Merge the lp-surface-tiling branch into master.

This branch implemented dual representations of texture/drawing surfaces:
one in the conventional linear layout and the other the tiled layout which
is used by the fragment shader pipe.  Per-tile flags indicate the layout
of each image tile.  In many situations this lets us avoid converting
image data between the two layouts.

Squashed commit of the following:

commit 563a7e3cc552fdcfcaf9ac0d4b1683c3ba2ae732
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 14:48:21 2010 -0600

    llvmpipe: convert points/lines to triangles with draw module

    This isn't the most efficient way to render points/lines but it allows us
    to run more tests.

commit a8aa763e8a717533f2b13bb6ea53cbccbede68c9
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 14:47:28 2010 -0600

    llvmpipe: call llvmpipe_get_texture_tile() for depth/stencil

    The returned pointer isn't used, but the tile status/layout info
    gets updated.  Helps to fix glReadPixels(DEPTH / STENCIL).

commit 463bc64af266194acbea71cd52e26a79b8c8a260
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 10:58:48 2010 -0600

    llvmpipe: add store_color to debug cmd_names list

commit 784cc73fb334a9d7b7c93cbd8a1445cdf742ff58
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 10:57:43 2010 -0600

    llvmpipe: fix debug build

commit 792c93171ec075664f55720ffed397ac2834a4fc
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 10:49:01 2010 -0600

    llvmpipe: fix cube mapping

commit 882b1035db88c3dd8aebe28dc971ac30a9ee39e3
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 09:53:30 2010 -0600

    llvmpipe: remove some older/unused code

commit b807d32b23145301e8842824664d9f06b9c5502e
Author: Brian Paul <[email protected]>
Date:   Thu Apr 8 09:29:50 2010 -0600

    llvmpipe: silence warning

commit 7b337e64fec92836ccdf9d96216289dd58418e35
Author: Brian Paul <[email protected]>
Date:   Wed Apr 7 17:06:08 2010 -0600

    llvmpipe: clean-up, comments in lp_surface_copy()

commit c52fa36f249cc652fa8d5fdd94d6574127c08c41
Author: Brian Paul <[email protected]>
Date:   Wed Apr 7 16:51:42 2010 -0600

    llvmpipe: overhaul tiled/linear memory management

    Now we keep per-tile layout info (linear vs. tiled (or neither or both)
    and convert from one layout to the other on demand.

commit 4a50ccfd470547c9be0704005818a87014e9c0e9
Author: Brian Paul <[email protected]>
Date:   Wed Apr 7 16:51:27 2010 -0600

    llvmpipe: added tile read/write counters

commit b7d0ea9c687ac8773b083791623826fa604adf21
Author: Brian Paul <[email protected]>
Date:   Mon Apr 5 14:54:04 2010 -0600

    llvmpipe: rename some functions

commit ee45c6e5b95cbd3c8cccc9aa4d45d8aef11e20c4
Author: Brian Paul <[email protected]>
Date:   Mon Apr 5 14:42:15 2010 -0600

    llvmpipe: re-org some get block/tile pointer code

commit 26ce97c16c0b6520ff1538803baa772d8c3b1280
Author: Brian Paul <[email protected]>
Date:   Mon Apr 5 14:34:13 2010 -0600

    llvmpipe: disable bad assertions

commit 5c670481248c4d46f87f13bf3af5655925e7002d
Author: Brian Paul <[email protected]>
Date:   Fri Apr 2 16:36:11 2010 -0600

    llvmpipe: add a special-case optimization to lp_surface_copy()

    Be more efficient when copying tiled image to linear image.
    Before, the fallback path was always converting the whole source image
    to linear.  Now we can convert just a sub region.

commit faa684645e64d6024b3a11e4e08da825e8220b2e
Author: Brian Paul <[email protected]>
Date:   Fri Apr 2 16:15:16 2010 -0600

    llvmpipe: assorted texture and tile/line conversion code change s

    The tiled/linear conversion functions take x/y positions now to
    allow converting only sub-regions.

    More texture-related helper functions.

commit baad81ec5318d44bfac1e37c7643afc0836607bb
Author: Brian Paul <[email protected]>
Date:   Tue Mar 30 13:18:40 2010 -0600

    llvmpipe: convert tiled->linear upon PIPE_FLUSH_SWAPBUFFERS

    If we know we're about to do a swapbuffers we should immediately
    convert the tiled color tiles to linear instead of later in
    llvmpipe_texture_unmap() since we can take advantage of threading/
    parallelism here.

commit 928dd41256811daeddb7506a49a34dbad04beaf8
Author: Brian Paul <[email protected]>
Date:   Tue Mar 30 09:16:58 2010 -0600

    llvmpipe: polish-up the llvmpipe_flush() code

commit dd6014abcf86c517d159b8175e0eaeb167ea2ef6
Author: Brian Paul <[email protected]>
Date:   Tue Mar 30 09:15:17 2010 -0600

    llvmpipe: SETUP_x enum clean-up

commit 0b1ce6da2b28a41f3389685ab93e10b43c950f5d
Author: Brian Paul <[email protected]>
Date:   Fri Mar 26 10:43:37 2010 -0600

    llvmpipe: remove unused vars

commit 4562663480f88162ed4452cb05569eecb67f9f39
Author: Brian Paul <[email protected]>
Date:   Fri Mar 26 10:31:55 2010 -0600

    llvmpipe: cope with non-existant color/depth buffers

    The fragment jit functions always grab these pointers, even if they're
    not used.

commit df4329edbaf204ed501f1eac0698b8198178f9af
Author: Brian Paul <[email protected]>
Date:   Thu Mar 25 15:20:15 2010 -0600

    llvmpipe: do all render target surface mapping/unmapping in the rast code

commit 3d0c25d5ba8b8f61e8366d4c97324e45d526ff41
Author: Brian Paul <[email protected]>
Date:   Thu Mar 25 14:31:21 2010 -0600

    llvmpipe: map z/stencil buffer on demand like color buffers

    Plus lots of code clean-up and loose ends taken care of.

commit c3b6fddd788aef09b4b84b843b7b1272231151e8
Author: Brian Paul <[email protected]>
Date:   Thu Mar 25 13:15:03 2010 -0600

    llvmpipe: remove unused write_zstencil field

commit 63374d97836926a6357e9d6dd24a509a8e155c56
Author: Brian Paul <[email protected]>
Date:   Thu Mar 25 09:45:59 2010 -0600

    llvmpipe: add missing lp_rast_end() call

    Fixes crash on window resize when LP_NUM_THREADS=0.

commit 92fe9952161cc06f6edc58778e9e5a8b9ea447dc
Author: Brian Paul <[email protected]>
Date:   Wed Mar 24 10:15:19 2010 -0600

    llvmpipe: add tiled/linear conversion for 16-bit Z images

commit 6605fa28c147f30df351da0e4413cab33e4db5da
Author: Brian Paul <[email protected]>
Date:   Tue Mar 23 16:06:41 2010 -0600

    llvmpipe: implement tiled/linear conversion for Z/stencil images

commit 804528d84ffa292ef9d49d3666cdd3fa099ff3ff
Author: Brian Paul <[email protected]>
Date:   Tue Mar 23 16:05:45 2010 -0600

    llvmpipe: added texture stride comment

commit 66a88c012edf670c4ac887a912f02dcff93266dd
Author: Brian Paul <[email protected]>
Date:   Tue Mar 23 16:04:07 2010 -0600

    llvmpipe: remove unused vars

commit e2ca8d1328316dc8b36d5f688c16d109e49a6870
Author: Brian Paul <[email protected]>
Date:   Mon Mar 22 18:53:11 2010 -0600

    llvmpipe: checkpoint WIP: overhaul texture/surface mapping

    Conversion between tiled and linear surfaces is working everywhere now.
    The LP_TEXTURE_READ/READ_WRITE/WRITE_ALL flags let us avoid unnecessary
    image layout conversions.

    Still some loose ends, temporary/debug code, etc.
    Need to implement tiled/linear conversion for depth/stencil images.

commit f2730a03839ee8984c1f537b7cbebba24961397a
Author: Brian Paul <[email protected]>
Date:   Mon Mar 22 14:41:58 2010 -0600

    llvmpipe: rename/repurpose lp_rast_store_color()

commit e192a47552c5d20d2caef452ca7697e2cd852c9b
Author: Brian Paul <[email protected]>
Date:   Mon Mar 22 14:38:51 2010 -0600

    llvmpipe: remove lp_rast_load_color()

commit 3cff0bde4b4ab980e1c3e812700419091527c76b
Author: Brian Paul <[email protected]>
Date:   Mon Mar 22 14:11:38 2010 -0600

    llvmpipe: remove/consolidate texture image code

commit 3a2f08b6a550c69ef5e874f482be30252cbf8bfa
Author: Brian Paul <[email protected]>
Date:   Fri Mar 19 17:03:14 2010 -0600

    llvmpipe: checkpoint WIP: directly render to tiled texture buffers

    We're now directly writing colors into the tiled texture image buffers.

    This is a checkpoint commit with lots of dead code and temporary hacks.
    Everything will get cleaned up eventually.

commit c5ca987e03870849514d4e3c99af143722a09695
Author: Brian Paul <[email protected]>
Date:   Fri Mar 19 16:41:14 2010 -0600

    llvmpipe: refactor code, create tile_pixel_offset()

commit 2133e8273e937cbac09cd7264d6ce53af9764ddb
Author: Brian Paul <[email protected]>
Date:   Fri Mar 19 14:55:11 2010 -0600

    llvmpipe: pass LP_TEXTURE_LINEAR/TILED flags around

commit b9b9d4b82b01f4588721fdc8444740f859b4a021
Author: Brian Paul <[email protected]>
Date:   Fri Mar 19 14:51:05 2010 -0600

    llvmpipe: checkpoint WIP: hanlde co-existing tiled/linear texture data

    Cube maps are temporarily broken, maybe other things.

commit 4cd322e6889940b5f155fcb69041b685b9ef9273
Author: Brian Paul <[email protected]>
Date:   Fri Mar 19 11:34:43 2010 -0600

    progs/demos: add other modes/patterns to dissolve demo

1 files changed, 209 insertions, 113 deletions

diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index fccc63c28fe..4574f411456 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -42,15 +42,15 @@
 #include "lp_scene.h"
 
 
-/* Begin rasterizing a scene:
+/**
+ * Begin rasterizing a scene.
+ * Called once per scene by one thread.
  */
-static boolean
+static void
 lp_rast_begin( struct lp_rasterizer *rast,
                struct lp_scene *scene )
 {
    const struct pipe_framebuffer_state *fb = &scene->fb;
-   boolean write_color = fb->nr_cbufs != 0;
-   boolean write_zstencil = fb->zsbuf != NULL;
    int i;
 
    rast->curr_scene = scene;
@@ -58,59 +58,147 @@ lp_rast_begin( struct lp_rasterizer *rast,
    LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);
 
    rast->state.nr_cbufs = scene->fb.nr_cbufs;
-   rast->state.write_zstencil = write_zstencil;
-   rast->state.write_color = write_color;
    
    for (i = 0; i < rast->state.nr_cbufs; i++) {
       struct pipe_surface *cbuf = scene->fb.cbufs[i];
-      rast->cbuf[i].map = scene->cbuf_map[i];
       rast->cbuf[i].format = cbuf->texture->format;
-      rast->cbuf[i].width = cbuf->width;
-      rast->cbuf[i].height = cbuf->height;
-      rast->cbuf[i].stride = llvmpipe_resource_stride(cbuf->texture, cbuf->level);
+      rast->cbuf[i].tiles_per_row = align(cbuf->width, TILE_SIZE) / TILE_SIZE;
+      rast->cbuf[i].blocksize = 
+         util_format_get_blocksize(cbuf->texture->format);
+      rast->cbuf[i].map = llvmpipe_resource_map(cbuf->texture,
+                                               cbuf->face,
+                                               cbuf->level,
+                                               cbuf->zslice,
+                                               LP_TEX_USAGE_READ_WRITE,
+                                               LP_TEX_LAYOUT_NONE);
    }
 
-   if (write_zstencil) {
+   if (fb->zsbuf) {
       struct pipe_surface *zsbuf = scene->fb.zsbuf;
-      rast->zsbuf.map = scene->zsbuf_map;
       rast->zsbuf.stride = llvmpipe_resource_stride(zsbuf->texture, zsbuf->level);
       rast->zsbuf.blocksize = 
          util_format_get_blocksize(zsbuf->texture->format);
+
+      rast->zsbuf.map = llvmpipe_resource_map(zsbuf->texture,
+                                             zsbuf->face,
+                                             zsbuf->level,
+                                             zsbuf->zslice,
+                                             LP_TEX_USAGE_READ_WRITE,
+                                             LP_TEX_LAYOUT_NONE);
+      assert(rast->zsbuf.map);
    }
 
    lp_scene_bin_iter_begin( scene );
-   
-   return TRUE;
 }
 
 
 static void
 lp_rast_end( struct lp_rasterizer *rast )
 {
-   int i;
-
-   lp_scene_reset( rast->curr_scene );
+   struct lp_scene *scene = rast->curr_scene;
+   unsigned i;
 
-   for (i = 0; i < rast->state.nr_cbufs; i++)
+   /* Unmap color buffers */
+   for (i = 0; i < rast->state.nr_cbufs; i++) {
+      struct pipe_surface *cbuf = scene->fb.cbufs[i];
+      llvmpipe_resource_unmap(cbuf->texture,
+                             cbuf->face,
+                             cbuf->level,
+                             cbuf->zslice);
       rast->cbuf[i].map = NULL;
+   }
+
+   /* Unmap z/stencil buffer */
+   if (rast->zsbuf.map) {
+      struct pipe_surface *zsbuf = scene->fb.zsbuf;
+      llvmpipe_resource_unmap(zsbuf->texture,
+                             zsbuf->face,
+                             zsbuf->level,
+                             zsbuf->zslice);
+      rast->zsbuf.map = NULL;
+   }
+
+   lp_scene_reset( rast->curr_scene );
 
-   rast->zsbuf.map = NULL;
    rast->curr_scene = NULL;
+
+   if (0)
+      printf("Post render scene: tile read: %d  tile write: %d\n",
+             tile_read_count, tile_write_count);
 }
 
+
 /**
  * Begining rasterization of a tile.
  * \param x  window X position of the tile, in pixels
  * \param y  window Y position of the tile, in pixels
  */
 static void
-lp_rast_start_tile(struct lp_rasterizer_task *task,
+lp_rast_tile_begin(struct lp_rasterizer_task *task,
                    unsigned x, unsigned y)
 {
+   struct lp_rasterizer *rast = task->rast;
+   struct lp_scene *scene = rast->curr_scene;
+   enum lp_texture_usage usage;
+   unsigned buf;
+
    LP_DBG(DEBUG_RAST, "%s %d,%d\n", __FUNCTION__, x, y);
 
+   assert(x % TILE_SIZE == 0);
+   assert(y % TILE_SIZE == 0);
+
    task->x = x;
    task->y = y;
+
+   if (scene->has_color_clear)
+      usage = LP_TEX_USAGE_WRITE_ALL;
+   else
+      usage = LP_TEX_USAGE_READ_WRITE;
+
+   /* get pointers to color tile(s) */
+   for (buf = 0; buf < rast->state.nr_cbufs; buf++) {
+      struct pipe_surface *cbuf = rast->curr_scene->fb.cbufs[buf];
+      struct llvmpipe_resource *lpt;
+      assert(cbuf);
+      lpt = llvmpipe_resource(cbuf->texture);
+      task->color_tiles[buf] = llvmpipe_get_texture_tile(lpt,
+                                                         cbuf->face,
+                                                         cbuf->level,
+                                                         usage,
+                                                         x, y);
+      assert(task->color_tiles[buf]);
+   }
+
+   /* get pointer to depth/stencil tile */
+   {
+      struct pipe_surface *zsbuf = rast->curr_scene->fb.zsbuf;
+      if (zsbuf) {
+         struct llvmpipe_resource *lpt = llvmpipe_resource(zsbuf->texture);
+
+         if (scene->has_depth_clear)
+            usage = LP_TEX_USAGE_WRITE_ALL;
+         else
+            usage = LP_TEX_USAGE_READ_WRITE;
+
+         /* "prime" the tile: convert data from linear to tiled if necessary
+          * and update the tile's layout info.
+          */
+         (void) llvmpipe_get_texture_tile(lpt,
+                                          zsbuf->face,
+                                          zsbuf->level,
+                                          usage,
+                                          x, y);
+         /* Get actual pointer to the tile data.  Note that depth/stencil
+          * data is tiled differently than color data.
+          */
+         task->depth_tile = lp_rast_get_depth_block_pointer(rast, x, y);
+
+         assert(task->depth_tile);
+      }
+      else {
+         task->depth_tile = NULL;
+      }
+   }
 }
 
 
@@ -124,7 +212,7 @@ lp_rast_clear_color(struct lp_rasterizer_task *task,
 {
    struct lp_rasterizer *rast = task->rast;
    const uint8_t *clear_color = arg.clear_color;
-   uint8_t **color_tile = task->tile.color;
+
    unsigned i;
 
    LP_DBG(DEBUG_RAST, "%s 0x%x,0x%x,0x%x,0x%x\n", __FUNCTION__, 
@@ -138,7 +226,8 @@ lp_rast_clear_color(struct lp_rasterizer_task *task,
        clear_color[2] == clear_color[3]) {
       /* clear to grayscale value {x, x, x, x} */
       for (i = 0; i < rast->state.nr_cbufs; i++) {
-	 memset(color_tile[i], clear_color[0], TILE_SIZE * TILE_SIZE * 4);
+         uint8_t *ptr = task->color_tiles[i];
+	 memset(ptr, clear_color[0], TILE_SIZE * TILE_SIZE * 4);
       }
    }
    else {
@@ -149,8 +238,9 @@ lp_rast_clear_color(struct lp_rasterizer_task *task,
        */
       const unsigned chunk = TILE_SIZE / 4;
       for (i = 0; i < rast->state.nr_cbufs; i++) {
-         uint8_t *c = color_tile[i];
+         uint8_t *c = task->color_tiles[i];
          unsigned j;
+
          for (j = 0; j < 4 * TILE_SIZE; j++) {
             memset(c, clear_color[0], chunk);
             c += chunk;
@@ -161,7 +251,6 @@ lp_rast_clear_color(struct lp_rasterizer_task *task,
             memset(c, clear_color[3], chunk);
             c += chunk;
          }
-         assert(c - color_tile[i] == TILE_SIZE * TILE_SIZE * 4);
       }
    }
 
@@ -178,23 +267,15 @@ lp_rast_clear_zstencil(struct lp_rasterizer_task *task,
                        const union lp_rast_cmd_arg arg)
 {
    struct lp_rasterizer *rast = task->rast;
-   const unsigned tile_x = task->x;
-   const unsigned tile_y = task->y;
    const unsigned height = TILE_SIZE / TILE_VECTOR_HEIGHT;
    const unsigned width = TILE_SIZE * TILE_VECTOR_HEIGHT;
-   unsigned block_size = rast->zsbuf.blocksize;
+   const unsigned block_size = rast->zsbuf.blocksize;
+   const unsigned dst_stride = rast->zsbuf.stride * TILE_VECTOR_HEIGHT;
    uint8_t *dst;
-   unsigned dst_stride = rast->zsbuf.stride * TILE_VECTOR_HEIGHT;
    unsigned i, j;
 
    LP_DBG(DEBUG_RAST, "%s 0x%x\n", __FUNCTION__, arg.clear_zstencil);
 
-   /*assert(rast->zsbuf.map);*/
-   if (!rast->zsbuf.map)
-      return;
-
-   LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);
-
    /*
     * Clear the aera of the swizzled depth/depth buffer matching this tile, in
     * stripes of TILE_VECTOR_HEIGHT x TILE_SIZE at a time.
@@ -203,7 +284,9 @@ lp_rast_clear_zstencil(struct lp_rasterizer_task *task,
     * TILE_VECTOR_HEIGHT x TILE_VECTOR_WIDTH pixels have consecutive offsets.
     */
 
-   dst = lp_rast_depth_pointer(rast, tile_x, tile_y);
+   dst = task->depth_tile;
+
+   assert(dst == lp_rast_get_depth_block_pointer(rast, task->x, task->y));
 
    switch (block_size) {
    case 1:
@@ -236,32 +319,73 @@ lp_rast_clear_zstencil(struct lp_rasterizer_task *task,
  * Load tile color from the framebuffer surface.
  * This is a bin command called during bin processing.
  */
+#if 0
 void
 lp_rast_load_color(struct lp_rasterizer_task *task,
                    const union lp_rast_cmd_arg arg)
 {
    struct lp_rasterizer *rast = task->rast;
-   const unsigned x = task->x, y = task->y;
-   unsigned i;
+   unsigned buf;
+   enum lp_texture_usage usage;
 
    LP_DBG(DEBUG_RAST, "%s at %u, %u\n", __FUNCTION__, x, y);
 
-   for (i = 0; i < rast->state.nr_cbufs; i++) {
-      if (x >= rast->cbuf[i].width || y >= rast->cbuf[i].height)
-	 continue;
+   if (scene->has_color_clear)
+      usage = LP_TEX_USAGE_WRITE_ALL;
+   else
+      usage = LP_TEX_USAGE_READ_WRITE;
+
+   /* Get pointers to color tile(s).
+    * This will convert linear data to tiled if needed.
+    */
+   for (buf = 0; buf < rast->state.nr_cbufs; buf++) {
+      struct pipe_surface *cbuf = rast->curr_scene->fb.cbufs[buf];
+      struct llvmpipe_texture *lpt;
+      assert(cbuf);
+      lpt = llvmpipe_texture(cbuf->texture);
+      task->color_tiles[buf] = llvmpipe_get_texture_tile(lpt,
+                                                         cbuf->face,
+                                                         cbuf->level,
+                                                         usage,
+                                                         task->x, task->y);
+      assert(task->color_tiles[buf]);
+   }
+}
+#endif
 
-      lp_tile_read_4ub(rast->cbuf[i].format,
-		       task->tile.color[i],
-		       rast->cbuf[i].map, 
-		       rast->cbuf[i].stride,
-		       x, y,
-		       TILE_SIZE, TILE_SIZE);
 
-      LP_COUNT(nr_color_tile_load);
+/**
+ * Convert the color tile from tiled to linear layout.
+ * This is generally only done when we're flushing the scene just prior to
+ * SwapBuffers.  If we didn't do this here, we'd have to convert the entire
+ * tiled color buffer to linear layout in the llvmpipe_texture_unmap()
+ * function.  It's better to do it here to take advantage of
+ * threading/parallelism.
+ * This is a bin command which is stored in all bins.
+ */
+void
+lp_rast_store_color( struct lp_rasterizer_task *task,
+                     const union lp_rast_cmd_arg arg)
+{
+   struct lp_rasterizer *rast = task->rast;
+   struct lp_scene *scene = rast->curr_scene;
+   unsigned buf;
+
+   for (buf = 0; buf < rast->state.nr_cbufs; buf++) {
+      struct pipe_surface *cbuf = scene->fb.cbufs[buf];
+      const unsigned face = cbuf->face, level = cbuf->level;
+      struct llvmpipe_resource *lpt = llvmpipe_resource(cbuf->texture);
+      /* this will convert the tiled data to linear if needed */
+      (void) llvmpipe_get_texture_tile_linear(lpt, face,level,
+                                              LP_TEX_USAGE_READ,
+                                              task->x, task->y);
    }
 }
 
 
+/**
+ * This is a bin command called during bin processing.
+ */
 void
 lp_rast_set_state(struct lp_rasterizer_task *task,
                   const union lp_rast_cmd_arg arg)
@@ -275,7 +399,6 @@ lp_rast_set_state(struct lp_rasterizer_task *task,
 }
 
 
-
 /**
  * Run the shader on all blocks in a tile.  This is used when a tile is
  * completely contained inside a triangle.
@@ -287,7 +410,6 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
 {
    struct lp_rasterizer *rast = task->rast;
    const struct lp_rast_state *state = task->current_state;
-   struct lp_rast_tile *tile = &task->tile;
    const struct lp_rast_shader_inputs *inputs = arg.shade_tile;
    const unsigned tile_x = task->x, tile_y = task->y;
    unsigned x, y;
@@ -299,19 +421,17 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
       for (x = 0; x < TILE_SIZE; x += 4) {
          uint8_t *color[PIPE_MAX_COLOR_BUFS];
          uint32_t *depth;
-         unsigned block_offset, i;
-
-         /* offset of the 16x16 pixel block within the tile */
-         block_offset = ((y / 4) * (16 * 16) + (x / 4) * 16);
+         unsigned i;
 
          /* color buffer */
          for (i = 0; i < rast->state.nr_cbufs; i++)
-            color[i] = tile->color[i] + 4 * block_offset;
+            color[i] = lp_rast_get_color_block_pointer(task, i,
+                                                       tile_x + x, tile_y + y);
 
          /* depth buffer */
-         depth = lp_rast_depth_pointer(rast, tile_x + x, tile_y + y);
+         depth = lp_rast_get_depth_block_pointer(rast, tile_x + x, tile_y + y);
 
-         /* run shader */
+         /* run shader on 4x4 block */
          state->jit_function[RAST_WHOLE]( &state->jit_context,
                                           tile_x + x, tile_y + y,
                                           inputs->facing,
@@ -330,6 +450,8 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
 /**
  * Compute shading for a 4x4 block of pixels.
  * This is a bin command called during bin processing.
+ * \param x  X position of quad in window coords
+ * \param y  Y position of quad in window coords
  */
 void lp_rast_shade_quads( struct lp_rasterizer_task *task,
                           const struct lp_rast_shader_inputs *inputs,
@@ -338,12 +460,9 @@ void lp_rast_shade_quads( struct lp_rasterizer_task *task,
 {
    const struct lp_rast_state *state = task->current_state;
    struct lp_rasterizer *rast = task->rast;
-   struct lp_rast_tile *tile = &task->tile;
    uint8_t *color[PIPE_MAX_COLOR_BUFS];
    void *depth;
    unsigned i;
-   unsigned ix, iy;
-   int block_offset;
 
    assert(state);
 
@@ -354,28 +473,23 @@ void lp_rast_shade_quads( struct lp_rasterizer_task *task,
    assert((x % 4) == 0);
    assert((y % 4) == 0);
 
-   ix = x % TILE_SIZE;
-   iy = y % TILE_SIZE;
-
-   /* offset of the 16x16 pixel block within the tile */
-   block_offset = ((iy / 4) * (16 * 16) + (ix / 4) * 16);
-
    /* color buffer */
-   for (i = 0; i < rast->state.nr_cbufs; i++)
-      color[i] = tile->color[i] + 4 * block_offset;
+   for (i = 0; i < rast->state.nr_cbufs; i++) {
+      color[i] = lp_rast_get_color_block_pointer(task, i, x, y);
+      assert(lp_check_alignment(color[i], 16));
+   }
 
    /* depth buffer */
-   depth = lp_rast_depth_pointer(rast, x, y);
+   depth = lp_rast_get_depth_block_pointer(rast, x, y);
 
 
-   assert(lp_check_alignment(tile->color[0], 16));
    assert(lp_check_alignment(state->jit_context.blend_color, 16));
 
    assert(lp_check_alignment(inputs->step[0], 16));
    assert(lp_check_alignment(inputs->step[1], 16));
    assert(lp_check_alignment(inputs->step[2], 16));
 
-   /* run shader */
+   /* run shader on 4x4 block */
    state->jit_function[RAST_EDGE_TEST]( &state->jit_context,
                                         x, y,
                                         inputs->facing,
@@ -445,39 +559,31 @@ outline_subtiles(uint8_t *tile)
 
 
 /**
- * Write the rasterizer's color tile to the framebuffer.
+ * Called when we're done writing to a color tile.
  */
 static void
-lp_rast_store_color(struct lp_rasterizer_task *task)
+lp_rast_tile_end(struct lp_rasterizer_task *task)
 {
+#if DEBUG
    struct lp_rasterizer *rast = task->rast;
-   const unsigned x = task->x, y = task->y;
-   unsigned i;
-
-   for (i = 0; i < rast->state.nr_cbufs; i++) {
-      if (x >= rast->cbuf[i].width)
-	 continue;
-
-      if (y >= rast->cbuf[i].height)
-	 continue;
+   unsigned buf;
 
-      LP_DBG(DEBUG_RAST, "%s [%u] %d,%d\n", __FUNCTION__,
-	     task->thread_index, x, y);
+   for (buf = 0; buf < rast->state.nr_cbufs; buf++) {
+      uint8_t *color = lp_rast_get_color_block_pointer(task, buf,
+                                                       task->x, task->y);
 
       if (LP_DEBUG & DEBUG_SHOW_SUBTILES)
-         outline_subtiles(task->tile.color[i]);
+         outline_subtiles(color);
       else if (LP_DEBUG & DEBUG_SHOW_TILES)
-         outline_tile(task->tile.color[i]);
-
-      lp_tile_write_4ub(rast->cbuf[i].format,
-			task->tile.color[i],
-			rast->cbuf[i].map, 
-			rast->cbuf[i].stride,
-			x, y,
-			TILE_SIZE, TILE_SIZE);
-
-      LP_COUNT(nr_color_tile_store);
+         outline_tile(color);
    }
+#else
+   (void) outline_subtiles;
+#endif
+
+   /* debug */
+   memset(task->color_tiles, 0, sizeof(task->color_tiles));
+   task->depth_tile = NULL;
 }
 
 
@@ -512,7 +618,7 @@ rasterize_bin(struct lp_rasterizer_task *task,
    struct cmd_block *block;
    unsigned k;
 
-   lp_rast_start_tile( task, x * TILE_SIZE, y * TILE_SIZE );
+   lp_rast_tile_begin( task, x * TILE_SIZE, y * TILE_SIZE );
 
    /* simply execute each of the commands in the block list */
    for (block = commands->head; block; block = block->next) {
@@ -521,10 +627,7 @@ rasterize_bin(struct lp_rasterizer_task *task,
       }
    }
 
-   /* Write the rasterizer's tiles to the framebuffer.
-    */
-   if (task->rast->state.write_color)
-      lp_rast_store_color(task);
+   lp_rast_tile_end(task);
 
    /* Free data for this bin.
     */
@@ -539,12 +642,12 @@ static struct {
    const char *name;
 } cmd_names[] = 
 {
-   RAST(load_color),
    RAST(clear_color),
    RAST(clear_zstencil),
    RAST(triangle),
    RAST(shade_tile),
    RAST(set_state),
+   RAST(store_color),
    RAST(fence),
 };
 
@@ -597,8 +700,7 @@ is_empty_bin( const struct cmd_bin *bin )
    }
 
    for (i = 0; i < head->count; i++)
-      if (head->cmd[i] != lp_rast_load_color &&
-          head->cmd[i] != lp_rast_set_state) {
+      if (head->cmd[i] != lp_rast_set_state) {
          return FALSE;
       }
 
@@ -658,6 +760,9 @@ lp_rast_queue_scene( struct lp_rasterizer *rast,
       rasterize_scene( &rast->tasks[0], scene );
 
       lp_scene_reset( scene );
+
+      lp_rast_end( rast );
+
       rast->curr_scene = NULL;
    }
    else {
@@ -798,7 +903,7 @@ struct lp_rasterizer *
 lp_rast_create( void )
 {
    struct lp_rasterizer *rast;
-   unsigned i, cbuf;
+   unsigned i;
 
    rast = CALLOC_STRUCT(lp_rasterizer);
    if(!rast)
@@ -808,10 +913,6 @@ lp_rast_create( void )
 
    for (i = 0; i < Elements(rast->tasks); i++) {
       struct lp_rasterizer_task *task = &rast->tasks[i];
-
-      for (cbuf = 0; cbuf < PIPE_MAX_COLOR_BUFS; cbuf++ )
-	 task->tile.color[cbuf] = align_malloc(TILE_SIZE * TILE_SIZE * 4, 16);
-
       task->rast = rast;
       task->thread_index = i;
    }
@@ -829,12 +930,7 @@ lp_rast_create( void )
  */
 void lp_rast_destroy( struct lp_rasterizer *rast )
 {
-   unsigned i, cbuf;
-
-   for (i = 0; i < Elements(rast->tasks); i++) {
-      for (cbuf = 0; cbuf < PIPE_MAX_COLOR_BUFS; cbuf++ )
-	 align_free(rast->tasks[i].tile.color[cbuf]);
-   }
+   unsigned i;
 
    /* Set exit_flag and signal each thread's work_ready semaphore.
     * Each thread will be woken up, notice that the exit_flag is set and