 0;
 }
 
 
-static void prepare_psp_urb_cbs(struct brw_context *brw)
+static int prepare_psp_urb_cbs(struct brw_context *brw)
 {
    brw_add_validated_bo(brw, brw->vs.state_bo);
    brw_add_validated_bo(brw, brw->gs.state_bo);
@@ -162,13 +159,26 @@ static void prepare_psp_urb_cbs(struct brw_context *brw)
    brw_add_validated_bo(brw, brw->sf.state_bo);
    brw_add_validated_bo(brw, brw->wm.state_bo);
    brw_add_validated_bo(brw, brw->cc.state_bo);
+   return 0;
 }
 
-static void upload_psp_urb_cbs(struct brw_context *brw )
+static int upload_psp_urb_cbs(struct brw_context *brw )
 {
-   upload_pipelined_state_pointers(brw);
-   brw_upload_urb_fence(brw);
-   brw_upload_cs_urb_state(brw);
+   int ret;
+   
+   ret = upload_pipelined_state_pointers(brw);
+   if (ret)
+      return ret;
+
+   ret = brw_upload_urb_fence(brw);
+   if (ret)
+      return ret;
+
+   ret = brw_upload_cs_urb_state(brw);
+   if (ret)
+      return ret;
+
+   return 0;
 }
 
 const struct brw_tracked_state brw_psp_urb_cbs = {
@@ -187,20 +197,22 @@ const struct brw_tracked_state brw_psp_urb_cbs = {
    .emit = upload_psp_urb_cbs,
 };
 
-static void prepare_depthbuffer(struct brw_context *brw)
+static int prepare_depthbuffer(struct brw_context *brw)
 {
-   struct intel_region *region = brw->state.depth_region;
+   struct pipe_surface *zsbuf = brw->curr.fb.zsbuf;
 
-   if (region != NULL)
-      brw_add_validated_bo(brw, region->buffer);
+   if (zsbuf)
+      brw_add_validated_bo(brw, brw_surface_bo(zsbuf));
+
+   return 0;
 }
 
-static void emit_depthbuffer(struct brw_context *brw)
+static int emit_depthbuffer(struct brw_context *brw)
 {
-   struct intel_region *region = brw->state.depth_region;
+   struct pipe_surface *surface = brw->curr.fb.zsbuf;
    unsigned int len = (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? 6 : 5;
 
-   if (region == NULL) {
+   if (surface == NULL) {
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
       OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
       OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
@@ -214,38 +226,45 @@ static void emit_depthbuffer(struct brw_context *brw)
 
       ADVANCE_BATCH();
    } else {
+      struct brw_winsys_buffer *bo;
       unsigned int format;
+      unsigned int pitch;
+      unsigned int cpp;
 
-      switch (region->cpp) {
-      case 2:
+      switch (surface->format) {
+      case PIPE_FORMAT_Z16_UNORM:
 	 format = BRW_DEPTHFORMAT_D16_UNORM;
+	 cpp = 2;
+	 break;
+      case PIPE_FORMAT_Z24S8_UNORM:
+	 format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT;
+	 cpp = 4;
 	 break;
-      case 4:
-	 if (intel->depth_buffer_is_float)
-	    format = BRW_DEPTHFORMAT_D32_FLOAT;
-	 else
-	    format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT;
+      case PIPE_FORMAT_Z32_FLOAT:
+	 format = BRW_DEPTHFORMAT_D32_FLOAT;
+	 cpp = 4;
 	 break;
       default:
 	 assert(0);
-	 return;
+	 return PIPE_ERROR_BAD_INPUT;
       }
 
-      assert(region->tiling != I915_TILING_X);
+      bo = brw_surface_bo(surface);
+      pitch = brw_surface_pitch(surface);
 
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
       OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
-      OUT_BATCH(((region->pitch * region->cpp) - 1) |
+      OUT_BATCH(((pitch * cpp) - 1) |
 		(format << 18) |
 		(BRW_TILEWALK_YMAJOR << 26) |
-		((region->tiling != I915_TILING_NONE) << 27) |
+		((surface->layout != PIPE_SURFACE_LAYOUT_LINEAR) << 27) |
 		(BRW_SURFACE_2D << 29));
-      OUT_RELOC(region->buffer,
+      OUT_RELOC(bo,
 		I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-		0);
+		surface->offset);
       OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) |
-		((region->pitch - 1) << 6) |
-		((region->height - 1) << 19));
+		((pitch - 1) << 6) |
+		((surface->height - 1) << 19));
       OUT_BATCH(0);
 
       if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))
@@ -253,6 +272,8 @@ static void emit_depthbuffer(struct brw_context *brw)
 
       ADVANCE_BATCH();
    }
+
+   return 0;
 }
 
 const struct brw_tracked_state brw_depthbuffer = {
@@ -271,37 +292,15 @@ const struct brw_tracked_state brw_depthbuffer = {
  * Polygon stipple packet
  */
 
-static void upload_polygon_stipple(struct brw_context *brw)
+static int upload_polygon_stipple(struct brw_context *brw)
 {
-   struct brw_polygon_stipple bps;
-   GLuint i;
-
-   memset(&bps, 0, sizeof(bps));
-   bps.header.opcode = CMD_POLY_STIPPLE_PATTERN;
-   bps.header.length = sizeof(bps)/4-2;
-
-   /* Polygon stipple is provided in OpenGL order, i.e. bottom
-    * row first.  If we're rendering to a window (i.e. the
-    * default frame buffer object, 0), then we need to invert
-    * it to match our pixel layout.  But if we're rendering
-    * to a FBO (i.e. any named frame buffer object), we *don't*
-    * need to invert - we already match the layout.
-    */
-   if (ctx->DrawBuffer->Name == 0) {
-      for (i = 0; i < 32; i++)
-         bps.stipple[i] = ctx->PolygonStipple[31 - i]; /* invert */
-   }
-   else {
-      for (i = 0; i < 32; i++)
-         bps.stipple[i] = ctx->PolygonStipple[i]; /* don't invert */
-   }
-
-   BRW_CACHED_BATCH_STRUCT(brw, &bps);
+   BRW_CACHED_BATCH_STRUCT(brw, &brw->curr.bps);
+   return 0;
 }
 
 const struct brw_tracked_state brw_polygon_stipple = {
    .dirty = {
-      .mesa = _NEW_POLYGONSTIPPLE,
+      .mesa = PIPE_NEW_POLYGON_STIPPLE,
       .brw = 0,
       .cache = 0
    },
@@ -313,37 +312,26 @@ const struct brw_tracked_state brw_polygon_stipple = {
  * Polygon stipple offset packet
  */
 
-static void upload_polygon_stipple_offset(struct brw_context *brw)
+static int upload_polygon_stipple_offset(struct brw_context *brw)
 {
    struct brw_polygon_stipple_offset bpso;
 
+   /* This is invarient state in gallium:
+    */
    memset(&bpso, 0, sizeof(bpso));
    bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET;
    bpso.header.length = sizeof(bpso)/4-2;
-
-   /* Never need to offset stipple coordinates.
-    *
-    * XXX: is it ever necessary to invert Y values?
-    */
-   if (0) {
-      int x = 0, y = 0, h = 0;
-      bpso.bits0.x_offset = (32 - (x & 31)) & 31;
-      bpso.bits0.y_offset = (32 - ((y + h) & 31)) & 31;
-   }
-   else {
-      bpso.bits0.y_offset = 0;
-      bpso.bits0.x_offset = 0;
-   }
+   bpso.bits0.y_offset = 0;
+   bpso.bits0.x_offset = 0;
 
    BRW_CACHED_BATCH_STRUCT(brw, &bpso);
+   return 0;
 }
 
-#define _NEW_WINDOW_POS 0x40000000
-
 const struct brw_tracked_state brw_polygon_stipple_offset = {
    .dirty = {
-      .mesa = _NEW_WINDOW_POS,
-      .brw = 0,
+      .mesa = 0,
+      .brw = BRW_NEW_CONTEXT,
       .cache = 0
    },
    .emit = upload_polygon_stipple_offset
@@ -352,12 +340,12 @@ const struct brw_tracked_state brw_polygon_stipple_offset = {
 /**********************************************************************
  * AA Line parameters
  */
-static void upload_aa_line_parameters(struct brw_context *brw)
+static int upload_aa_line_parameters(struct brw_context *brw)
 {
    struct brw_aa_line_parameters balp;
    
    if (BRW_IS_965(brw))
-      return;
+      return 0;
 
    /* use legacy aa line coverage computation */
    memset(&balp, 0, sizeof(balp));
@@ -365,6 +353,7 @@ static void upload_aa_line_parameters(struct brw_context *brw)
    balp.header.length = sizeof(balp) / 4 - 2;
    
    BRW_CACHED_BATCH_STRUCT(brw, &balp);
+   return 0;
 }
 
 const struct brw_tracked_state brw_aa_line_parameters = {
@@ -380,31 +369,16 @@ const struct brw_tracked_state brw_aa_line_parameters = {
  * Line stipple packet
  */
 
-static void upload_line_stipple(struct brw_context *brw)
+static int upload_line_stipple(struct brw_context *brw)
 {
-   struct brw_line_stipple bls;
-   GLfloat tmp;
-   GLint tmpi;
-
-   memset(&bls, 0, sizeof(bls));
-   bls.header.opcode = CMD_LINE_STIPPLE_PATTERN;
-   bls.header.length = sizeof(bls)/4 - 2;
-
-   bls.bits0.pattern = ctx->Line.StipplePattern;
-   bls.bits1.repeat_count = ctx->Line.StippleFactor;
-
-   tmp = 1.0 / (GLfloat) ctx->Line.StippleFactor;
-   tmpi = tmp * (1<<13);
-
-
-   bls.bits1.inverse_repeat_count = tmpi;
-
-   BRW_CACHED_BATCH_STRUCT(brw, &bls);
+   struct brw_line_stipple *bls = NULL; //brw->curr.rast->bls;
+   BRW_CACHED_BATCH_STRUCT(brw, bls);
+   return 0;
 }
 
 const struct brw_tracked_state brw_line_stipple = {
    .dirty = {
-      .mesa = _NEW_LINE,
+      .mesa = PIPE_NEW_RAST,
       .brw = 0,
       .cache = 0
    },
@@ -416,7 +390,7 @@ const struct brw_tracked_state brw_line_stipple = {
  * Misc invarient state packets
  */
 
-static void upload_invarient_state( struct brw_context *brw )
+static int upload_invarient_state( struct brw_context *brw )
 {
    {
       /* 0x61040000  Pipeline Select */
@@ -424,7 +398,10 @@ static void upload_invarient_state( struct brw_context *brw )
       struct brw_pipeline_select ps;
 
       memset(&ps, 0, sizeof(ps));
-      ps.header.opcode = CMD_PIPELINE_SELECT(brw);
+      if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))
+	 ps.header.opcode = CMD_PIPELINE_SELECT_GM45;
+      else
+	 ps.header.opcode = CMD_PIPELINE_SELECT_965;
       ps.header.pipeline_select = 0;
       BRW_BATCH_STRUCT(brw, &ps);
    }
@@ -460,12 +437,18 @@ static void upload_invarient_state( struct brw_context *brw )
       struct brw_vf_statistics vfs;
       memset(&vfs, 0, sizeof(vfs));
 
-      vfs.opcode = CMD_VF_STATISTICS(brw);
-      if (INTEL_DEBUG & DEBUG_STATS)
+      if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) 
+	 vfs.opcode = CMD_VF_STATISTICS_GM45;
+      else 
+	 vfs.opcode = CMD_VF_STATISTICS_965;
+
+      if (BRW_DEBUG & DEBUG_STATS)
 	 vfs.statistics_enable = 1; 
 
       BRW_BATCH_STRUCT(brw, &vfs);
    }
+   
+   return 0;
 }
 
 const struct brw_tracked_state brw_invarient_state = {
@@ -485,7 +468,7 @@ const struct brw_tracked_state brw_invarient_state = {
  * state pools.  This comes at the expense of memory, and more expensive cache
  * misses.
  */
-static void upload_state_base_address( struct brw_context *brw )
+static int upload_state_base_address( struct brw_context *brw )
 {
    /* Output the structure (brw_state_base_address) directly to the
     * batchbuffer, so we can emit relocations inline.
@@ -511,6 +494,7 @@ static void upload_state_base_address( struct brw_context *brw )
        OUT_BATCH(1); /* Indirect object upper bound */
        ADVANCE_BATCH();
    }
+   return 0;
 }
 
 const struct brw_tracked_state brw_state_base_address = {
diff --git a/src/gallium/drivers/i965/brw_pipe_blend.c b/src/gallium/drivers/i965/brw_pipe_blend.c
index 17895d2782..54d09d9e45 100644
--- a/src/gallium/drivers/i965/brw_pipe_blend.c
+++ b/src/gallium/drivers/i965/brw_pipe_blend.c
@@ -43,3 +43,22 @@
    if (INTEL_DEBUG & DEBUG_STATS)
       cc.cc5.statistics_enable = 1;
 }
+
+
+
+static void brw_set_blend_color(struct pipe_context *pipe,
+				const float *blend_color)
+{
+   struct brw_context *brw = brw_context(pipe);
+   struct brw_blend_constant_color *bcc = &brw->curr.blend_color.bcc;
+
+   memset(bcc, 0, sizeof(*bcc));      
+   bcc->header.opcode = CMD_BLEND_CONSTANT_COLOR;
+   bcc->header.length = sizeof(*bcc)/4-2;
+   bcc->blend_constant_color[0] = blend_color[0];
+   bcc->blend_constant_color[1] = blend_color[1];
+   bcc->blend_constant_color[2] = blend_color[2];
+   bcc->blend_constant_color[3] = blend_color[3];
+
+   brw->state.dirty.pipe |= PIPE_NEW_BLEND_COLOR;
+}
diff --git a/src/gallium/drivers/i965/brw_pipe_rast.c b/src/gallium/drivers/i965/brw_pipe_rast.c
index ff64dbd48d..86822d478a 100644
--- a/src/gallium/drivers/i965/brw_pipe_rast.c
+++ b/src/gallium/drivers/i965/brw_pipe_rast.c
@@ -44,3 +44,23 @@ calculate_clip_key_rast()
       }
    }
 }
+
+
+static void
+calculate_line_stipple_rast()
+{
+   GLfloat tmp;
+   GLint tmpi;
+
+   memset(&bls, 0, sizeof(bls));
+   bls.header.opcode = CMD_LINE_STIPPLE_PATTERN;
+   bls.header.length = sizeof(bls)/4 - 2;
+   bls.bits0.pattern = brw->curr.rast.line_stipple_pattern;
+   bls.bits1.repeat_count = brw->curr.rast.line_stipple_factor + 1;
+
+   tmp = 1.0 / (GLfloat) bls.bits1.repeat_count;
+   tmpi = tmp * (1<<13);
+
+   bls.bits1.inverse_repeat_count = tmpi;
+
+}
diff --git a/src/gallium/drivers/i965/brw_screen.h b/src/gallium/drivers/i965/brw_screen.h
index b0be0e1f8a..eafd8ddf77 100644
--- a/src/gallium/drivers/i965/brw_screen.h
+++ b/src/gallium/drivers/i965/brw_screen.h
@@ -95,4 +95,11 @@ brw_buffer_is_user_buffer( const struct pipe_buffer *buf )
    return ((const struct brw_buffer *)buf)->is_user_buffer;
 }
 
+struct brw_winsys_buffer *
+brw_surface_bo( struct pipe_surface *surface );
+
+unsigned
+brw_surface_pitch( const struct pipe_surface *surface );
+
+
 #endif /* BRW_SCREEN_H */
diff --git a/src/gallium/drivers/i965/brw_sf.c b/src/gallium/drivers/i965/brw_sf.c
index e2db2e76e6..1b73b3fd51 100644
--- a/src/gallium/drivers/i965/brw_sf.c
+++ b/src/gallium/drivers/i965/brw_sf.c
@@ -131,7 +131,7 @@ 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->outputs_written; 
+   key.attrs = brw->vs.prog_data->nr_outputs_written; 
 
    /* BRW_NEW_REDUCED_PRIMITIVE */
    switch (brw->reduced_primitive) {
diff --git a/src/gallium/drivers/i965/brw_state.h b/src/gallium/drivers/i965/brw_state.h
index 9bf34c3fe4..663fc839df 100644
--- a/src/gallium/drivers/i965/brw_state.h
+++ b/src/gallium/drivers/i965/brw_state.h
@@ -33,9 +33,11 @@
 #ifndef BRW_STATE_H
 #define BRW_STATE_H
 
-#include "brw_context.h"
+#include "pipe/p_error.h"
 #include "util/u_memory.h"
 
+#include "brw_context.h"
+
 static inline void
 brw_add_validated_bo(struct brw_context *brw, struct brw_winsys_buffer *bo)
 {
diff --git a/src/gallium/drivers/i965/brw_urb.c b/src/gallium/drivers/i965/brw_urb.c
index 18d79c5ebb..a2277519ad 100644
--- a/src/gallium/drivers/i965/brw_urb.c
+++ b/src/gallium/drivers/i965/brw_urb.c
@@ -221,7 +221,7 @@ const struct brw_tracked_state brw_recalculate_urb_fence = {
 
 
 
-void brw_upload_urb_fence(struct brw_context *brw)
+int brw_upload_urb_fence(struct brw_context *brw)
 {
    struct brw_urb_fence uf;
    memset(&uf, 0, sizeof(uf));
@@ -247,4 +247,5 @@ void brw_upload_urb_fence(struct brw_context *brw)
    uf.bits1.cs_fence  = URB_SIZES(brw);
 
    BRW_BATCH_STRUCT(brw, &uf);
+   return 0;
 }
diff --git a/src/gallium/drivers/i965/brw_vs.c b/src/gallium/drivers/i965/brw_vs.c
index dcd687ac34..010ac115d3 100644
--- a/src/gallium/drivers/i965/brw_vs.c
+++ b/src/gallium/drivers/i965/brw_vs.c
@@ -51,11 +51,11 @@ static void do_vs_prog( struct brw_context *brw,
    brw_init_compile(brw, &c.func);
    c.vp = vp;
 
-   c.prog_data.outputs_written = vp->program.Base.OutputsWritten;
+   c.prog_data.nr_outputs_written = vp->program.Base.OutputsWritten;
    c.prog_data.inputs_read = vp->program.Base.InputsRead;
 
    if (c.key.copy_edgeflag) {
-      c.prog_data.outputs_written |= 1<<VERT_RESULT_EDGE;
+      c.prog_data.nr_outputs_written |= 1<<VERT_RESULT_EDGE;
       c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG;
    }
 
diff --git a/src/gallium/drivers/i965/brw_vs_emit.c b/src/gallium/drivers/i965/brw_vs_emit.c
index e946944295..086f54799e 100644
--- a/src/gallium/drivers/i965/brw_vs_emit.c
+++ b/src/gallium/drivers/i965/brw_vs_emit.c
@@ -140,35 +140,33 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    c->first_overflow_output = 0;
 
    if (BRW_IS_IGDNG(c->func.brw))
-       mrf = 8;
+      mrf = 8;
    else
-       mrf = 4;
+      mrf = 4;
 
-   for (i = 0; i < VERT_RESULT_MAX; i++) {
-      if (c->prog_data.outputs_written & (1 << i)) {
-	 c->nr_outputs++;
-         assert(i < Elements(c->regs[PROGRAM_OUTPUT]));
-	 if (i == VERT_RESULT_HPOS) {
-	    c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
-	    reg++;
+   for (i = 0; i < c->prog_data.nr_outputs_written; i++) {
+      c->nr_outputs++;
+      assert(i < Elements(c->regs[PROGRAM_OUTPUT]));
+      if (i == VERT_RESULT_HPOS) {
+	 c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
+	 reg++;
+      }
+      else if (i == VERT_RESULT_PSIZ) {
+	 c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
+	 reg++;
+	 mrf++;		/* just a placeholder?  XXX fix later stages & remove this */
+      }
+      else {
+	 if (mrf < 16) {
+	    c->regs[PROGRAM_OUTPUT][i] = brw_message_reg(mrf);
+	    mrf++;
 	 }
-	 else if (i == VERT_RESULT_PSIZ) {
+	 else {
+	    /* too many vertex results to fit in MRF, use GRF for overflow */
+	    if (!c->first_overflow_output)
+	       c->first_overflow_output = i;
 	    c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
 	    reg++;
-	    mrf++;		/* just a placeholder?  XXX fix later stages & remove this */
-	 }
-	 else {
-            if (mrf < 16) {
-               c->regs[PROGRAM_OUTPUT][i] = brw_message_reg(mrf);
-               mrf++;
-            }
-            else {
-               /* too many vertex results to fit in MRF, use GRF for overflow */
-               if (!c->first_overflow_output)
-                  c->first_overflow_output = i;
-               c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
-               reg++;
-            }
 	 }
       }
    }     
@@ -238,9 +236,9 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    attributes_in_vue = MAX2(c->nr_outputs, c->nr_inputs);
 
    if (BRW_IS_IGDNG(c->func.brw))
-       c->prog_data.urb_entry_size = (attributes_in_vue + 6 + 3) / 4;
+      c->prog_data.urb_entry_size = (attributes_in_vue + 6 + 3) / 4;
    else
-       c->prog_data.urb_entry_size = (attributes_in_vue + 2 + 3) / 4;
+      c->prog_data.urb_entry_size = (attributes_in_vue + 2 + 3) / 4;
 
    c->prog_data.total_grf = reg;
 
@@ -1050,8 +1048,9 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    /* Update the header for point size, user clipping flags, and -ve rhw
     * workaround.
     */
-   if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) ||
-       c->key.nr_userclip || BRW_IS_965(p->brw))
+   if (c->prog_data.writes_psiz ||
+       c->key.nr_userclip || 
+       BRW_IS_965(p->brw))
    {
       struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
       GLuint i;
@@ -1060,7 +1059,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 
       brw_set_access_mode(p, BRW_ALIGN_16);	
 
-      if (c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) {
+      if (c->prog_data.writes_psiz) {
 	 struct brw_reg psiz = c->regs[PROGRAM_OUTPUT][VERT_RESULT_PSIZ];
 	 brw_MUL(p, brw_writemask(header1, BRW_WRITEMASK_W), brw_swizzle1(psiz, 0), brw_imm_f(1<<11));
 	 brw_AND(p, brw_writemask(header1, BRW_WRITEMASK_W), header1, brw_imm_ud(0x7ff<<8));
@@ -1149,12 +1148,10 @@ static void emit_vertex_write( struct brw_vs_compile *c)
        * at mrf[4] atm...
        */
       GLuint i, mrf = 0;
-      for (i = c->first_overflow_output; i < VERT_RESULT_MAX; i++) {
-         if (c->prog_data.outputs_written & (1 << i)) {
-            /* move from GRF to MRF */
-            brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]);
-            mrf++;
-         }
+      for (i = c->first_overflow_output; i < c->prog_data.nr_outputs_written; i++) {
+	 /* move from GRF to MRF */
+	 brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]);
+	 mrf++;
       }
 
       brw_urb_WRITE(p,
diff --git a/src/gallium/drivers/i965/brw_wm.c b/src/gallium/drivers/i965/brw_wm.c
index 4948ea0dff..764708f7df 100644
--- a/src/gallium/drivers/i965/brw_wm.c
+++ b/src/gallium/drivers/i965/brw_wm.c
@@ -310,7 +310,7 @@ static void brw_wm_populate_key( struct brw_context *brw,
    }
 
    /* CACHE_NEW_VS_PROG */
-   key->vp_outputs_written = brw->vs.prog_data->outputs_written; /* bitmask */
+   key->vp_nr_outputs_written = brw->vs.prog_data->nr_outputs_written;
 
    /* The unique fragment program ID */
    key->program_string_id = fp->id;
diff --git a/src/gallium/drivers/i965/brw_wm.h b/src/gallium/drivers/i965/brw_wm.h
index e06de95a8a..bf241f5fa4 100644
--- a/src/gallium/drivers/i965/brw_wm.h
+++ b/src/gallium/drivers/i965/brw_wm.h
@@ -76,7 +76,7 @@ struct brw_wm_prog_key {
 
    GLuint program_string_id:32;
    GLuint drawable_height;
-   GLuint vp_outputs_written;
+   GLuint vp_nr_outputs_written;
 };