Merge remote branch 'upstream/gallium-0.1' into nouveau-gallium-0.1

60 files changed, 3084 insertions, 1956 deletions

diff --git a/src/gallium/README.portability b/src/gallium/README.portability
index 18a97f449b4..d5d5987a7ff 100644
--- a/src/gallium/README.portability
+++ b/src/gallium/README.portability
@@ -6,7 +6,7 @@
 The state tracker and winsys driver support a rather limited number of
 platforms. However, the pipe drivers are meant to run in a wide number of
 platforms. Hence the pipe drivers, the auxiliary modules, and all public
-headers in general, should stricly follow these guidelines to ensure
+headers in general, should strictly follow these guidelines to ensure
 
 
 = Compiler Support =
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index a1a3a9efaf3..af4af8ac1d4 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -292,7 +292,7 @@ enum pipe_error cso_set_blend(struct cso_context *ctx,
       if (!cso)
          return PIPE_ERROR_OUT_OF_MEMORY;
 
-      cso->state = *templ;
+      memcpy(&cso->state, templ, sizeof(*templ));
       cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state);
       cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state;
       cso->context = ctx->pipe;
@@ -350,7 +350,7 @@ enum pipe_error cso_single_sampler(struct cso_context *ctx,
          if (!cso)
             return PIPE_ERROR_OUT_OF_MEMORY;
 
-         cso->state = *templ;
+         memcpy(&cso->state, templ, sizeof(*templ));
          cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state);
          cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state;
          cso->context = ctx->pipe;
@@ -508,7 +508,7 @@ enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx,
       if (!cso)
          return PIPE_ERROR_OUT_OF_MEMORY;
 
-      cso->state = *templ;
+      memcpy(&cso->state, templ, sizeof(*templ));
       cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state);
       cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state;
       cso->context = ctx->pipe;
@@ -564,7 +564,7 @@ enum pipe_error cso_set_rasterizer(struct cso_context *ctx,
       if (!cso)
          return PIPE_ERROR_OUT_OF_MEMORY;
 
-      cso->state = *templ;
+      memcpy(&cso->state, templ, sizeof(*templ));
       cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state);
       cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state;
       cso->context = ctx->pipe;
@@ -726,7 +726,7 @@ enum pipe_error cso_set_vertex_shader(struct cso_context *ctx,
       if (!cso)
          return PIPE_ERROR_OUT_OF_MEMORY;
 
-      cso->state = *templ;
+      memcpy(cso->state, templ, sizeof(*templ));
       cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state);
       cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state;
       cso->context = ctx->pipe;
@@ -769,8 +769,7 @@ void cso_restore_vertex_shader(struct cso_context *ctx)
 enum pipe_error cso_set_framebuffer(struct cso_context *ctx,
                                     const struct pipe_framebuffer_state *fb)
 {
-   /* XXX this memcmp() fails to detect buffer size changes */
-   if (1/*memcmp(&ctx->fb, fb, sizeof(*fb))*/) {
+   if (memcmp(&ctx->fb, fb, sizeof(*fb)) != 0) {
       ctx->fb = *fb;
       ctx->pipe->set_framebuffer_state(ctx->pipe, fb);
    }
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
index 346a96e37fd..b1ed8aa24ea 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -415,8 +415,11 @@ aaline_create_texture(struct aaline_stage *aaline)
 
       assert(aaline->texture->width[level] == aaline->texture->height[level]);
 
-      surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0);
-      data = pipe_surface_map(surface);
+      /* This texture is new, no need to flush. 
+       */
+      surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+      data = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE);
       if (data == NULL)
          return FALSE;
 
@@ -440,9 +443,8 @@ aaline_create_texture(struct aaline_stage *aaline)
       }
 
       /* unmap */
-      pipe_surface_unmap(surface);
-      pipe_surface_reference(&surface, NULL);
-      pipe->texture_update(pipe, aaline->texture, 0, (1 << level));
+      screen->surface_unmap(screen, surface);
+      screen->tex_surface_release(screen, &surface);
    }
    return TRUE;
 }
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
index 2dfd1800d55..4c92416eb19 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -376,8 +376,14 @@ pstip_update_texture(struct pstip_stage *pstip)
    uint i, j;
    ubyte *data;
 
-   surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0);
-   data = pipe_surface_map(surface);
+   /* XXX: want to avoid flushing just because we use stipple: 
+    */
+   pipe->flush( pipe, PIPE_FLUSH_TEXTURE_CACHE, NULL );
+
+   surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+   data = screen->surface_map(screen, surface,
+                              PIPE_BUFFER_USAGE_CPU_WRITE);
 
    /*
     * Load alpha texture.
@@ -399,9 +405,8 @@ pstip_update_texture(struct pstip_stage *pstip)
    }
 
    /* unmap */
-   pipe_surface_unmap(surface);
-   pipe_surface_reference(&surface, NULL);
-   pipe->texture_update(pipe, pstip->texture, 0, 0x1);
+   screen->surface_unmap(screen, surface);
+   screen->tex_surface_release(screen, &surface);
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
index 86a7d1c7303..67b9a9503d2 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -427,9 +427,6 @@ static void vbuf_destroy( struct draw_stage *stage )
    if(vbuf->indices)
       align_free( vbuf->indices );
    
-   if(vbuf->translate)
-      vbuf->translate->release( vbuf->translate );
-
    if (vbuf->render)
       vbuf->render->destroy( vbuf->render );
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
index 100117a9aef..b96335b7895 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -184,7 +184,8 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw )
 
 void draw_pt_fetch_destroy( struct pt_fetch *fetch )
 {
-   translate_cache_destroy(fetch->cache);
+   if (fetch->cache)
+      translate_cache_destroy(fetch->cache);
 
    FREE(fetch);
 }
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
index b7b970a297f..4ea7d4359ff 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
@@ -268,7 +268,8 @@ static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
 {
    struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
 
-   translate_cache_destroy(feme->cache);
+   if (feme->cache)
+      translate_cache_destroy(feme->cache);
 
    FREE(middle);
 }
diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile
index d654dbcc919..ff09011b660 100644
--- a/src/gallium/auxiliary/pipebuffer/Makefile
+++ b/src/gallium/auxiliary/pipebuffer/Makefile
@@ -11,6 +11,7 @@ C_SOURCES = \
 	pb_bufmgr_mm.c \
 	pb_bufmgr_pool.c \
 	pb_bufmgr_slab.c \
+	pb_validate.c \
 	pb_winsys.c
 
 include ../../Makefile.template
diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript
index 604a217982e..9db0c0eae3b 100644
--- a/src/gallium/auxiliary/pipebuffer/SConscript
+++ b/src/gallium/auxiliary/pipebuffer/SConscript
@@ -10,6 +10,7 @@ pipebuffer = env.ConvenienceLibrary(
 		'pb_bufmgr_mm.c',
 		'pb_bufmgr_pool.c',
 		'pb_bufmgr_slab.c',
+		'pb_validate.c',
 		'pb_winsys.c',
 	])
 
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
index 49705cb8627..857ea53c786 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
@@ -146,6 +146,8 @@ pb_map(struct pb_buffer *buf,
        unsigned flags)
 {
    assert(buf);
+   if(!buf)
+      return NULL;
    return buf->vtbl->map(buf, flags);
 }
 
@@ -154,6 +156,8 @@ static INLINE void
 pb_unmap(struct pb_buffer *buf)
 {
    assert(buf);
+   if(!buf)
+      return;
    buf->vtbl->unmap(buf);
 }
 
@@ -163,6 +167,12 @@ pb_get_base_buffer( struct pb_buffer *buf,
 		    struct pb_buffer **base_buf,
 		    unsigned *offset )
 {
+   assert(buf);
+   if(!buf) {
+      base_buf = NULL;
+      offset = 0;
+      return;
+   }
    buf->vtbl->get_base_buffer(buf, base_buf, offset);
 }
 
@@ -171,7 +181,8 @@ static INLINE void
 pb_destroy(struct pb_buffer *buf)
 {
    assert(buf);
-   assert(buf->vtbl);
+   if(!buf)
+      return;
    buf->vtbl->destroy(buf);
 }
 
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
index 2fa08429715..7f236887a90 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
@@ -168,6 +168,28 @@ _fenced_buffer_remove(struct fenced_buffer *fenced_buf)
 }
 
 
+static INLINE enum pipe_error
+_fenced_buffer_finish(struct fenced_buffer *fenced_buf)
+{
+   struct fenced_buffer_list *fenced_list = fenced_buf->list;
+   struct pipe_winsys *winsys = fenced_list->winsys;
+
+   debug_warning("waiting for GPU");
+
+   assert(fenced_buf->fence);
+   if(fenced_buf->fence) {
+      if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) {
+	 return PIPE_ERROR;
+      }
+      /* Remove from the fenced list */
+      _fenced_buffer_remove(fenced_buf); /* TODO: remove consequents */
+   }
+
+   fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+   return PIPE_OK;
+}
+
+
 /**
  * Free as many fenced buffers from the list head as possible. 
  */
@@ -207,40 +229,6 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list,
 }
 
 
-/**
- * Serialize writes, but allow concurrent reads.
- */
-static INLINE enum pipe_error
-fenced_buffer_serialize(struct fenced_buffer *fenced_buf, unsigned flags)
-{
-   struct fenced_buffer_list *fenced_list = fenced_buf->list;
-   struct pipe_winsys *winsys = fenced_list->winsys;
-
-   /* Allow concurrent reads */
-   if(((fenced_buf->flags | flags) & PIPE_BUFFER_USAGE_WRITE) == 0)
-      return PIPE_OK;
-
-   /* Wait for the CPU to finish */
-   if(fenced_buf->mapcount) {
-      /* FIXME: Use thread conditions variables to signal when mapcount 
-       * reaches zero */
-      debug_warning("attemp to write concurrently to buffer");
-      /* XXX: we must not fail here in order to support texture mipmap generation
-      return PIPE_ERROR_RETRY;
-       */
-   }
-
-   /* Wait for the GPU to finish */
-   if(fenced_buf->fence) {
-      if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0)
-	 return PIPE_ERROR_RETRY; 
-      _fenced_buffer_remove(fenced_buf);
-   }
-
-   return PIPE_OK;
-}
-
-
 static void
 fenced_buffer_destroy(struct pb_buffer *buf)
 {
@@ -280,15 +268,28 @@ fenced_buffer_map(struct pb_buffer *buf,
 {
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);
    void *map;
-   assert((flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE) == 0);
+
+   assert(!(flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE));
+   flags &= PIPE_BUFFER_USAGE_CPU_READ_WRITE;
    
-   if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK)
-      return NULL;
+   /* Check for GPU read/write access */
+   if(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) {
+      /* Wait for the GPU to finish writing */
+      _fenced_buffer_finish(fenced_buf);
+   }
+
+   /* Check for CPU write access (read is OK) */
+   if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
+      /* this is legal -- just for debugging */
+      debug_warning("concurrent CPU writes");
+   }
    
    map = pb_map(fenced_buf->buffer, flags);
-   if(map)
+   if(map) {
       ++fenced_buf->mapcount;
-   fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+      fenced_buf->flags |= flags;
+   }
+
    return map;
 }
 
@@ -298,10 +299,12 @@ fenced_buffer_unmap(struct pb_buffer *buf)
 {
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);
    assert(fenced_buf->mapcount);
-   pb_unmap(fenced_buf->buffer);
-   --fenced_buf->mapcount;
-   if(!fenced_buf->mapcount)
-      fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+   if(fenced_buf->mapcount) {
+      pb_unmap(fenced_buf->buffer);
+      --fenced_buf->mapcount;
+      if(!fenced_buf->mapcount)
+	 fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+   }
 }
 
 
@@ -334,8 +337,10 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list,
       return NULL;
    
    buf = CALLOC_STRUCT(fenced_buffer);
-   if(!buf)
+   if(!buf) {
+      pb_reference(&buffer, NULL);
       return NULL;
+   }
    
    buf->base.base.refcount = 1;
    buf->base.base.alignment = buffer->base.alignment;
@@ -374,7 +379,7 @@ buffer_fence(struct pb_buffer *buf,
    fenced_list = fenced_buf->list;
    winsys = fenced_list->winsys;
    
-   if(fence == fenced_buf->fence) {
+   if(!fence || fence == fenced_buf->fence) {
       /* Handle the same fence case specially, not only because it is a fast 
        * path, but mostly to avoid serializing two writes with the same fence, 
        * as that would bring the hardware down to synchronous operation without
@@ -384,11 +389,6 @@ buffer_fence(struct pb_buffer *buf,
       return;
    }
    
-   if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK) {
-      /* FIXME: propagate error */
-      (void)0;
-   }
-   
    _glthread_LOCK_MUTEX(fenced_list->mutex);
    if (fenced_buf->fence)
       _fenced_buffer_remove(fenced_buf);
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
index 8de286e3f94..f6cc7a525b7 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
@@ -50,7 +50,8 @@
 #define PB_BUFMGR_H_
 
 
-#include <stddef.h>
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
 
 
 #ifdef __cplusplus
@@ -68,7 +69,6 @@ struct pipe_winsys;
  */
 struct pb_manager
 {
-   /* XXX: we will likely need more allocation flags */
    struct pb_buffer *
    (*create_buffer)( struct pb_manager *mgr, 
 	             size_t size,
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
index b931455056e..45ba158a4d9 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
@@ -47,9 +47,6 @@
 #include "pb_bufmgr.h"
 
 
-#define DRI_SLABPOOL_ALLOC_RETRIES 100
-
-
 struct pb_slab;
 
 struct pb_slab_buffer
@@ -313,7 +310,6 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr,
    static struct pb_slab_buffer *buf;
    struct pb_slab *slab;
    struct list_head *list;
-   int count = DRI_SLABPOOL_ALLOC_RETRIES;
 
    /* check size */
    assert(size == mgr->bufSize);
@@ -331,23 +327,14 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr,
    /* XXX: check for compatible buffer usage too? */
    
    _glthread_LOCK_MUTEX(mgr->mutex);
-   while (mgr->slabs.next == &mgr->slabs && count > 0) {
-      if (mgr->slabs.next != &mgr->slabs)
-	 break;
-
-      _glthread_UNLOCK_MUTEX(mgr->mutex);
-      if (count != DRI_SLABPOOL_ALLOC_RETRIES)
-	 util_time_sleep(1);
-      _glthread_LOCK_MUTEX(mgr->mutex);
+   if (mgr->slabs.next == &mgr->slabs) {
       (void) pb_slab_create(mgr);
-      count--;
+      if (mgr->slabs.next == &mgr->slabs) {
+	 _glthread_UNLOCK_MUTEX(mgr->mutex);
+	 return NULL;
+      }
    }
-
    list = mgr->slabs.next;
-   if (list == &mgr->slabs) {
-      _glthread_UNLOCK_MUTEX(mgr->mutex);
-      return NULL;
-   }
    slab = LIST_ENTRY(struct pb_slab, list, head);
    if (--slab->numFree == 0)
       LIST_DELINIT(list);
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c
new file mode 100644
index 00000000000..362fd896f31
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c
@@ -0,0 +1,153 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Buffer validation.
+ * 
+ * @author Jose Fonseca <[email protected]>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+#include "pipe/p_util.h"
+#include "pipe/p_debug.h"
+
+#include "pb_buffer.h"
+#include "pb_buffer_fenced.h"
+#include "pb_validate.h"
+
+
+#define PB_VALIDATE_INITIAL_SIZE 1 /* 512 */ 
+
+
+struct pb_validate
+{
+   struct pb_buffer **buffers;
+   unsigned used;
+   unsigned size;
+};
+
+
+enum pipe_error
+pb_validate_add_buffer(struct pb_validate *vl,
+                       struct pb_buffer *buf)
+{
+   assert(buf);
+   if(!buf)
+      return PIPE_ERROR;
+
+   /* We only need to store one reference for each buffer, so avoid storing
+    * consecutive references for the same buffer. It might not be the more 
+    * common pasttern, but it is easy to implement.
+    */
+   if(vl->used && vl->buffers[vl->used - 1] == buf) {
+      return PIPE_OK;
+   }
+   
+   /* Grow the table */
+   if(vl->used == vl->size) {
+      unsigned new_size;
+      struct pb_buffer **new_buffers;
+      
+      new_size = vl->size * 2;
+      if(!new_size)
+	 return PIPE_ERROR_OUT_OF_MEMORY;
+
+      new_buffers = (struct pb_buffer **)REALLOC(vl->buffers,
+                                                 vl->size*sizeof(struct pb_buffer *),
+                                                 new_size*sizeof(struct pb_buffer *));
+      if(!new_buffers)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      
+      memset(new_buffers + vl->size, 0, (new_size - vl->size)*sizeof(struct pb_buffer *));
+      
+      vl->size = new_size;
+      vl->buffers = new_buffers;
+   }
+   
+   assert(!vl->buffers[vl->used]);
+   pb_reference(&vl->buffers[vl->used], buf);
+   ++vl->used;
+   
+   return PIPE_OK;
+}
+
+
+enum pipe_error
+pb_validate_validate(struct pb_validate *vl) 
+{
+   /* FIXME: go through each buffer, ensure its not mapped, its address is 
+    * available -- requires a new pb_buffer interface */
+   return PIPE_OK;
+}
+
+
+void
+pb_validate_fence(struct pb_validate *vl,
+                  struct pipe_fence_handle *fence)
+{
+   unsigned i;
+   for(i = 0; i < vl->used; ++i) {
+      buffer_fence(vl->buffers[i], fence);
+      pb_reference(&vl->buffers[i], NULL);
+   }
+   vl->used = 0;
+}
+
+
+void
+pb_validate_destroy(struct pb_validate *vl)
+{
+   unsigned i;
+   for(i = 0; i < vl->used; ++i)
+      pb_reference(&vl->buffers[i], NULL);
+   FREE(vl->buffers);
+   FREE(vl);
+}
+
+
+struct pb_validate *
+pb_validate_create()
+{
+   struct pb_validate *vl;
+   
+   vl = CALLOC_STRUCT(pb_validate);
+   if(!vl)
+      return NULL;
+   
+   vl->size = PB_VALIDATE_INITIAL_SIZE;
+   vl->buffers = (struct pb_buffer **)CALLOC(vl->size, sizeof(struct pb_buffer *));
+   if(!vl->buffers) {
+      FREE(vl);
+      return NULL;
+   }
+
+   return vl;
+}
+
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.h b/src/gallium/auxiliary/pipebuffer/pb_validate.h
new file mode 100644
index 00000000000..3db1d5330b9
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.h
@@ -0,0 +1,91 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Buffer validation.
+ * 
+ * @author Jose Fonseca <[email protected]>
+ */
+
+#ifndef PB_VALIDATE_H_
+#define PB_VALIDATE_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pb_buffer;
+struct pipe_fence_handle;
+
+
+/**
+ * Buffer validation list.
+ * 
+ * It holds a list of buffers to be validated and fenced when flushing.
+ */
+struct pb_validate;
+
+
+enum pipe_error
+pb_validate_add_buffer(struct pb_validate *vl,
+                       struct pb_buffer *buf);
+
+/**
+ * Validate all buffers for hardware access.
+ * 
+ * Should be called right before issuing commands to the hardware.
+ */
+enum pipe_error
+pb_validate_validate(struct pb_validate *vl);
+
+/**
+ * Fence all buffers and clear the list.
+ * 
+ * Should be called right before issuing commands to the hardware.
+ */
+void
+pb_validate_fence(struct pb_validate *vl,
+                  struct pipe_fence_handle *fence);
+
+struct pb_validate *
+pb_validate_create(void);
+
+void
+pb_validate_destroy(struct pb_validate *vl);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*PB_VALIDATE_H_*/
diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript
index d55d2c70818..0309de1ac22 100644
--- a/src/gallium/auxiliary/util/SConscript
+++ b/src/gallium/auxiliary/util/SConscript
@@ -5,6 +5,7 @@ util = env.ConvenienceLibrary(
 	source = [
 		'p_debug.c',
 		'p_debug_mem.c',
+		'p_debug_prof.c',
 		'p_tile.c',
 		'p_util.c',
 		'u_blit.c',
diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c
index 4ec17466624..ce7fb589562 100644
--- a/src/gallium/auxiliary/util/p_debug.c
+++ b/src/gallium/auxiliary/util/p_debug.c
@@ -67,7 +67,7 @@ void _debug_vprintf(const char *format, va_list ap)
    static char buf[512 + 1] = {'\0'};
    size_t len = strlen(buf);
    int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap);
-   if(ret > (int)(sizeof(buf) - len - 1) || strchr(buf + len, '\n')) {
+   if(ret > (int)(sizeof(buf) - len - 1) || util_strchr(buf + len, '\n')) {
       _EngDebugPrint("%s", buf);
       buf[0] = '\0';
    }
@@ -154,6 +154,7 @@ debug_get_option(const char *name, const char *dfault)
 {
    const char *result;
 #ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
+#ifdef DEBUG
    ULONG_PTR iFile = 0;
    const void *pMap = NULL;
    const char *sol, *eol, *sep;
@@ -184,6 +185,9 @@ debug_get_option(const char *name, const char *dfault)
       EngUnmapFile(iFile);
    }
 #else
+   result = dfault;
+#endif
+#else
    
    result = getenv(name);
    if(!result)
@@ -203,15 +207,15 @@ debug_get_bool_option(const char *name, boolean dfault)
    
    if(str == NULL)
       result = dfault;
-   else if(!strcmp(str, "n"))
+   else if(!util_strcmp(str, "n"))
       result = FALSE;
-   else if(!strcmp(str, "no"))
+   else if(!util_strcmp(str, "no"))
       result = FALSE;
-   else if(!strcmp(str, "0"))
+   else if(!util_strcmp(str, "0"))
       result = FALSE;
-   else if(!strcmp(str, "f"))
+   else if(!util_strcmp(str, "f"))
       result = FALSE;
-   else if(!strcmp(str, "false"))
+   else if(!util_strcmp(str, "false"))
       result = FALSE;
    else
       result = TRUE;
@@ -244,7 +248,7 @@ debug_get_flags_option(const char *name,
    else {
       result = 0;
       while( flags->name ) {
-	 if (!strcmp(str, "all") || strstr(str, flags->name ))
+	 if (!util_strcmp(str, "all") || util_strstr(str, flags->name ))
 	    result |= flags->value;
 	 ++flags;
       }
@@ -299,10 +303,10 @@ debug_dump_flags(const struct debug_named_value *names,
    while(names->name) {
       if((names->value & value) == names->value) {
 	 if (!first)
-	    strncat(output, "|", sizeof(output));
+	    util_strncat(output, "|", sizeof(output));
 	 else
 	    first = 0;
-	 strncat(output, names->name, sizeof(output));
+	 util_strncat(output, names->name, sizeof(output));
 	 value &= ~names->value;
       }
       ++names;
@@ -310,12 +314,12 @@ debug_dump_flags(const struct debug_named_value *names,
    
    if (value) {
       if (!first)
-	 strncat(output, "|", sizeof(output));
+	 util_strncat(output, "|", sizeof(output));
       else
 	 first = 0;
       
       util_snprintf(rest, sizeof(rest), "0x%08lx", value);
-      strncat(output, rest, sizeof(output));
+      util_strncat(output, rest, sizeof(output));
    }
    
    if(first)
@@ -325,101 +329,159 @@ debug_dump_flags(const struct debug_named_value *names,
 }
 
 
+static const struct debug_named_value pipe_format_names[] = {
+#ifdef DEBUG
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_NONE),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8R8G8B8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8R8G8B8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8A8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8X8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A1R5G5B5_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A4R4G4B4_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R5G6B5_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_I8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8L8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR_REV),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_S8Z24_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z24S8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8Z24_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z24X8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_S8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64B64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64B64A64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_L8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT3_RGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT5_RGBA),
+#endif
+   DEBUG_NAMED_VALUE_END
+};
 
-
-
+#ifdef DEBUG
+void debug_print_format(const char *msg, unsigned fmt )
+{
+   debug_printf("%s: %s\n", msg, debug_dump_enum(pipe_format_names, fmt)); 
+}
+#endif
 
 char *pf_sprint_name( char *str, enum pipe_format format )
 {
-   strcpy( str, "PIPE_FORMAT_" );
-   switch (pf_layout( format )) {
-   case PIPE_FORMAT_LAYOUT_RGBAZS:
-      {
-         pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format;
-         uint                 i;
-         uint                 scale = 1 << (pf_exp8( rgbazs ) * 3);
-
-         for (i = 0; i < 4; i++) {
-            uint  size = pf_size_xyzw( rgbazs, i );
-
-            if (size == 0) {
-               break;
-            }
-            switch (pf_swizzle_xyzw( rgbazs, i )) {
-            case PIPE_FORMAT_COMP_R:
-               strcat( str, "R" );
-               break;
-            case PIPE_FORMAT_COMP_G:
-               strcat( str, "G" );
-               break;
-            case PIPE_FORMAT_COMP_B:
-               strcat( str, "B" );
-               break;
-            case PIPE_FORMAT_COMP_A:
-               strcat( str, "A" );
-               break;
-            case PIPE_FORMAT_COMP_0:
-               strcat( str, "0" );
-               break;
-            case PIPE_FORMAT_COMP_1:
-               strcat( str, "1" );
-               break;
-            case PIPE_FORMAT_COMP_Z:
-               strcat( str, "Z" );
-               break;
-            case PIPE_FORMAT_COMP_S:
-               strcat( str, "S" );
-               break;
-            }
-            util_snprintf( &str[strlen( str )], 32, "%u", size * scale );
-         }
-         if (i != 0) {
-            strcat( str, "_" );
-         }
-         switch (pf_type( rgbazs )) {
-         case PIPE_FORMAT_TYPE_UNKNOWN:
-            strcat( str, "NONE" );
-            break;
-         case PIPE_FORMAT_TYPE_FLOAT:
-            strcat( str, "FLOAT" );
-            break;
-         case PIPE_FORMAT_TYPE_UNORM:
-            strcat( str, "UNORM" );
-            break;
-         case PIPE_FORMAT_TYPE_SNORM:
-            strcat( str, "SNORM" );
-            break;
-         case PIPE_FORMAT_TYPE_USCALED:
-            strcat( str, "USCALED" );
-            break;
-         case PIPE_FORMAT_TYPE_SSCALED:
-            strcat( str, "SSCALED" );
-            break;
-         }
-      }
-      break;
-   case PIPE_FORMAT_LAYOUT_YCBCR:
-      {
-         pipe_format_ycbcr_t  ycbcr = (pipe_format_ycbcr_t) format;
-
-         strcat( str, "YCBCR" );
-         if (pf_rev( ycbcr )) {
-            strcat( str, "_REV" );
-         }
-      }
-      break;
-   }
+   strcpy( str, debug_dump_enum(pipe_format_names, format) );
    return str;
 }
 
 
 #ifdef DEBUG
-void debug_print_format(const char *msg, unsigned fmt )
+void debug_dump_image(const char *prefix,
+                      unsigned format, unsigned cpp,
+                      unsigned width, unsigned height,
+                      unsigned pitch,
+                      const void *data)     
 {
-   char fmtstr[80];
- 
-   pf_sprint_name(fmtstr, (enum pipe_format)fmt);
+#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
+   static unsigned no = 0; 
+   char filename[256];
+   WCHAR wfilename[sizeof(filename)];
+   ULONG_PTR iFile = 0;
+   struct {
+      unsigned format;
+      unsigned cpp;
+      unsigned width;
+      unsigned height;
+   } header;
+   unsigned char *pMap = NULL;
+   unsigned i;
 
-   debug_printf("%s: %s\n", msg, fmtstr); 
+   util_snprintf(filename, sizeof(filename), "\\??\\c:\\%03u%s.raw", ++no, prefix);
+   for(i = 0; i < sizeof(filename); ++i)
+      wfilename[i] = (WCHAR)filename[i];
+   
+   pMap = (unsigned char *)EngMapFile(wfilename, sizeof(header) + cpp*width*height, &iFile);
+   if(!pMap)
+      return;
+   
+   header.format = format;
+   header.cpp = cpp;
+   header.width = width;
+   header.height = height;
+   memcpy(pMap, &header, sizeof(header));
+   pMap += sizeof(header);
+   
+   for(i = 0; i < height; ++i) {
+      memcpy(pMap, (unsigned char *)data + cpp*pitch*i, cpp*width);
+      pMap += cpp*width;
+   }
+      
+   EngUnmapFile(iFile);
+#endif
 }
 #endif
diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c
index 3b5e4fbaeea..78497c5f6af 100644
--- a/src/gallium/auxiliary/util/p_debug_mem.c
+++ b/src/gallium/auxiliary/util/p_debug_mem.c
@@ -211,6 +211,7 @@ debug_memory_begin(void)
 void 
 debug_memory_end(unsigned long start_no)
 {
+   size_t total_size = 0;
    struct list_head *entry;
 
    entry = list.prev;
@@ -220,9 +221,15 @@ debug_memory_end(unsigned long start_no)
       hdr = LIST_ENTRY(struct debug_memory_header, entry, head);
       ptr = data_from_header(hdr);
       if(start_no <= hdr->no && hdr->no < last_no ||
-	 last_no < start_no && (hdr->no < last_no || start_no <= hdr->no))
+	 last_no < start_no && (hdr->no < last_no || start_no <= hdr->no)) {
 	 debug_printf("%s:%u:%s: %u bytes at %p not freed\n",
 		      hdr->file, hdr->line, hdr->function,
 		      hdr->size, ptr);
+	 total_size += hdr->size;
+      }
+   }
+   if(total_size) {
+      debug_printf("Total of %u KB of system memory apparently leaked\n",
+		   (total_size + 1023)/1024);
    }
 }
diff --git a/src/gallium/auxiliary/util/p_debug_prof.c b/src/gallium/auxiliary/util/p_debug_prof.c
new file mode 100644
index 00000000000..958f99c327c
--- /dev/null
+++ b/src/gallium/auxiliary/util/p_debug_prof.c
@@ -0,0 +1,175 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Poor-man profiling.
+ * 
+ * @author José Fonseca <[email protected]>
+ * 
+ * @sa http://blogs.msdn.com/joshpoley/archive/2008/03/12/poor-man-s-profiler.aspx
+ * @sa http://www.johnpanzer.com/aci_cuj/index.html
+ */
+
+#include "pipe/p_config.h" 
+
+#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+
+#include <windows.h>
+#include <winddi.h>
+
+#include "pipe/p_debug.h" 
+#include "util/u_string.h" 
+
+
+#define PROFILE_FILE_SIZE 4*1024*1024
+#define FILE_NAME_SIZE 256
+
+static WCHAR wFileName[FILE_NAME_SIZE];
+static ULONG_PTR iFile = 0;
+static void *pMap = NULL;
+static void *pMapEnd = NULL;
+
+
+/**
+ * Called at the start of every method or function.
+ * 
+ * @sa http://msdn.microsoft.com/en-us/library/c63a9b7h.aspx
+ */
+void __declspec(naked) __cdecl 
+_penter(void) {
+   _asm {
+      push ebx
+      mov ebx, [pMap]
+      test ebx, ebx
+      jz done
+      cmp ebx, [pMapEnd]
+      je done
+      push eax
+      push edx
+      mov eax, [esp+12]
+      and eax, 0xfffffffe
+      mov [ebx], eax
+      add ebx, 4
+      rdtsc
+      mov [ebx], eax
+      add ebx, 4
+      mov [pMap], ebx
+      pop edx
+      pop eax
+done:
+      pop ebx
+      ret  
+   }
+}
+
+
+/**
+ * Called at the end of Calls the end of every method or function.
+ * 
+ * @sa http://msdn.microsoft.com/en-us/library/xc11y76y.aspx
+ */
+void __declspec(naked) __cdecl 
+_pexit(void) {
+   _asm {
+      push ebx
+      mov ebx, [pMap]
+      test ebx, ebx
+      jz done
+      cmp ebx, [pMapEnd]
+      je done
+      push eax
+      push edx
+      mov eax, [esp+12]
+      or eax, 0x00000001
+      mov [ebx], eax
+      add ebx, 4
+      rdtsc
+      mov [ebx], eax
+      add ebx, 4
+      mov [pMap], ebx
+      pop edx
+      pop eax
+done:
+      pop ebx
+      ret
+   }
+}
+
+
+void __declspec(naked) 
+__debug_profile_reference1(void) {
+   _asm {
+      call _penter
+      call _pexit
+      ret
+   }
+}
+
+
+void __declspec(naked) 
+__debug_profile_reference2(void) {
+   _asm {
+      call _penter
+      call __debug_profile_reference1
+      call _pexit
+      ret
+   }
+}
+
+
+void
+debug_profile_start(void)
+{
+   static unsigned no = 0; 
+   char filename[FILE_NAME_SIZE];
+   unsigned i;
+
+   util_snprintf(filename, sizeof(filename), "\\??\\c:\\%03u.prof", ++no);
+   for(i = 0; i < FILE_NAME_SIZE; ++i)
+      wFileName[i] = (WCHAR)filename[i];
+   
+   pMap = EngMapFile(wFileName, PROFILE_FILE_SIZE, &iFile);
+   if(pMap) {
+      pMapEnd = (unsigned char*)pMap + PROFILE_FILE_SIZE;
+      /* reference functions for calibration purposes */
+      __debug_profile_reference2();
+   }
+}
+
+void 
+debug_profile_stop(void)
+{
+   if(iFile) {
+      EngUnmapFile(iFile);
+      /* TODO: truncate file */
+   }
+   iFile = 0;
+   pMapEnd = pMap = NULL; 
+}
+
+#endif /* PROFILE */
diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c
index 63e1cc6013a..5728757d2fb 100644
--- a/src/gallium/auxiliary/util/p_tile.c
+++ b/src/gallium/auxiliary/util/p_tile.c
@@ -50,6 +50,7 @@ pipe_get_tile_raw(struct pipe_context *pipe,
                   uint x, uint y, uint w, uint h,
                   void *p, int dst_stride)
 {
+   struct pipe_screen *screen = pipe->screen;
    const uint cpp = ps->cpp;
    const ubyte *pSrc;
    const uint src_stride = ps->pitch * cpp;
@@ -63,7 +64,11 @@ pipe_get_tile_raw(struct pipe_context *pipe,
    if (pipe_clip_tile(x, y, &w, &h, ps))
       return;
 
-   pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp;
+   pSrc = (const ubyte *) screen->surface_map(screen, ps,
+                                              PIPE_BUFFER_USAGE_CPU_READ);
+   assert(pSrc);                /* XXX: proper error handling! */
+
+   pSrc += (y * ps->pitch + x) * cpp;
    pDest = (ubyte *) p;
 
    for (i = 0; i < h; i++) {
@@ -72,7 +77,7 @@ pipe_get_tile_raw(struct pipe_context *pipe,
       pSrc += src_stride;
    }
 
-   pipe_surface_unmap(ps);
+   screen->surface_unmap(screen, ps);
 }
 
 
@@ -86,6 +91,7 @@ pipe_put_tile_raw(struct pipe_context *pipe,
                   uint x, uint y, uint w, uint h,
                   const void *p, int src_stride)
 {
+   struct pipe_screen *screen = pipe->screen;
    const uint cpp = ps->cpp;
    const ubyte *pSrc;
    const uint dst_stride = ps->pitch * cpp;
@@ -100,7 +106,11 @@ pipe_put_tile_raw(struct pipe_context *pipe,
       return;
 
    pSrc = (const ubyte *) p;
-   pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp;
+
+   pDest = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE);
+   assert(pDest);               /* XXX: proper error handling */
+
+   pDest += (y * ps->pitch + x) * cpp;
 
    for (i = 0; i < h; i++) {
       memcpy(pDest, pSrc, w * cpp);
@@ -108,7 +118,7 @@ pipe_put_tile_raw(struct pipe_context *pipe,
       pSrc += src_stride;
    }
 
-   pipe_surface_unmap(ps);
+   screen->surface_unmap(screen, ps);
 }
 
 
@@ -834,18 +844,26 @@ pipe_get_tile_z(struct pipe_context *pipe,
                 uint x, uint y, uint w, uint h,
                 uint *z)
 {
+   struct pipe_screen *screen = pipe->screen;
    const uint dstStride = w;
+   void *map;
    uint *pDest = z;
    uint i, j;
 
    if (pipe_clip_tile(x, y, &w, &h, ps))
       return;
 
+   map = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ);
+   if (!map) {
+      assert(0);
+      return;
+   }
+
    switch (ps->format) {
    case PIPE_FORMAT_Z32_UNORM:
       {
          const uint *pSrc
-            = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x);
+            = (const uint *)map  + (y * ps->pitch + x);
          for (i = 0; i < h; i++) {
             memcpy(pDest, pSrc, 4 * w);
             pDest += dstStride;
@@ -857,7 +875,7 @@ pipe_get_tile_z(struct pipe_context *pipe,
    case PIPE_FORMAT_X8Z24_UNORM:
       {
          const uint *pSrc
-            = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x);
+            = (const uint *)map + (y * ps->pitch + x);
          for (i = 0; i < h; i++) {
             for (j = 0; j < w; j++) {
                /* convert 24-bit Z to 32-bit Z */
@@ -871,7 +889,7 @@ pipe_get_tile_z(struct pipe_context *pipe,
    case PIPE_FORMAT_Z16_UNORM:
       {
          const ushort *pSrc
-            = (const ushort *) pipe_surface_map(ps) + (y * ps->pitch + x);
+            = (const ushort *)map + (y * ps->pitch + x);
          for (i = 0; i < h; i++) {
             for (j = 0; j < w; j++) {
                /* convert 16-bit Z to 32-bit Z */
@@ -886,7 +904,7 @@ pipe_get_tile_z(struct pipe_context *pipe,
       assert(0);
    }
 
-   pipe_surface_unmap(ps);
+   screen->surface_unmap(screen, ps);
 }
 
 
@@ -896,17 +914,25 @@ pipe_put_tile_z(struct pipe_context *pipe,
                 uint x, uint y, uint w, uint h,
                 const uint *zSrc)
 {
+   struct pipe_screen *screen = pipe->screen;
    const uint srcStride = w;
    const uint *pSrc = zSrc;
+   void *map;
    uint i, j;
 
    if (pipe_clip_tile(x, y, &w, &h, ps))
       return;
 
+   map = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE);
+   if (!map) {
+      assert(0);
+      return;
+   }
+
    switch (ps->format) {
    case PIPE_FORMAT_Z32_UNORM:
       {
-         uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x);
+         uint *pDest = (uint *) map + (y * ps->pitch + x);
          for (i = 0; i < h; i++) {
             memcpy(pDest, pSrc, 4 * w);
             pDest += ps->pitch;
@@ -917,7 +943,7 @@ pipe_put_tile_z(struct pipe_context *pipe,
    case PIPE_FORMAT_S8Z24_UNORM:
    case PIPE_FORMAT_X8Z24_UNORM:
       {
-         uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x);
+         uint *pDest = (uint *) map + (y * ps->pitch + x);
          for (i = 0; i < h; i++) {
             for (j = 0; j < w; j++) {
                /* convert 32-bit Z to 24-bit Z (0 stencil) */
@@ -930,7 +956,7 @@ pipe_put_tile_z(struct pipe_context *pipe,
       break;
    case PIPE_FORMAT_Z16_UNORM:
       {
-         ushort *pDest = (ushort *) pipe_surface_map(ps) + (y * ps->pitch + x);
+         ushort *pDest = (ushort *) map + (y * ps->pitch + x);
          for (i = 0; i < h; i++) {
             for (j = 0; j < w; j++) {
                /* convert 32-bit Z to 16-bit Z */
@@ -945,7 +971,7 @@ pipe_put_tile_z(struct pipe_context *pipe,
       assert(0);
    }
 
-   pipe_surface_unmap(ps);
+   screen->surface_unmap(screen, ps);
 }
 
 
diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c
index 568d62ced19..999a3e50995 100644
--- a/src/gallium/auxiliary/util/u_blit.c
+++ b/src/gallium/auxiliary/util/u_blit.c
@@ -287,7 +287,8 @@ util_blit_pixels(struct blit_state *ctx,
    if (!tex)
       return;
 
-   texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0);
+   texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0, 
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
 
    /* load temp texture */
    pipe->surface_copy(pipe, FALSE,
@@ -295,7 +296,9 @@ util_blit_pixels(struct blit_state *ctx,
                       src, srcLeft, srcTop, /* src */
                       srcW, srcH);     /* size */
 
-   pipe->texture_update(pipe, tex, 0, 1 << 0);
+   /* free the surface, update the texture if necessary.
+    */
+   screen->tex_surface_release(screen, &texSurf);
 
    /* save state (restored below) */
    cso_save_blend(ctx->cso);
@@ -356,8 +359,6 @@ util_blit_pixels(struct blit_state *ctx,
    cso_restore_vertex_shader(ctx->cso);
    cso_restore_viewport(ctx->cso);
 
-   /* free the texture */
-   pipe_surface_reference(&texSurf, NULL);
    screen->texture_release(screen, &tex);
 }
 
diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c
index e659edb0881..bf143815d8b 100644
--- a/src/gallium/auxiliary/util/u_draw_quad.c
+++ b/src/gallium/auxiliary/util/u_draw_quad.c
@@ -82,52 +82,51 @@ util_draw_texquad(struct pipe_context *pipe,
 {
    struct pipe_buffer *vbuf;
    uint numAttribs = 2, vertexBytes, i, j;
-   float *v;
 
    vertexBytes = 4 * (4 * numAttribs * sizeof(float));
 
    /* XXX create one-time */
    vbuf = pipe->winsys->buffer_create(pipe->winsys, 32,
                                       PIPE_BUFFER_USAGE_VERTEX, vertexBytes);
-   assert(vbuf);
-
-   v = (float *) pipe->winsys->buffer_map(pipe->winsys, vbuf,
-                                          PIPE_BUFFER_USAGE_CPU_WRITE);
-
-   /*
-    * Load vertex buffer
-    */
-   for (i = j = 0; i < 4; i++) {
-      v[j + 2] = z;   /* z */
-      v[j + 3] = 1.0; /* w */
-      v[j + 6] = 0.0; /* r */
-      v[j + 7] = 1.0; /* q */
-      j += 8;
+   if (vbuf) {
+      float *v = (float *) pipe->winsys->buffer_map(pipe->winsys, vbuf,
+                                             PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (v) {
+         /*
+          * Load vertex buffer
+          */
+         for (i = j = 0; i < 4; i++) {
+            v[j + 2] = z;   /* z */
+            v[j + 3] = 1.0; /* w */
+            v[j + 6] = 0.0; /* r */
+            v[j + 7] = 1.0; /* q */
+            j += 8;
+         }
+
+         v[0] = x0;
+         v[1] = y0;
+         v[4] = 0.0; /*s*/
+         v[5] = 0.0; /*t*/
+
+         v[8] = x1;
+         v[9] = y0;
+         v[12] = 1.0;
+         v[13] = 0.0;
+
+         v[16] = x1;
+         v[17] = y1;
+         v[20] = 1.0;
+         v[21] = 1.0;
+
+         v[24] = x0;
+         v[25] = y1;
+         v[28] = 0.0;
+         v[29] = 1.0;
+
+         pipe->winsys->buffer_unmap(pipe->winsys, vbuf);
+         util_draw_vertex_buffer(pipe, vbuf, PIPE_PRIM_TRIANGLE_FAN, 4, 2);
+      }
+
+      pipe_buffer_reference(pipe->winsys, &vbuf, NULL);
    }
-
-   v[0] = x0;
-   v[1] = y0;
-   v[4] = 0.0; /*s*/
-   v[5] = 0.0; /*t*/
-
-   v[8] = x1;
-   v[9] = y0;
-   v[12] = 1.0;
-   v[13] = 0.0;
-
-   v[16] = x1;
-   v[17] = y1;
-   v[20] = 1.0;
-   v[21] = 1.0;
-
-   v[24] = x0;
-   v[25] = y1;
-   v[28] = 0.0;
-   v[29] = 1.0;
-
-   pipe->winsys->buffer_unmap(pipe->winsys, vbuf);
-
-   util_draw_vertex_buffer(pipe, vbuf, PIPE_PRIM_TRIANGLE_FAN, 4, 2);
-
-   pipe_buffer_reference(pipe->winsys, &vbuf, NULL);
 }
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index c53c5122681..7d71aefda9d 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -589,8 +589,11 @@ make_1d_mipmap(struct gen_mipmap_state *ctx,
       struct pipe_surface *srcSurf, *dstSurf;
       void *srcMap, *dstMap;
       
-      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice);
-      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice);
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
 
       srcMap = ((ubyte *) winsys->buffer_map(winsys, srcSurf->buffer,
                                             PIPE_BUFFER_USAGE_CPU_READ)
@@ -629,8 +632,10 @@ make_2d_mipmap(struct gen_mipmap_state *ctx,
       struct pipe_surface *srcSurf, *dstSurf;
       ubyte *srcMap, *dstMap;
       
-      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice);
-      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice);
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
 
       srcMap = ((ubyte *) winsys->buffer_map(winsys, srcSurf->buffer,
                                             PIPE_BUFFER_USAGE_CPU_READ)
@@ -891,10 +896,14 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
    for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
       const uint srcLevel = dstLevel - 1;
 
+      struct pipe_surface *surf = 
+         screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                 PIPE_BUFFER_USAGE_GPU_WRITE);
+
       /*
        * Setup framebuffer / dest surface
        */
-      fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice);
+      fb.cbufs[0] = surf;
       fb.width = pt->width[dstLevel];
       fb.height = pt->height[dstLevel];
       cso_set_framebuffer(ctx->cso, &fb);
@@ -925,7 +934,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
       pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
 
       /* need to signal that the texture has changed _after_ rendering to it */
-      pipe->texture_update(pipe, pt, face, (1 << dstLevel));
+      pipe_surface_reference( &surf, NULL );
    }
 
    /* restore state we changed */
diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c
index f3f16a8d94f..dd5eca7fca3 100644
--- a/src/gallium/auxiliary/util/u_hash_table.c
+++ b/src/gallium/auxiliary/util/u_hash_table.c
@@ -67,6 +67,13 @@ struct hash_table_item
 };
 
 
+static INLINE struct hash_table_item *
+hash_table_item(struct cso_hash_iter iter)
+{
+   return (struct hash_table_item *)cso_hash_iter_data(iter);
+}
+
+
 struct hash_table *
 hash_table_create(unsigned (*hash)(void *key),
                   int (*compare)(void *key1, void *key2))
@@ -90,7 +97,27 @@ hash_table_create(unsigned (*hash)(void *key),
 }
 
 
-static struct hash_table_item *
+static INLINE struct cso_hash_iter
+hash_table_find_iter(struct hash_table *ht,
+                     void *key, 
+                     unsigned key_hash)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+   
+   iter = cso_hash_find(ht->cso, key_hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      if (!ht->compare(item->key, key))
+         break;
+      iter = cso_hash_iter_next(iter);
+   }
+   
+   return iter;
+}
+
+
+static INLINE struct hash_table_item *
 hash_table_find_item(struct hash_table *ht,
                      void *key, 
                      unsigned key_hash)
@@ -117,6 +144,7 @@ hash_table_set(struct hash_table *ht,
 {
    unsigned key_hash;
    struct hash_table_item *item;
+   struct cso_hash_iter iter;
 
    assert(ht);
 
@@ -136,9 +164,8 @@ hash_table_set(struct hash_table *ht,
    item->key = key;
    item->value = value;
    
-   cso_hash_insert(ht->cso, key_hash, item);
-   /* FIXME: there is no OOM propagation in cso_hash */
-   if(0) {
+   iter = cso_hash_insert(ht->cso, key_hash, item);
+   if(cso_hash_iter_is_null(iter)) {
       FREE(item);
       return PIPE_ERROR_OUT_OF_MEMORY;
    }
@@ -171,19 +198,39 @@ hash_table_remove(struct hash_table *ht,
                   void *key)
 {
    unsigned key_hash;
+   struct cso_hash_iter iter;
    struct hash_table_item *item;
 
    assert(ht);
 
    key_hash = ht->hash(key);
 
-   item = hash_table_find_item(ht, key, key_hash);
-   if(!item)
+   iter = hash_table_find_iter(ht, key, key_hash);
+   if(cso_hash_iter_is_null(iter))
       return;
    
-   /* FIXME: cso_hash_take takes the first element of the collision list 
-    * indiscriminately, so we can not take the item down. */
-   item->value = NULL;
+   item = hash_table_item(iter);
+   assert(item);
+   FREE(item);
+   
+   cso_hash_erase(ht->cso, iter);
+}
+
+
+void 
+hash_table_clear(struct hash_table *ht)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+
+   assert(ht);
+   
+   iter = cso_hash_first_node(ht->cso);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter));
+      FREE(item);
+      iter = cso_hash_first_node(ht->cso);
+   }
 }
 
 
@@ -196,6 +243,8 @@ hash_table_foreach(struct hash_table *ht,
    struct hash_table_item *item;
    enum pipe_error result;
    
+   assert(ht);
+   
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_iter_data(iter);
@@ -212,7 +261,17 @@ hash_table_foreach(struct hash_table *ht,
 void
 hash_table_destroy(struct hash_table *ht)
 {
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+   
    assert(ht);
+   
+   iter = cso_hash_first_node(ht->cso);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      FREE(item);
+      iter = cso_hash_iter_next(iter);
+   }
 
    cso_hash_delete(ht->cso);
    
diff --git a/src/gallium/auxiliary/util/u_hash_table.h b/src/gallium/auxiliary/util/u_hash_table.h
index 1583bd7548c..feee881582e 100644
--- a/src/gallium/auxiliary/util/u_hash_table.h
+++ b/src/gallium/auxiliary/util/u_hash_table.h
@@ -75,6 +75,10 @@ hash_table_remove(struct hash_table *ht,
                   void *key);
 
 
+void
+hash_table_clear(struct hash_table *ht);
+
+
 enum pipe_error
 hash_table_foreach(struct hash_table *ht,
                    enum pipe_error (*callback)(void *key, void *value, void *data),
diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h
index 0b917c005f1..655e2c82592 100644
--- a/src/gallium/auxiliary/util/u_pack_color.h
+++ b/src/gallium/auxiliary/util/u_pack_color.h
@@ -101,6 +101,19 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a,
          *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
       }
       return;
+   case PIPE_FORMAT_A8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = a;
+      }
+      return;
+   case PIPE_FORMAT_L8_UNORM:
+   case PIPE_FORMAT_I8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = r;
+      }
+      return;
    case PIPE_FORMAT_R32G32B32A32_FLOAT:
       {
          float *d = (float *) dest;
@@ -198,6 +211,19 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest)
          *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
       }
       return;
+   case PIPE_FORMAT_A8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = a;
+      }
+      return;
+   case PIPE_FORMAT_L8_UNORM:
+   case PIPE_FORMAT_I8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = r;
+      }
+      return;
    case PIPE_FORMAT_R32G32B32A32_FLOAT:
       {
          float *d = (float *) dest;
diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h
index b99d4e8021c..73c88d87b4c 100644
--- a/src/gallium/auxiliary/util/u_string.h
+++ b/src/gallium/auxiliary/util/u_string.h
@@ -41,6 +41,8 @@
 #include <stddef.h>
 #include <stdarg.h>
 
+#include "pipe/p_compiler.h"
+
 
 #ifdef __cplusplus
 extern "C" {
@@ -48,11 +50,129 @@ extern "C" {
 
    
 #ifdef WIN32
+   
 int util_vsnprintf(char *, size_t, const char *, va_list);
 int util_snprintf(char *str, size_t size, const char *format, ...);
+
+static INLINE void 
+util_vsprintf(char *str, const char *format, va_list ap)
+{
+   util_vsnprintf(str, (size_t)-1, format, ap);
+}
+
+static INLINE void 
+util_sprintf(char *str, const char *format, ...)
+{
+   va_list ap;
+   va_start(ap, format);
+   util_vsnprintf(str, (size_t)-1, format, ap);
+   va_end(ap);
+}
+
+static INLINE char *
+util_strchr(const char *s, char c)
+{
+   while(*s) {
+      if(*s == c)
+	 return (char *)s;
+      ++s;
+   }
+   return NULL;
+}
+
+static INLINE char*
+util_strncat(char *dst, const char *src, size_t n)
+{
+   char *p = dst + strlen(dst);
+   const char *q = src;
+   size_t i;
+
+   for (i = 0; i < n && *q != '\0'; ++i)
+       *p++ = *q++;
+   *p = '\0';
+
+   return dst;
+}
+
+static INLINE int
+util_strcmp(const char *s1, const char *s2)
+{
+   unsigned char u1, u2;
+
+   while (1) {
+      u1 = (unsigned char) *s1++;
+      u2 = (unsigned char) *s2++;
+      if (u1 != u2)
+	 return u1 - u2;
+      if (u1 == '\0')
+	 return 0;
+   }
+   return 0;
+}
+
+static INLINE int
+util_strncmp(const char *s1, const char *s2, size_t n)
+{
+   unsigned char u1, u2;
+
+   while (n-- > 0) {
+      u1 = (unsigned char) *s1++;
+      u2 = (unsigned char) *s2++;
+      if (u1 != u2)
+	 return u1 - u2;
+      if (u1 == '\0')
+	 return 0;
+   }
+   return 0;
+}
+
+static INLINE char *
+util_strstr(const char *haystack, const char *needle)
+{
+   const char *p = haystack;
+   int len = strlen(needle);
+
+   for (; (p = util_strchr(p, *needle)) != 0; p++) {
+      if (util_strncmp(p, needle, len) == 0) {
+	 return (char *)p;
+      }
+   }
+   return NULL;
+}
+
+static INLINE void *
+util_memmove(void *dest, const void *src, size_t n)
+{
+   char *p = (char *)dest;
+   const char *q = (const char *)src;
+   if (dest < src) {
+      while (n--)
+	 *p++ = *q++;
+   }
+   else
+   {
+      p += n;
+      q += n;
+      while (n--)
+	 *--p = *--q;
+   }
+   return dest;
+}
+
+
 #else
+
 #define util_vsnprintf vsnprintf
 #define util_snprintf snprintf
+#define util_vsprintf vsprintf
+#define util_sprintf sprintf
+#define util_strchr strchr
+#define util_strcmp strcmp
+#define util_strncmp strncmp
+#define util_strncat strncat
+#define util_strstr strstr
+#define util_memmove memmove
+
 #endif
 
 
diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c
index cb95ba516f9..014a3e31d50 100644
--- a/src/gallium/drivers/failover/fo_context.c
+++ b/src/gallium/drivers/failover/fo_context.c
@@ -147,8 +147,8 @@ struct pipe_context *failover_create( struct pipe_context *hw,
    failover->pipe.texture_create = hw->texture_create;
    failover->pipe.texture_release = hw->texture_release;
    failover->pipe.get_tex_surface = hw->get_tex_surface;
-#endif
    failover->pipe.texture_update = hw->texture_update;
+#endif
 
    failover->pipe.flush = hw->flush;
 
diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h
index 38e6a348845..53fc5ed0795 100644
--- a/src/gallium/drivers/i915simple/i915_context.h
+++ b/src/gallium/drivers/i915simple/i915_context.h
@@ -1,341 +1,341 @@
- /**************************************************************************

- * 

- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.

- * All Rights Reserved.

- * 

- * Permission is hereby granted, free of charge, to any person obtaining a

- * copy of this software and associated documentation files (the

- * "Software"), to deal in the Software without restriction, including

- * without limitation the rights to use, copy, modify, merge, publish,

- * distribute, sub license, and/or sell copies of the Software, and to

- * permit persons to whom the Software is furnished to do so, subject to

- * the following conditions:

- * 

- * The above copyright notice and this permission notice (including the

- * next paragraph) shall be included in all copies or substantial portions

- * of the Software.

- * 

- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS

- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.

- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR

- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,

- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

- * 

- **************************************************************************/

-

-#ifndef I915_CONTEXT_H

-#define I915_CONTEXT_H

-

-

-#include "pipe/p_context.h"

-#include "pipe/p_defines.h"

-#include "pipe/p_state.h"

-

-#include "draw/draw_vertex.h"

-

-#include "tgsi/util/tgsi_scan.h"

-

-

-#define I915_TEX_UNITS 8

-

-#define I915_DYNAMIC_MODES4       0

-#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */

-#define I915_DYNAMIC_DEPTHSCALE_1 2 

-#define I915_DYNAMIC_IAB          3

-#define I915_DYNAMIC_BC_0         4 /* just the header */

-#define I915_DYNAMIC_BC_1         5

-#define I915_DYNAMIC_BFO_0        6 

-#define I915_DYNAMIC_BFO_1        7

-#define I915_DYNAMIC_STP_0        8 

-#define I915_DYNAMIC_STP_1        9 

-#define I915_DYNAMIC_SC_ENA_0     10 

-#define I915_DYNAMIC_SC_RECT_0    11 

-#define I915_DYNAMIC_SC_RECT_1    12 

-#define I915_DYNAMIC_SC_RECT_2    13 

-#define I915_MAX_DYNAMIC          14

-

-

-#define I915_IMMEDIATE_S0         0

-#define I915_IMMEDIATE_S1         1

-#define I915_IMMEDIATE_S2         2

-#define I915_IMMEDIATE_S3         3

-#define I915_IMMEDIATE_S4         4

-#define I915_IMMEDIATE_S5         5

-#define I915_IMMEDIATE_S6         6

-#define I915_IMMEDIATE_S7         7

-#define I915_MAX_IMMEDIATE        8

-

-/* These must mach the order of LI0_STATE_* bits, as they will be used

- * to generate hardware packets:

- */

-#define I915_CACHE_STATIC         0 

-#define I915_CACHE_DYNAMIC        1 /* handled specially */

-#define I915_CACHE_SAMPLER        2

-#define I915_CACHE_MAP            3

-#define I915_CACHE_PROGRAM        4

-#define I915_CACHE_CONSTANTS      5

-#define I915_MAX_CACHE            6

-

-#define I915_MAX_CONSTANT  32

-

-

-/** See constant_flags[] below */

-#define I915_CONSTFLAG_USER 0x1f

-

-

-/**

- * Subclass of pipe_shader_state

- */

-struct i915_fragment_shader

-{

-   struct pipe_shader_state state;

-

-   struct tgsi_shader_info info;

-

-   uint *program;

-   uint program_len;

-

-   /**

-    * constants introduced during translation.

-    * These are placed at the end of the constant buffer and grow toward

-    * the beginning (eg: slot 31, 30 29, ...)

-    * User-provided constants start at 0.

-    * This allows both types of constants to co-exist (until there's too many)

-    * and doesn't require regenerating/changing the fragment program to

-    * shuffle constants around.

-    */

-   uint num_constants;

-   float constants[I915_MAX_CONSTANT][4];

-

-   /**

-    * Status of each constant

-    * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding

-    * slot of the user's constant buffer. (set by pipe->set_constant_buffer())

-    * Else, the bitmask indicates which components are occupied by immediates.

-    */

-   ubyte constant_flags[I915_MAX_CONSTANT];

-};

-

-

-struct i915_cache_context;

-

-/* Use to calculate differences between state emitted to hardware and

- * current driver-calculated state.  

- */

-struct i915_state 

-{

-   unsigned immediate[I915_MAX_IMMEDIATE];

-   unsigned dynamic[I915_MAX_DYNAMIC];

-

-   float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4];

-   /** number of constants passed in through a constant buffer */

-   uint num_user_constants[PIPE_SHADER_TYPES];

-

-   /* texture sampler state */

-   unsigned sampler[I915_TEX_UNITS][3];

-   unsigned sampler_enable_flags;

-   unsigned sampler_enable_nr;

-

-   /* texture image buffers */

-   unsigned texbuffer[I915_TEX_UNITS][2];

-

-   /** Describes the current hardware vertex layout */

-   struct vertex_info vertex_info;

-   

-   unsigned id;			/* track lost context events */

-};

-

-struct i915_blend_state {

-   unsigned iab;

-   unsigned modes4;

-   unsigned LIS5;

-   unsigned LIS6;

-};

-

-struct i915_depth_stencil_state {

-   unsigned stencil_modes4;

-   unsigned bfo[2];

-   unsigned stencil_LIS5;

-   unsigned depth_LIS6;

-};

-

-struct i915_rasterizer_state {

-   int light_twoside : 1;

-   unsigned st;

-   enum interp_mode color_interp;

-

-   unsigned LIS4;

-   unsigned LIS7;

-   unsigned sc[1];

-

-   const struct pipe_rasterizer_state *templ;

-

-   union { float f; unsigned u; } ds[2];

-};

-

-struct i915_sampler_state {

-   unsigned state[3];

-   const struct pipe_sampler_state *templ;

-};

-

-

-struct i915_texture {

-   struct pipe_texture base;

-

-   /* Derived from the above:

-    */

-   unsigned pitch;

-   unsigned depth_pitch;          /* per-image on i945? */

-   unsigned total_height;

-

-   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];

-

-   /* Explicitly store the offset of each image for each cube face or

-    * depth value.  Pretty much have to accept that hardware formats

-    * are going to be so diverse that there is no unified way to

-    * compute the offsets of depth/cube images within a mipmap level,

-    * so have to store them as a lookup table:

-    */

-   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */

-

-   /* Includes image offset tables:

-    */

-   unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS];

-

-   /* The data is held here:

-    */

-   struct pipe_buffer *buffer;

-};

-

-struct i915_context

-{

-   struct pipe_context pipe;

-   struct i915_winsys *winsys;

-   struct draw_context *draw;

-

-   /* The most recent drawing state as set by the driver:

-    */

-   const struct i915_blend_state           *blend;

-   const struct i915_sampler_state         *sampler[PIPE_MAX_SAMPLERS];

-   const struct i915_depth_stencil_state   *depth_stencil;

-   const struct i915_rasterizer_state      *rasterizer;

-

-   struct i915_fragment_shader *fs;

-

-   struct pipe_blend_color blend_color;

-   struct pipe_clip_state clip;

-   struct pipe_constant_buffer constants[PIPE_SHADER_TYPES];

-   struct pipe_framebuffer_state framebuffer;

-   struct pipe_poly_stipple poly_stipple;

-   struct pipe_scissor_state scissor;

-   struct i915_texture *texture[PIPE_MAX_SAMPLERS];

-   struct pipe_viewport_state viewport;

-   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];

-

-   unsigned dirty;

-

-   unsigned num_samplers;

-   unsigned num_textures;

-   unsigned num_vertex_elements;

-   unsigned num_vertex_buffers;

-

-   unsigned *batch_start;

-

-   /** Vertex buffer */

-   struct pipe_buffer *vbo;

-

-   struct i915_state current;

-   unsigned hardware_dirty;

-   

-   unsigned debug;

-};

-

-/* A flag for each state_tracker state object:

- */

-#define I915_NEW_VIEWPORT      0x1

-#define I915_NEW_RASTERIZER    0x2

-#define I915_NEW_FS            0x4

-#define I915_NEW_BLEND         0x8

-#define I915_NEW_CLIP          0x10

-#define I915_NEW_SCISSOR       0x20

-#define I915_NEW_STIPPLE       0x40

-#define I915_NEW_FRAMEBUFFER   0x80

-#define I915_NEW_ALPHA_TEST    0x100

-#define I915_NEW_DEPTH_STENCIL 0x200

-#define I915_NEW_SAMPLER       0x400

-#define I915_NEW_TEXTURE       0x800

-#define I915_NEW_CONSTANTS     0x1000

-#define I915_NEW_VBO           0x2000

-#define I915_NEW_VS            0x4000

-

-

-/* Driver's internally generated state flags:

- */

-#define I915_NEW_VERTEX_FORMAT    0x10000

-

-

-/* Dirty flags for hardware emit

- */

-#define I915_HW_STATIC            (1<<I915_CACHE_STATIC)

-#define I915_HW_DYNAMIC           (1<<I915_CACHE_DYNAMIC)

-#define I915_HW_SAMPLER           (1<<I915_CACHE_SAMPLER)

-#define I915_HW_MAP               (1<<I915_CACHE_MAP)

-#define I915_HW_PROGRAM           (1<<I915_CACHE_PROGRAM)

-#define I915_HW_CONSTANTS         (1<<I915_CACHE_CONSTANTS)

-#define I915_HW_IMMEDIATE         (1<<(I915_MAX_CACHE+0))

-#define I915_HW_INVARIENT         (1<<(I915_MAX_CACHE+1))

-

-

-/***********************************************************************

- * i915_prim_emit.c: 

- */

-struct draw_stage *i915_draw_render_stage( struct i915_context *i915 );

-

-

-/***********************************************************************

- * i915_prim_vbuf.c: 

- */

-struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 );

-

-

-/***********************************************************************

- * i915_state_emit.c: 

- */

-void i915_emit_hardware_state(struct i915_context *i915 );

-

-

-

-/***********************************************************************

- * i915_clear.c: 

- */

-void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps,

-		unsigned clearValue);

-

-

-/***********************************************************************

- * i915_surface.c: 

- */

-void i915_init_surface_functions( struct i915_context *i915 );

-

-void i915_init_state_functions( struct i915_context *i915 );

-void i915_init_flush_functions( struct i915_context *i915 );

-void i915_init_string_functions( struct i915_context *i915 );

-

-

-

-

-/***********************************************************************

- * Inline conversion functions.  These are better-typed than the

- * macros used previously:

- */

-static INLINE struct i915_context *

-i915_context( struct pipe_context *pipe )

-{

-   return (struct i915_context *)pipe;

-}

-

-

-

-#endif

+ /**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef I915_CONTEXT_H
+#define I915_CONTEXT_H
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "draw/draw_vertex.h"
+
+#include "tgsi/util/tgsi_scan.h"
+
+
+#define I915_TEX_UNITS 8
+
+#define I915_DYNAMIC_MODES4       0
+#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */
+#define I915_DYNAMIC_DEPTHSCALE_1 2 
+#define I915_DYNAMIC_IAB          3
+#define I915_DYNAMIC_BC_0         4 /* just the header */
+#define I915_DYNAMIC_BC_1         5
+#define I915_DYNAMIC_BFO_0        6 
+#define I915_DYNAMIC_BFO_1        7
+#define I915_DYNAMIC_STP_0        8 
+#define I915_DYNAMIC_STP_1        9 
+#define I915_DYNAMIC_SC_ENA_0     10 
+#define I915_DYNAMIC_SC_RECT_0    11 
+#define I915_DYNAMIC_SC_RECT_1    12 
+#define I915_DYNAMIC_SC_RECT_2    13 
+#define I915_MAX_DYNAMIC          14
+
+
+#define I915_IMMEDIATE_S0         0
+#define I915_IMMEDIATE_S1         1
+#define I915_IMMEDIATE_S2         2
+#define I915_IMMEDIATE_S3         3
+#define I915_IMMEDIATE_S4         4
+#define I915_IMMEDIATE_S5         5
+#define I915_IMMEDIATE_S6         6
+#define I915_IMMEDIATE_S7         7
+#define I915_MAX_IMMEDIATE        8
+
+/* These must mach the order of LI0_STATE_* bits, as they will be used
+ * to generate hardware packets:
+ */
+#define I915_CACHE_STATIC         0 
+#define I915_CACHE_DYNAMIC        1 /* handled specially */
+#define I915_CACHE_SAMPLER        2
+#define I915_CACHE_MAP            3
+#define I915_CACHE_PROGRAM        4
+#define I915_CACHE_CONSTANTS      5
+#define I915_MAX_CACHE            6
+
+#define I915_MAX_CONSTANT  32
+
+
+/** See constant_flags[] below */
+#define I915_CONSTFLAG_USER 0x1f
+
+
+/**
+ * Subclass of pipe_shader_state
+ */
+struct i915_fragment_shader
+{
+   struct pipe_shader_state state;
+
+   struct tgsi_shader_info info;
+
+   uint *program;
+   uint program_len;
+
+   /**
+    * constants introduced during translation.
+    * These are placed at the end of the constant buffer and grow toward
+    * the beginning (eg: slot 31, 30 29, ...)
+    * User-provided constants start at 0.
+    * This allows both types of constants to co-exist (until there's too many)
+    * and doesn't require regenerating/changing the fragment program to
+    * shuffle constants around.
+    */
+   uint num_constants;
+   float constants[I915_MAX_CONSTANT][4];
+
+   /**
+    * Status of each constant
+    * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding
+    * slot of the user's constant buffer. (set by pipe->set_constant_buffer())
+    * Else, the bitmask indicates which components are occupied by immediates.
+    */
+   ubyte constant_flags[I915_MAX_CONSTANT];
+};
+
+
+struct i915_cache_context;
+
+/* Use to calculate differences between state emitted to hardware and
+ * current driver-calculated state.  
+ */
+struct i915_state 
+{
+   unsigned immediate[I915_MAX_IMMEDIATE];
+   unsigned dynamic[I915_MAX_DYNAMIC];
+
+   float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4];
+   /** number of constants passed in through a constant buffer */
+   uint num_user_constants[PIPE_SHADER_TYPES];
+
+   /* texture sampler state */
+   unsigned sampler[I915_TEX_UNITS][3];
+   unsigned sampler_enable_flags;
+   unsigned sampler_enable_nr;
+
+   /* texture image buffers */
+   unsigned texbuffer[I915_TEX_UNITS][2];
+
+   /** Describes the current hardware vertex layout */
+   struct vertex_info vertex_info;
+   
+   unsigned id;			/* track lost context events */
+};
+
+struct i915_blend_state {
+   unsigned iab;
+   unsigned modes4;
+   unsigned LIS5;
+   unsigned LIS6;
+};
+
+struct i915_depth_stencil_state {
+   unsigned stencil_modes4;
+   unsigned bfo[2];
+   unsigned stencil_LIS5;
+   unsigned depth_LIS6;
+};
+
+struct i915_rasterizer_state {
+   int light_twoside : 1;
+   unsigned st;
+   enum interp_mode color_interp;
+
+   unsigned LIS4;
+   unsigned LIS7;
+   unsigned sc[1];
+
+   const struct pipe_rasterizer_state *templ;
+
+   union { float f; unsigned u; } ds[2];
+};
+
+struct i915_sampler_state {
+   unsigned state[3];
+   const struct pipe_sampler_state *templ;
+};
+
+
+struct i915_texture {
+   struct pipe_texture base;
+
+   /* Derived from the above:
+    */
+   unsigned pitch;
+   unsigned depth_pitch;          /* per-image on i945? */
+   unsigned total_height;
+
+   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* Explicitly store the offset of each image for each cube face or
+    * depth value.  Pretty much have to accept that hardware formats
+    * are going to be so diverse that there is no unified way to
+    * compute the offsets of depth/cube images within a mipmap level,
+    * so have to store them as a lookup table:
+    */
+   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */
+
+   /* Includes image offset tables:
+    */
+   unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* The data is held here:
+    */
+   struct pipe_buffer *buffer;
+};
+
+struct i915_context
+{
+   struct pipe_context pipe;
+   struct i915_winsys *winsys;
+   struct draw_context *draw;
+
+   /* The most recent drawing state as set by the driver:
+    */
+   const struct i915_blend_state           *blend;
+   const struct i915_sampler_state         *sampler[PIPE_MAX_SAMPLERS];
+   const struct i915_depth_stencil_state   *depth_stencil;
+   const struct i915_rasterizer_state      *rasterizer;
+
+   struct i915_fragment_shader *fs;
+
+   struct pipe_blend_color blend_color;
+   struct pipe_clip_state clip;
+   struct pipe_constant_buffer constants[PIPE_SHADER_TYPES];
+   struct pipe_framebuffer_state framebuffer;
+   struct pipe_poly_stipple poly_stipple;
+   struct pipe_scissor_state scissor;
+   struct i915_texture *texture[PIPE_MAX_SAMPLERS];
+   struct pipe_viewport_state viewport;
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+
+   unsigned dirty;
+
+   unsigned num_samplers;
+   unsigned num_textures;
+   unsigned num_vertex_elements;
+   unsigned num_vertex_buffers;
+
+   unsigned *batch_start;
+
+   /** Vertex buffer */
+   struct pipe_buffer *vbo;
+
+   struct i915_state current;
+   unsigned hardware_dirty;
+   
+   unsigned debug;
+};
+
+/* A flag for each state_tracker state object:
+ */
+#define I915_NEW_VIEWPORT      0x1
+#define I915_NEW_RASTERIZER    0x2
+#define I915_NEW_FS            0x4
+#define I915_NEW_BLEND         0x8
+#define I915_NEW_CLIP          0x10
+#define I915_NEW_SCISSOR       0x20
+#define I915_NEW_STIPPLE       0x40
+#define I915_NEW_FRAMEBUFFER   0x80
+#define I915_NEW_ALPHA_TEST    0x100
+#define I915_NEW_DEPTH_STENCIL 0x200
+#define I915_NEW_SAMPLER       0x400
+#define I915_NEW_TEXTURE       0x800
+#define I915_NEW_CONSTANTS     0x1000
+#define I915_NEW_VBO           0x2000
+#define I915_NEW_VS            0x4000
+
+
+/* Driver's internally generated state flags:
+ */
+#define I915_NEW_VERTEX_FORMAT    0x10000
+
+
+/* Dirty flags for hardware emit
+ */
+#define I915_HW_STATIC            (1<<I915_CACHE_STATIC)
+#define I915_HW_DYNAMIC           (1<<I915_CACHE_DYNAMIC)
+#define I915_HW_SAMPLER           (1<<I915_CACHE_SAMPLER)
+#define I915_HW_MAP               (1<<I915_CACHE_MAP)
+#define I915_HW_PROGRAM           (1<<I915_CACHE_PROGRAM)
+#define I915_HW_CONSTANTS         (1<<I915_CACHE_CONSTANTS)
+#define I915_HW_IMMEDIATE         (1<<(I915_MAX_CACHE+0))
+#define I915_HW_INVARIENT         (1<<(I915_MAX_CACHE+1))
+
+
+/***********************************************************************
+ * i915_prim_emit.c: 
+ */
+struct draw_stage *i915_draw_render_stage( struct i915_context *i915 );
+
+
+/***********************************************************************
+ * i915_prim_vbuf.c: 
+ */
+struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 );
+
+
+/***********************************************************************
+ * i915_state_emit.c: 
+ */
+void i915_emit_hardware_state(struct i915_context *i915 );
+
+
+
+/***********************************************************************
+ * i915_clear.c: 
+ */
+void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		unsigned clearValue);
+
+
+/***********************************************************************
+ * i915_surface.c: 
+ */
+void i915_init_surface_functions( struct i915_context *i915 );
+
+void i915_init_state_functions( struct i915_context *i915 );
+void i915_init_flush_functions( struct i915_context *i915 );
+void i915_init_string_functions( struct i915_context *i915 );
+
+
+
+
+/***********************************************************************
+ * Inline conversion functions.  These are better-typed than the
+ * macros used previously:
+ */
+static INLINE struct i915_context *
+i915_context( struct pipe_context *pipe )
+{
+   return (struct i915_context *)pipe;
+}
+
+
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c
index 646cfd921d1..ba8f183bdf0 100644
--- a/src/gallium/drivers/i915simple/i915_screen.c
+++ b/src/gallium/drivers/i915simple/i915_screen.c
@@ -201,6 +201,35 @@ i915_destroy_screen( struct pipe_screen *screen )
 }
 
 
+static void *
+i915_surface_map( struct pipe_screen *screen,
+                  struct pipe_surface *surface,
+                  unsigned flags )
+{
+   char *map = screen->winsys->buffer_map( screen->winsys, surface->buffer, flags );
+   if (map == NULL)
+      return NULL;
+
+   if (surface->texture &&
+       (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) 
+   {
+      /* Do something to notify contexts of a texture change.  
+       */
+      /* i915_screen(screen)->timestamp++; */
+   }
+   
+   return map + surface->offset;
+}
+
+static void
+i915_surface_unmap(struct pipe_screen *screen,
+                   struct pipe_surface *surface)
+{
+   screen->winsys->buffer_unmap( screen->winsys, surface->buffer );
+}
+
+
+
 /**
  * Create a new i915_screen object
  */
@@ -244,6 +273,8 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id)
    i915screen->screen.get_param = i915_get_param;
    i915screen->screen.get_paramf = i915_get_paramf;
    i915screen->screen.is_format_supported = i915_is_format_supported;
+   i915screen->screen.surface_map = i915_surface_map;
+   i915screen->screen.surface_unmap = i915_surface_unmap;
 
    i915_init_screen_texture_functions(&i915screen->screen);
 
diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c
index 964bd7c871c..e6c4671700e 100644
--- a/src/gallium/drivers/i915simple/i915_state.c
+++ b/src/gallium/drivers/i915simple/i915_state.c
@@ -466,7 +466,7 @@ void i915_delete_fs_state(struct pipe_context *pipe, void *shader)
       FREE(ifs->program);
    ifs->program_len = 0;
 
-   FREE(ifs->state.tokens);
+   FREE((struct tgsi_token *)ifs->state.tokens);
 
    FREE(ifs);
 }
diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c
index f4fbedbe9bc..98367ac0739 100644
--- a/src/gallium/drivers/i915simple/i915_surface.c
+++ b/src/gallium/drivers/i915simple/i915_surface.c
@@ -51,17 +51,25 @@ i915_surface_copy(struct pipe_context *pipe,
    assert( dst->cpp == src->cpp );
 
    if (0) {
-      pipe_copy_rect(pipe_surface_map(dst),
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+      
+      const void *src_map = pipe->screen->surface_map( pipe->screen,
+                                                       src,
+                                                       PIPE_BUFFER_USAGE_CPU_READ );
+      
+      pipe_copy_rect(dst_map,
                      dst->cpp,
                      dst->pitch,
                      dstx, dsty, 
                      width, height, 
-                     pipe_surface_map(src), 
+                     src_map, 
                      do_flip ? -(int) src->pitch : src->pitch, 
                      srcx, do_flip ? 1 - srcy - height : srcy);
 
-      pipe_surface_unmap(src);
-      pipe_surface_unmap(dst);
+      pipe->screen->surface_unmap(pipe->screen, src);
+      pipe->screen->surface_unmap(pipe->screen, dst);
    }
    else {
       i915_copy_blit( i915_context(pipe),
@@ -92,7 +100,10 @@ i915_surface_fill(struct pipe_context *pipe,
 {
    if (0) {
       unsigned i, j;
-      void *dst_map = pipe_surface_map(dst);
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+
 
       switch (dst->cpp) {
       case 1: {
@@ -126,7 +137,7 @@ i915_surface_fill(struct pipe_context *pipe,
 	 break;
       }
 
-      pipe_surface_unmap( dst );
+      pipe->screen->surface_unmap(pipe->screen, dst);
    }
    else {
       i915_fill_blit( i915_context(pipe),
diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c
index c39e747705b..df11ba0544c 100644
--- a/src/gallium/drivers/i915simple/i915_texture.c
+++ b/src/gallium/drivers/i915simple/i915_texture.c
@@ -105,6 +105,51 @@ i915_miptree_set_image_offset(struct i915_texture *tex,
 }
 
 
+/* Hack it up to use the old winsys->surface_alloc_storage()
+ * method for now:
+ */
+static boolean
+i915_displaytarget_layout(struct pipe_screen *screen,
+                          struct i915_texture *tex)
+{
+   struct pipe_winsys *ws = screen->winsys;
+   struct pipe_surface surf;
+   unsigned flags = (PIPE_BUFFER_USAGE_CPU_READ |
+                     PIPE_BUFFER_USAGE_CPU_WRITE |
+                     PIPE_BUFFER_USAGE_GPU_READ |
+                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+
+   memset(&surf, 0, sizeof(surf));
+
+   ws->surface_alloc_storage( ws, 
+                              &surf,
+                              tex->base.width[0], 
+                              tex->base.height[0],
+                              tex->base.format,
+                              flags,
+                              tex->base.tex_usage);
+      
+   /* Now extract the goodies: 
+    */
+   i915_miptree_set_level_info( tex, 0, 1, 0, 0,
+                                tex->base.width[0],
+                                tex->base.height[0],
+                                1 );
+   i915_miptree_set_image_offset( tex, 0, 0, 0, 0 );
+
+   tex->buffer = surf.buffer;
+   tex->pitch = surf.pitch;
+   tex->total_height = 0;
+
+
+   return tex->buffer != NULL;
+}
+
+
+
+
+
 static void
 i945_miptree_layout_2d( struct i915_texture *tex )
 {
@@ -480,39 +525,54 @@ i945_miptree_layout(struct i915_texture * tex)
 
 
 static struct pipe_texture *
-i915_texture_create_screen(struct pipe_screen *screen,
-                           const struct pipe_texture *templat)
+i915_texture_create(struct pipe_screen *screen,
+                    const struct pipe_texture *templat)
 {
+   struct i915_screen *i915screen = i915_screen(screen);
+   struct pipe_winsys *ws = screen->winsys;
    struct i915_texture *tex = CALLOC_STRUCT(i915_texture);
 
-   if (tex) {
-      struct i915_screen *i915screen = i915_screen(screen);
-      struct pipe_winsys *ws = screen->winsys;
-
-      tex->base = *templat;
-      tex->base.refcount = 1;
-      tex->base.screen = screen;
+   if (!tex) 
+      return NULL;
 
-      if (i915screen->is_i945 ? i945_miptree_layout(tex) :
-	  i915_miptree_layout(tex))
-	 tex->buffer = ws->buffer_create(ws, 64,
-                                         PIPE_BUFFER_USAGE_PIXEL,
-                                         tex->pitch * tex->base.cpp *
-                                         tex->total_height);
+   tex->base = *templat;
+   tex->base.refcount = 1;
+   tex->base.screen = screen;
 
-      if (!tex->buffer) {
-	 FREE(tex);
-	 return NULL;
+   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
+      if (!i915_displaytarget_layout(screen, tex))
+         goto fail;
+   }
+   else {
+      if (i915screen->is_i945) {
+         if (!i945_miptree_layout(tex))
+            goto fail;
       }
+      else {
+         if (!i915_miptree_layout(tex))
+            goto fail;
+      }
+      
+      tex->buffer = ws->buffer_create(ws, 64,
+                                      PIPE_BUFFER_USAGE_PIXEL,
+                                      tex->pitch * tex->base.cpp *
+                                      tex->total_height);
+
+      if (!tex->buffer) 
+         goto fail;
    }
 
    return &tex->base;
+
+ fail:
+   FREE(tex);
+   return NULL;
 }
 
 
 static void
-i915_texture_release_screen(struct pipe_screen *screen,
-                            struct pipe_texture **pt)
+i915_texture_release(struct pipe_screen *screen,
+                     struct pipe_texture **pt)
 {
    if (!*pt)
       return;
@@ -541,21 +601,15 @@ i915_texture_release_screen(struct pipe_screen *screen,
 }
 
 
-static void
-i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture,
-                    uint face, uint levelsMask)
-{
-   /* no-op? */
-}
-
 
 /*
  * XXX note: same as code in sp_surface.c
  */
 static struct pipe_surface *
-i915_get_tex_surface_screen(struct pipe_screen *screen,
-                            struct pipe_texture *pt,
-                            unsigned face, unsigned level, unsigned zslice)
+i915_get_tex_surface(struct pipe_screen *screen,
+                     struct pipe_texture *pt,
+                     unsigned face, unsigned level, unsigned zslice,
+                     unsigned flags)
 {
    struct i915_texture *tex = (struct i915_texture *)pt;
    struct pipe_winsys *ws = screen->winsys;
@@ -579,6 +633,7 @@ i915_get_tex_surface_screen(struct pipe_screen *screen,
    if (ps) {
       assert(ps->refcount);
       assert(ps->winsys);
+      pipe_texture_reference(&ps->texture, pt);
       pipe_buffer_reference(ws, &ps->buffer, tex->buffer);
       ps->format = pt->format;
       ps->cpp = pt->cpp;
@@ -586,6 +641,7 @@ i915_get_tex_surface_screen(struct pipe_screen *screen,
       ps->height = pt->height[level];
       ps->pitch = tex->pitch;
       ps->offset = offset;
+      ps->usage = flags;
    }
    return ps;
 }
@@ -594,14 +650,38 @@ i915_get_tex_surface_screen(struct pipe_screen *screen,
 void
 i915_init_texture_functions(struct i915_context *i915)
 {
-   i915->pipe.texture_update = i915_texture_update;
+//   i915->pipe.texture_update = i915_texture_update;
 }
 
+static void
+i915_tex_surface_release(struct pipe_screen *screen,
+                         struct pipe_surface **surface)
+{
+   struct pipe_surface *surf = *surface;
+
+   if (--surf->refcount == 0) {
+
+      /* This really should not be possible, but it's actually
+       * happening quite a bit...  Will fix.
+       */
+      if (surf->status == PIPE_SURFACE_STATUS_CLEAR) {
+         debug_printf("XXX destroying a surface with pending clears...\n");
+         assert(0);
+      }
+
+      pipe_texture_reference(&surf->texture, NULL);
+      pipe_buffer_reference(screen->winsys, &surf->buffer, NULL);
+      FREE(surf);
+   }
+
+   *surface = NULL;
+}
 
 void
 i915_init_screen_texture_functions(struct pipe_screen *screen)
 {
-   screen->texture_create = i915_texture_create_screen;
-   screen->texture_release = i915_texture_release_screen;
-   screen->get_tex_surface = i915_get_tex_surface_screen;
+   screen->texture_create = i915_texture_create;
+   screen->texture_release = i915_texture_release;
+   screen->get_tex_surface = i915_get_tex_surface;
+   screen->tex_surface_release = i915_tex_surface_release;
 }
diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h
index eeccf367856..8ac6b4e6897 100644
--- a/src/gallium/drivers/i965simple/brw_context.h
+++ b/src/gallium/drivers/i965simple/brw_context.h
@@ -1,684 +1,684 @@
-/*

- 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]>

-  */

-

-

-#ifndef BRWCONTEXT_INC

-#define BRWCONTEXT_INC

-

-

-#include "pipe/p_context.h"

-#include "pipe/p_defines.h"

-#include "pipe/p_state.h"

-

-#include "tgsi/util/tgsi_scan.h"

-

-#include "brw_structs.h"

-#include "brw_winsys.h"

-

-

-/* Glossary:

- *

- * URB - uniform resource buffer.  A mid-sized buffer which is

- * partitioned between the fixed function units and used for passing

- * values (vertices, primitives, constants) between them.

- *

- * CURBE - constant URB entry.  An urb region (entry) used to hold

- * constant values which the fixed function units can be instructed to

- * preload into the GRF when spawining a thread.

- *

- * VUE - vertex URB entry.  An urb entry holding a vertex and usually

- * a vertex header.  The header contains control information and

- * things like primitive type, Begin/end flags and clip codes.

- *

- * PUE - primitive URB entry.  An urb entry produced by the setup (SF)

- * unit holding rasterization and interpolation parameters.

- *

- * GRF - general register file.  One of several register files

- * addressable by programmed threads.  The inputs (r0, payload, curbe,

- * urb) of the thread are preloaded to this area before the thread is

- * spawned.  The registers are individually 8 dwords wide and suitable

- * for general usage.  Registers holding thread input values are not

- * special and may be overwritten.

- *

- * MRF - message register file.  Threads communicate (and terminate)

- * by sending messages.  Message parameters are placed in contigous

- * MRF registers.  All program output is via these messages.  URB

- * entries are populated by sending a message to the shared URB

- * function containing the new data, together with a control word,

- * often an unmodified copy of R0.

- *

- * R0 - GRF register 0.  Typically holds control information used when

- * sending messages to other threads.

- *

- * EU or GEN4 EU: The name of the programmable subsystem of the

- * i965 hardware.  Threads are executed by the EU, the registers

- * described above are part of the EU architecture.

- *

- * Fixed function units:

- *

- * CS - Command streamer.  Notional first unit, little software

- * interaction.  Holds the URB entries used for constant data, ie the

- * CURBEs.

- *

- * VF/VS - Vertex Fetch / Vertex Shader.  The fixed function part of

- * this unit is responsible for pulling vertices out of vertex buffers

- * in vram and injecting them into the processing pipe as VUEs.  If

- * enabled, it first passes them to a VS thread which is a good place

- * for the driver to implement any active vertex shader.

- *

- * GS - Geometry Shader.  This corresponds to a new DX10 concept.  If

- * enabled, incoming strips etc are passed to GS threads in individual

- * line/triangle/point units.  The GS thread may perform arbitary

- * computation and emit whatever primtives with whatever vertices it

- * chooses.  This makes GS an excellent place to implement GL's

- * unfilled polygon modes, though of course it is capable of much

- * more.  Additionally, GS is used to translate away primitives not

- * handled by latter units, including Quads and Lineloops.

- *

- * CS - Clipper.  Mesa's clipping algorithms are imported to run on

- * this unit.  The fixed function part performs cliptesting against

- * the 6 fixed clipplanes and makes descisions on whether or not the

- * incoming primitive needs to be passed to a thread for clipping.

- * User clip planes are handled via cooperation with the VS thread.

- *

- * SF - Strips Fans or Setup: Triangles are prepared for

- * rasterization.  Interpolation coefficients are calculated.

- * Flatshading and two-side lighting usually performed here.

- *

- * WM - Windower.  Interpolation of vertex attributes performed here.

- * Fragment shader implemented here.  SIMD aspects of EU taken full

- * advantage of, as pixels are processed in blocks of 16.

- *

- * CC - Color Calculator.  No EU threads associated with this unit.

- * Handles blending and (presumably) depth and stencil testing.

- */

-

-#define BRW_MAX_CURBE                    (32*16)

-

-struct brw_context;

-struct brw_winsys;

-

-

-/* Raised when we receive new state across the pipe interface:

- */

-#define BRW_NEW_VIEWPORT                0x1

-#define BRW_NEW_RASTERIZER              0x2

-#define BRW_NEW_FS                      0x4

-#define BRW_NEW_BLEND                   0x8

-#define BRW_NEW_CLIP                    0x10

-#define BRW_NEW_SCISSOR                 0x20

-#define BRW_NEW_STIPPLE                 0x40

-#define BRW_NEW_FRAMEBUFFER             0x80

-#define BRW_NEW_ALPHA_TEST              0x100

-#define BRW_NEW_DEPTH_STENCIL           0x200

-#define BRW_NEW_SAMPLER                 0x400

-#define BRW_NEW_TEXTURE                 0x800

-#define BRW_NEW_CONSTANTS               0x1000

-#define BRW_NEW_VBO                     0x2000

-#define BRW_NEW_VS                      0x4000

-

-/* Raised for other internal events:

- */

-#define BRW_NEW_URB_FENCE               0x10000

-#define BRW_NEW_PSP                     0x20000

-#define BRW_NEW_CURBE_OFFSETS           0x40000

-#define BRW_NEW_REDUCED_PRIMITIVE       0x80000

-#define BRW_NEW_PRIMITIVE               0x100000

-#define BRW_NEW_SCENE                 0x200000

-#define BRW_NEW_SF_LINKAGE              0x400000

-

-extern int BRW_DEBUG;

-

-#define DEBUG_TEXTURE	0x1

-#define DEBUG_STATE	0x2

-#define DEBUG_IOCTL	0x4

-#define DEBUG_PRIMS	0x8

-#define DEBUG_VERTS	0x10

-#define DEBUG_FALLBACKS	0x20

-#define DEBUG_VERBOSE	0x40

-#define DEBUG_DRI       0x80

-#define DEBUG_DMA       0x100

-#define DEBUG_SANITY    0x200

-#define DEBUG_SYNC      0x400

-#define DEBUG_SLEEP     0x800

-#define DEBUG_PIXEL     0x1000

-#define DEBUG_STATS     0x2000

-#define DEBUG_TILE      0x4000

-#define DEBUG_SINGLE_THREAD   0x8000

-#define DEBUG_WM        0x10000

-#define DEBUG_URB       0x20000

-#define DEBUG_VS        0x40000

-#define DEBUG_BATCH	0x80000

-#define DEBUG_BUFMGR	0x100000

-#define DEBUG_BLIT	0x200000

-#define DEBUG_REGION	0x400000

-#define DEBUG_MIPTREE	0x800000

-

-#define DBG(...) do {						\

-   if (BRW_DEBUG & FILE_DEBUG_FLAG)				\

-      debug_printf(__VA_ARGS__);				\

-} while(0)

-

-#define PRINT(...) do {						\

-   debug_printf(brw->pipe.winsys, __VA_ARGS__);			\

-} while(0)

-

-struct brw_state_flags {

-   unsigned cache;

-   unsigned brw;

-};

-

-

-struct brw_vertex_program {

-   struct pipe_shader_state program;

-   struct tgsi_shader_info info;

-   int id;

-};

-

-

-struct brw_fragment_program {

-   struct pipe_shader_state program;

-   struct tgsi_shader_info info;

-   

-   boolean UsesDepth; /* XXX add this to tgsi_shader_info? */

-   int id;

-};

-

-

-struct pipe_setup_linkage {

-   struct {

-      unsigned vp_output:5;

-      unsigned interp_mode:4;

-      unsigned bf_vp_output:5;

-   } fp_input[PIPE_MAX_SHADER_INPUTS];

-

-   unsigned fp_input_count:5;

-   unsigned max_vp_output:5;

-};

-   

-

-

-struct brw_texture {

-   struct pipe_texture base;

-

-   /* Derived from the above:

-    */

-   unsigned pitch;

-   unsigned depth_pitch;          /* per-image on i945? */

-   unsigned total_height;

-

-   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];

-

-   /* Explicitly store the offset of each image for each cube face or

-    * depth value.  Pretty much have to accept that hardware formats

-    * are going to be so diverse that there is no unified way to

-    * compute the offsets of depth/cube images within a mipmap level,

-    * so have to store them as a lookup table:

-    */

-   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */

-

-   /* Includes image offset tables:

-    */

-   unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS];

-

-   /* The data is held here:

-    */

-   struct pipe_buffer *buffer;

-};

-

-/* Data about a particular attempt to compile a program.  Note that

- * there can be many of these, each in a different GL state

- * corresponding to a different brw_wm_prog_key struct, with different

- * compiled programs:

- */

-/* Data about a particular attempt to compile a program.  Note that

- * there can be many of these, each in a different GL state

- * corresponding to a different brw_wm_prog_key struct, with different

- * compiled programs:

- */

-

-struct brw_wm_prog_data {

-   unsigned curb_read_length;

-   unsigned urb_read_length;

-

-   unsigned first_curbe_grf;

-   unsigned total_grf;

-   unsigned total_scratch;

-

-   /* Internally generated constants for the CURBE.  These are loaded

-    * ahead of the data from the constant buffer.

-    */

-   const float internal_const[8];

-   unsigned nr_internal_consts;

-   unsigned max_const;

-

-   boolean error;

-};

-

-struct brw_sf_prog_data {

-   unsigned urb_read_length;

-   unsigned total_grf;

-

-   /* Each vertex may have upto 12 attributes, 4 components each,

-    * except WPOS which requires only 2.  (11*4 + 2) == 44 ==> 11

-    * rows.

-    *

-    * Actually we use 4 for each, so call it 12 rows.

-    */

-   unsigned urb_entry_size;

-};

-

-struct brw_clip_prog_data {

-   unsigned curb_read_length;	/* user planes? */

-   unsigned clip_mode;

-   unsigned urb_read_length;

-   unsigned total_grf;

-};

-

-struct brw_gs_prog_data {

-   unsigned urb_read_length;

-   unsigned total_grf;

-};

-

-struct brw_vs_prog_data {

-   unsigned curb_read_length;

-   unsigned urb_read_length;

-   unsigned total_grf;

-   unsigned outputs_written;

-

-   unsigned inputs_read;

-

-   unsigned max_const;

-

-   float    imm_buf[PIPE_MAX_CONSTANT][4];

-   unsigned num_imm;

-   unsigned num_consts;

-

-   /* Used for calculating urb partitions:

-    */

-   unsigned urb_entry_size;

-};

-

-

-#define BRW_MAX_TEX_UNIT 8

-#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1

-

-/* Create a fixed sized struct for caching binding tables:

- */

-struct brw_surface_binding_table {

-   unsigned surf_ss_offset[BRW_WM_MAX_SURF];

-};

-

-

-struct brw_cache;

-

-struct brw_mem_pool {

-   struct pipe_buffer *buffer;

-

-   unsigned size;

-   unsigned offset;		/* offset of first free byte */

-

-   struct brw_context *brw;

-};

-

-struct brw_cache_item {

-   unsigned hash;

-   unsigned key_size;		/* for variable-sized keys */

-   const void *key;

-

-   unsigned offset;		/* offset within pool's buffer */

-   unsigned data_size;

-

-   struct brw_cache_item *next;

-};

-

-

-

-struct brw_cache {

-   unsigned id;

-

-   const char *name;

-

-   struct brw_context *brw;

-   struct brw_mem_pool *pool;

-

-   struct brw_cache_item **items;

-   unsigned size, n_items;

-

-   unsigned key_size;		/* for fixed-size keys */

-   unsigned aux_size;

-

-   unsigned last_addr;			/* offset of active item */

-};

-

-

-

-

-/* Considered adding a member to this struct to document which flags

- * an update might raise so that ordering of the state atoms can be

- * checked or derived at runtime.  Dropped the idea in favor of having

- * a debug mode where the state is monitored for flags which are

- * raised that have already been tested against.

- */

-struct brw_tracked_state {

-   struct brw_state_flags dirty;

-   void (*update)( struct brw_context *brw );

-};

-

-

-/* Flags for brw->state.cache.

- */

-#define CACHE_NEW_CC_VP                  (1<<BRW_CC_VP)

-#define CACHE_NEW_CC_UNIT                (1<<BRW_CC_UNIT)

-#define CACHE_NEW_WM_PROG                (1<<BRW_WM_PROG)

-#define CACHE_NEW_SAMPLER_DEFAULT_COLOR  (1<<BRW_SAMPLER_DEFAULT_COLOR)

-#define CACHE_NEW_SAMPLER                (1<<BRW_SAMPLER)

-#define CACHE_NEW_WM_UNIT                (1<<BRW_WM_UNIT)

-#define CACHE_NEW_SF_PROG                (1<<BRW_SF_PROG)

-#define CACHE_NEW_SF_VP                  (1<<BRW_SF_VP)

-#define CACHE_NEW_SF_UNIT                (1<<BRW_SF_UNIT)

-#define CACHE_NEW_VS_UNIT                (1<<BRW_VS_UNIT)

-#define CACHE_NEW_VS_PROG                (1<<BRW_VS_PROG)

-#define CACHE_NEW_GS_UNIT                (1<<BRW_GS_UNIT)

-#define CACHE_NEW_GS_PROG                (1<<BRW_GS_PROG)

-#define CACHE_NEW_CLIP_VP                (1<<BRW_CLIP_VP)

-#define CACHE_NEW_CLIP_UNIT              (1<<BRW_CLIP_UNIT)

-#define CACHE_NEW_CLIP_PROG              (1<<BRW_CLIP_PROG)

-#define CACHE_NEW_SURFACE                (1<<BRW_SS_SURFACE)

-#define CACHE_NEW_SURF_BIND              (1<<BRW_SS_SURF_BIND)

-

-

-

-

-enum brw_mempool_id {

-   BRW_GS_POOL,

-   BRW_SS_POOL,

-   BRW_MAX_POOL

-};

-

-

-struct brw_cached_batch_item {

-   struct header *header;

-   unsigned sz;

-   struct brw_cached_batch_item *next;

-};

-

-

-

-/* Protect against a future where PIPE_MAX_ATTRIBS > 32.  Wouldn't life

- * be easier if C allowed arrays of packed elements?

- */

-#define ATTRIB_BIT_DWORDS  ((PIPE_MAX_ATTRIBS+31)/32)

-

-

-

-

-struct brw_vertex_info {

-   unsigned varying;  /* varying:1[PIPE_MAX_ATTRIBS] */

-   unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */

-};

-

-

-

-

-

-struct brw_context

-{

-   struct pipe_context pipe;

-   struct brw_winsys *winsys;

-

-   unsigned primitive;

-   unsigned reduced_primitive;

-

-   boolean emit_state_always;

-

-   struct {

-      struct brw_state_flags dirty;

-   } state;

-

-

-   struct {

-      const struct pipe_blend_state         *Blend;

-      const struct pipe_depth_stencil_alpha_state *DepthStencil;

-      const struct pipe_poly_stipple        *PolygonStipple;

-      const struct pipe_rasterizer_state    *Raster;

-      const struct pipe_sampler_state       *Samplers[PIPE_MAX_SAMPLERS];

-      const struct brw_vertex_program       *VertexProgram;

-      const struct brw_fragment_program     *FragmentProgram;

-

-      struct pipe_clip_state          Clip;

-      struct pipe_blend_color         BlendColor;

-      struct pipe_scissor_state       Scissor;

-      struct pipe_viewport_state      Viewport;

-      struct pipe_framebuffer_state   FrameBuffer;

-

-      const struct pipe_constant_buffer *Constants[2];

-      const struct brw_texture          *Texture[PIPE_MAX_SAMPLERS];

-   } attribs;

-

-   unsigned num_samplers;

-   unsigned num_textures;

-

-   struct brw_mem_pool pool[BRW_MAX_POOL];

-   struct brw_cache cache[BRW_MAX_CACHE];

-   struct brw_cached_batch_item *cached_batch_items;

-

-   struct {

-

-      /* Arrays with buffer objects to copy non-bufferobj arrays into

-       * for upload:

-       */

-      const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS];

-

-      struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS];

-

-#define BRW_NR_UPLOAD_BUFS 17

-#define BRW_UPLOAD_INIT_SIZE (128*1024)

-

-      /* Summary of size and varying of active arrays, so we can check

-       * for changes to this state:

-       */

-      struct brw_vertex_info info;

-   } vb;

-

-

-   unsigned hardware_dirty;

-   unsigned dirty;

-   unsigned pci_id;

-   /* BRW_NEW_URB_ALLOCATIONS:

-    */

-   struct {

-      unsigned vsize;		/* vertex size plus header in urb registers */

-      unsigned csize;		/* constant buffer size in urb registers */

-      unsigned sfsize;		/* setup data size in urb registers */

-

-      boolean constrained;

-

-      unsigned nr_vs_entries;

-      unsigned nr_gs_entries;

-      unsigned nr_clip_entries;

-      unsigned nr_sf_entries;

-      unsigned nr_cs_entries;

-

-/*       unsigned vs_size; */

-/*       unsigned gs_size; */

-/*       unsigned clip_size; */

-/*       unsigned sf_size; */

-/*       unsigned cs_size; */

-

-      unsigned vs_start;

-      unsigned gs_start;

-      unsigned clip_start;

-      unsigned sf_start;

-      unsigned cs_start;

-   } urb;

-

-

-   /* BRW_NEW_CURBE_OFFSETS:

-    */

-   struct {

-      unsigned wm_start;

-      unsigned wm_size;

-      unsigned clip_start;

-      unsigned clip_size;

-      unsigned vs_start;

-      unsigned vs_size;

-      unsigned total_size;

-

-      unsigned gs_offset;

-

-      float *last_buf;

-      unsigned last_bufsz;

-   } curbe;

-

-   struct {

-      struct brw_vs_prog_data *prog_data;

-

-      unsigned prog_gs_offset;

-      unsigned state_gs_offset;

-   } vs;

-

-   struct {

-      struct brw_gs_prog_data *prog_data;

-

-      boolean prog_active;

-      unsigned prog_gs_offset;

-      unsigned state_gs_offset;

-   } gs;

-

-   struct {

-      struct brw_clip_prog_data *prog_data;

-

-      unsigned prog_gs_offset;

-      unsigned vp_gs_offset;

-      unsigned state_gs_offset;

-   } clip;

-

-

-   struct {

-      struct brw_sf_prog_data *prog_data;

-

-      struct pipe_setup_linkage linkage;

-

-      unsigned prog_gs_offset;

-      unsigned vp_gs_offset;

-      unsigned state_gs_offset;

-   } sf;

-

-   struct {

-      struct brw_wm_prog_data *prog_data;

-

-//      struct brw_wm_compiler *compile_data;

-

-

-      /**

-       * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER

-       * cache

-       */

-      struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT];

-

-      unsigned render_surf;

-      unsigned nr_surfaces;

-

-      unsigned max_threads;

-      struct pipe_buffer *scratch_buffer;

-      unsigned scratch_buffer_size;

-

-      unsigned sampler_count;

-      unsigned sampler_gs_offset;

-

-      struct brw_surface_binding_table bind;

-      unsigned bind_ss_offset;

-

-      unsigned prog_gs_offset;

-      unsigned state_gs_offset;

-   } wm;

-

-

-   struct {

-      unsigned vp_gs_offset;

-      unsigned state_gs_offset;

-   } cc;

-

-

-   /* Used to give every program string a unique id

-    */

-   unsigned program_id;

-};

-

-

-#define BRW_PACKCOLOR8888(r,g,b,a)  ((r<<24) | (g<<16) | (b<<8) | a)

-

-

-/*======================================================================

- * brw_vtbl.c

- */

-void brw_do_flush( struct brw_context *brw,

-		   unsigned flags );

-

-

-/*======================================================================

- * brw_state.c

- */

-void brw_validate_state(struct brw_context *brw);

-void brw_init_state(struct brw_context *brw);

-void brw_destroy_state(struct brw_context *brw);

-

-

-/*======================================================================

- * brw_tex.c

- */

-void brwUpdateTextureState( struct brw_context *brw );

-

-

-/* brw_urb.c

- */

-void brw_upload_urb_fence(struct brw_context *brw);

-

-void brw_upload_constant_buffer_state(struct brw_context *brw);

-

-void brw_init_surface_functions(struct brw_context *brw);

-void brw_init_state_functions(struct brw_context *brw);

-void brw_init_flush_functions(struct brw_context *brw);

-void brw_init_string_functions(struct brw_context *brw);

-

-/*======================================================================

- * Inline conversion functions.  These are better-typed than the

- * macros used previously:

- */

-static inline struct brw_context *

-brw_context( struct pipe_context *ctx )

-{

-   return (struct brw_context *)ctx;

-}

-

-#endif

-

+/*
+ 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]>
+  */
+
+
+#ifndef BRWCONTEXT_INC
+#define BRWCONTEXT_INC
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "tgsi/util/tgsi_scan.h"
+
+#include "brw_structs.h"
+#include "brw_winsys.h"
+
+
+/* Glossary:
+ *
+ * URB - uniform resource buffer.  A mid-sized buffer which is
+ * partitioned between the fixed function units and used for passing
+ * values (vertices, primitives, constants) between them.
+ *
+ * CURBE - constant URB entry.  An urb region (entry) used to hold
+ * constant values which the fixed function units can be instructed to
+ * preload into the GRF when spawining a thread.
+ *
+ * VUE - vertex URB entry.  An urb entry holding a vertex and usually
+ * a vertex header.  The header contains control information and
+ * things like primitive type, Begin/end flags and clip codes.
+ *
+ * PUE - primitive URB entry.  An urb entry produced by the setup (SF)
+ * unit holding rasterization and interpolation parameters.
+ *
+ * GRF - general register file.  One of several register files
+ * addressable by programmed threads.  The inputs (r0, payload, curbe,
+ * urb) of the thread are preloaded to this area before the thread is
+ * spawned.  The registers are individually 8 dwords wide and suitable
+ * for general usage.  Registers holding thread input values are not
+ * special and may be overwritten.
+ *
+ * MRF - message register file.  Threads communicate (and terminate)
+ * by sending messages.  Message parameters are placed in contigous
+ * MRF registers.  All program output is via these messages.  URB
+ * entries are populated by sending a message to the shared URB
+ * function containing the new data, together with a control word,
+ * often an unmodified copy of R0.
+ *
+ * R0 - GRF register 0.  Typically holds control information used when
+ * sending messages to other threads.
+ *
+ * EU or GEN4 EU: The name of the programmable subsystem of the
+ * i965 hardware.  Threads are executed by the EU, the registers
+ * described above are part of the EU architecture.
+ *
+ * Fixed function units:
+ *
+ * CS - Command streamer.  Notional first unit, little software
+ * interaction.  Holds the URB entries used for constant data, ie the
+ * CURBEs.
+ *
+ * VF/VS - Vertex Fetch / Vertex Shader.  The fixed function part of
+ * this unit is responsible for pulling vertices out of vertex buffers
+ * in vram and injecting them into the processing pipe as VUEs.  If
+ * enabled, it first passes them to a VS thread which is a good place
+ * for the driver to implement any active vertex shader.
+ *
+ * GS - Geometry Shader.  This corresponds to a new DX10 concept.  If
+ * enabled, incoming strips etc are passed to GS threads in individual
+ * line/triangle/point units.  The GS thread may perform arbitary
+ * computation and emit whatever primtives with whatever vertices it
+ * chooses.  This makes GS an excellent place to implement GL's
+ * unfilled polygon modes, though of course it is capable of much
+ * more.  Additionally, GS is used to translate away primitives not
+ * handled by latter units, including Quads and Lineloops.
+ *
+ * CS - Clipper.  Mesa's clipping algorithms are imported to run on
+ * this unit.  The fixed function part performs cliptesting against
+ * the 6 fixed clipplanes and makes descisions on whether or not the
+ * incoming primitive needs to be passed to a thread for clipping.
+ * User clip planes are handled via cooperation with the VS thread.
+ *
+ * SF - Strips Fans or Setup: Triangles are prepared for
+ * rasterization.  Interpolation coefficients are calculated.
+ * Flatshading and two-side lighting usually performed here.
+ *
+ * WM - Windower.  Interpolation of vertex attributes performed here.
+ * Fragment shader implemented here.  SIMD aspects of EU taken full
+ * advantage of, as pixels are processed in blocks of 16.
+ *
+ * CC - Color Calculator.  No EU threads associated with this unit.
+ * Handles blending and (presumably) depth and stencil testing.
+ */
+
+#define BRW_MAX_CURBE                    (32*16)
+
+struct brw_context;
+struct brw_winsys;
+
+
+/* Raised when we receive new state across the pipe interface:
+ */
+#define BRW_NEW_VIEWPORT                0x1
+#define BRW_NEW_RASTERIZER              0x2
+#define BRW_NEW_FS                      0x4
+#define BRW_NEW_BLEND                   0x8
+#define BRW_NEW_CLIP                    0x10
+#define BRW_NEW_SCISSOR                 0x20
+#define BRW_NEW_STIPPLE                 0x40
+#define BRW_NEW_FRAMEBUFFER             0x80
+#define BRW_NEW_ALPHA_TEST              0x100
+#define BRW_NEW_DEPTH_STENCIL           0x200
+#define BRW_NEW_SAMPLER                 0x400
+#define BRW_NEW_TEXTURE                 0x800
+#define BRW_NEW_CONSTANTS               0x1000
+#define BRW_NEW_VBO                     0x2000
+#define BRW_NEW_VS                      0x4000
+
+/* Raised for other internal events:
+ */
+#define BRW_NEW_URB_FENCE               0x10000
+#define BRW_NEW_PSP                     0x20000
+#define BRW_NEW_CURBE_OFFSETS           0x40000
+#define BRW_NEW_REDUCED_PRIMITIVE       0x80000
+#define BRW_NEW_PRIMITIVE               0x100000
+#define BRW_NEW_SCENE                 0x200000
+#define BRW_NEW_SF_LINKAGE              0x400000
+
+extern int BRW_DEBUG;
+
+#define DEBUG_TEXTURE	0x1
+#define DEBUG_STATE	0x2
+#define DEBUG_IOCTL	0x4
+#define DEBUG_PRIMS	0x8
+#define DEBUG_VERTS	0x10
+#define DEBUG_FALLBACKS	0x20
+#define DEBUG_VERBOSE	0x40
+#define DEBUG_DRI       0x80
+#define DEBUG_DMA       0x100
+#define DEBUG_SANITY    0x200
+#define DEBUG_SYNC      0x400
+#define DEBUG_SLEEP     0x800
+#define DEBUG_PIXEL     0x1000
+#define DEBUG_STATS     0x2000
+#define DEBUG_TILE      0x4000
+#define DEBUG_SINGLE_THREAD   0x8000
+#define DEBUG_WM        0x10000
+#define DEBUG_URB       0x20000
+#define DEBUG_VS        0x40000
+#define DEBUG_BATCH	0x80000
+#define DEBUG_BUFMGR	0x100000
+#define DEBUG_BLIT	0x200000
+#define DEBUG_REGION	0x400000
+#define DEBUG_MIPTREE	0x800000
+
+#define DBG(...) do {						\
+   if (BRW_DEBUG & FILE_DEBUG_FLAG)				\
+      debug_printf(__VA_ARGS__);				\
+} while(0)
+
+#define PRINT(...) do {						\
+   debug_printf(brw->pipe.winsys, __VA_ARGS__);			\
+} while(0)
+
+struct brw_state_flags {
+   unsigned cache;
+   unsigned brw;
+};
+
+
+struct brw_vertex_program {
+   struct pipe_shader_state program;
+   struct tgsi_shader_info info;
+   int id;
+};
+
+
+struct brw_fragment_program {
+   struct pipe_shader_state program;
+   struct tgsi_shader_info info;
+   
+   boolean UsesDepth; /* XXX add this to tgsi_shader_info? */
+   int id;
+};
+
+
+struct pipe_setup_linkage {
+   struct {
+      unsigned vp_output:5;
+      unsigned interp_mode:4;
+      unsigned bf_vp_output:5;
+   } fp_input[PIPE_MAX_SHADER_INPUTS];
+
+   unsigned fp_input_count:5;
+   unsigned max_vp_output:5;
+};
+   
+
+
+struct brw_texture {
+   struct pipe_texture base;
+
+   /* Derived from the above:
+    */
+   unsigned pitch;
+   unsigned depth_pitch;          /* per-image on i945? */
+   unsigned total_height;
+
+   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* Explicitly store the offset of each image for each cube face or
+    * depth value.  Pretty much have to accept that hardware formats
+    * are going to be so diverse that there is no unified way to
+    * compute the offsets of depth/cube images within a mipmap level,
+    * so have to store them as a lookup table:
+    */
+   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */
+
+   /* Includes image offset tables:
+    */
+   unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* The data is held here:
+    */
+   struct pipe_buffer *buffer;
+};
+
+/* Data about a particular attempt to compile a program.  Note that
+ * there can be many of these, each in a different GL state
+ * corresponding to a different brw_wm_prog_key struct, with different
+ * compiled programs:
+ */
+/* Data about a particular attempt to compile a program.  Note that
+ * there can be many of these, each in a different GL state
+ * corresponding to a different brw_wm_prog_key struct, with different
+ * compiled programs:
+ */
+
+struct brw_wm_prog_data {
+   unsigned curb_read_length;
+   unsigned urb_read_length;
+
+   unsigned first_curbe_grf;
+   unsigned total_grf;
+   unsigned total_scratch;
+
+   /* Internally generated constants for the CURBE.  These are loaded
+    * ahead of the data from the constant buffer.
+    */
+   const float internal_const[8];
+   unsigned nr_internal_consts;
+   unsigned max_const;
+
+   boolean error;
+};
+
+struct brw_sf_prog_data {
+   unsigned urb_read_length;
+   unsigned total_grf;
+
+   /* Each vertex may have upto 12 attributes, 4 components each,
+    * except WPOS which requires only 2.  (11*4 + 2) == 44 ==> 11
+    * rows.
+    *
+    * Actually we use 4 for each, so call it 12 rows.
+    */
+   unsigned urb_entry_size;
+};
+
+struct brw_clip_prog_data {
+   unsigned curb_read_length;	/* user planes? */
+   unsigned clip_mode;
+   unsigned urb_read_length;
+   unsigned total_grf;
+};
+
+struct brw_gs_prog_data {
+   unsigned urb_read_length;
+   unsigned total_grf;
+};
+
+struct brw_vs_prog_data {
+   unsigned curb_read_length;
+   unsigned urb_read_length;
+   unsigned total_grf;
+   unsigned outputs_written;
+
+   unsigned inputs_read;
+
+   unsigned max_const;
+
+   float    imm_buf[PIPE_MAX_CONSTANT][4];
+   unsigned num_imm;
+   unsigned num_consts;
+
+   /* Used for calculating urb partitions:
+    */
+   unsigned urb_entry_size;
+};
+
+
+#define BRW_MAX_TEX_UNIT 8
+#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1
+
+/* Create a fixed sized struct for caching binding tables:
+ */
+struct brw_surface_binding_table {
+   unsigned surf_ss_offset[BRW_WM_MAX_SURF];
+};
+
+
+struct brw_cache;
+
+struct brw_mem_pool {
+   struct pipe_buffer *buffer;
+
+   unsigned size;
+   unsigned offset;		/* offset of first free byte */
+
+   struct brw_context *brw;
+};
+
+struct brw_cache_item {
+   unsigned hash;
+   unsigned key_size;		/* for variable-sized keys */
+   const void *key;
+
+   unsigned offset;		/* offset within pool's buffer */
+   unsigned data_size;
+
+   struct brw_cache_item *next;
+};
+
+
+
+struct brw_cache {
+   unsigned id;
+
+   const char *name;
+
+   struct brw_context *brw;
+   struct brw_mem_pool *pool;
+
+   struct brw_cache_item **items;
+   unsigned size, n_items;
+
+   unsigned key_size;		/* for fixed-size keys */
+   unsigned aux_size;
+
+   unsigned last_addr;			/* offset of active item */
+};
+
+
+
+
+/* Considered adding a member to this struct to document which flags
+ * an update might raise so that ordering of the state atoms can be
+ * checked or derived at runtime.  Dropped the idea in favor of having
+ * a debug mode where the state is monitored for flags which are
+ * raised that have already been tested against.
+ */
+struct brw_tracked_state {
+   struct brw_state_flags dirty;
+   void (*update)( struct brw_context *brw );
+};
+
+
+/* Flags for brw->state.cache.
+ */
+#define CACHE_NEW_CC_VP                  (1<<BRW_CC_VP)
+#define CACHE_NEW_CC_UNIT                (1<<BRW_CC_UNIT)
+#define CACHE_NEW_WM_PROG                (1<<BRW_WM_PROG)
+#define CACHE_NEW_SAMPLER_DEFAULT_COLOR  (1<<BRW_SAMPLER_DEFAULT_COLOR)
+#define CACHE_NEW_SAMPLER                (1<<BRW_SAMPLER)
+#define CACHE_NEW_WM_UNIT                (1<<BRW_WM_UNIT)
+#define CACHE_NEW_SF_PROG                (1<<BRW_SF_PROG)
+#define CACHE_NEW_SF_VP                  (1<<BRW_SF_VP)
+#define CACHE_NEW_SF_UNIT                (1<<BRW_SF_UNIT)
+#define CACHE_NEW_VS_UNIT                (1<<BRW_VS_UNIT)
+#define CACHE_NEW_VS_PROG                (1<<BRW_VS_PROG)
+#define CACHE_NEW_GS_UNIT                (1<<BRW_GS_UNIT)
+#define CACHE_NEW_GS_PROG                (1<<BRW_GS_PROG)
+#define CACHE_NEW_CLIP_VP                (1<<BRW_CLIP_VP)
+#define CACHE_NEW_CLIP_UNIT              (1<<BRW_CLIP_UNIT)
+#define CACHE_NEW_CLIP_PROG              (1<<BRW_CLIP_PROG)
+#define CACHE_NEW_SURFACE                (1<<BRW_SS_SURFACE)
+#define CACHE_NEW_SURF_BIND              (1<<BRW_SS_SURF_BIND)
+
+
+
+
+enum brw_mempool_id {
+   BRW_GS_POOL,
+   BRW_SS_POOL,
+   BRW_MAX_POOL
+};
+
+
+struct brw_cached_batch_item {
+   struct header *header;
+   unsigned sz;
+   struct brw_cached_batch_item *next;
+};
+
+
+
+/* Protect against a future where PIPE_MAX_ATTRIBS > 32.  Wouldn't life
+ * be easier if C allowed arrays of packed elements?
+ */
+#define ATTRIB_BIT_DWORDS  ((PIPE_MAX_ATTRIBS+31)/32)
+
+
+
+
+struct brw_vertex_info {
+   unsigned varying;  /* varying:1[PIPE_MAX_ATTRIBS] */
+   unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */
+};
+
+
+
+
+
+struct brw_context
+{
+   struct pipe_context pipe;
+   struct brw_winsys *winsys;
+
+   unsigned primitive;
+   unsigned reduced_primitive;
+
+   boolean emit_state_always;
+
+   struct {
+      struct brw_state_flags dirty;
+   } state;
+
+
+   struct {
+      const struct pipe_blend_state         *Blend;
+      const struct pipe_depth_stencil_alpha_state *DepthStencil;
+      const struct pipe_poly_stipple        *PolygonStipple;
+      const struct pipe_rasterizer_state    *Raster;
+      const struct pipe_sampler_state       *Samplers[PIPE_MAX_SAMPLERS];
+      const struct brw_vertex_program       *VertexProgram;
+      const struct brw_fragment_program     *FragmentProgram;
+
+      struct pipe_clip_state          Clip;
+      struct pipe_blend_color         BlendColor;
+      struct pipe_scissor_state       Scissor;
+      struct pipe_viewport_state      Viewport;
+      struct pipe_framebuffer_state   FrameBuffer;
+
+      const struct pipe_constant_buffer *Constants[2];
+      const struct brw_texture          *Texture[PIPE_MAX_SAMPLERS];
+   } attribs;
+
+   unsigned num_samplers;
+   unsigned num_textures;
+
+   struct brw_mem_pool pool[BRW_MAX_POOL];
+   struct brw_cache cache[BRW_MAX_CACHE];
+   struct brw_cached_batch_item *cached_batch_items;
+
+   struct {
+
+      /* Arrays with buffer objects to copy non-bufferobj arrays into
+       * for upload:
+       */
+      const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS];
+
+      struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS];
+
+#define BRW_NR_UPLOAD_BUFS 17
+#define BRW_UPLOAD_INIT_SIZE (128*1024)
+
+      /* Summary of size and varying of active arrays, so we can check
+       * for changes to this state:
+       */
+      struct brw_vertex_info info;
+   } vb;
+
+
+   unsigned hardware_dirty;
+   unsigned dirty;
+   unsigned pci_id;
+   /* BRW_NEW_URB_ALLOCATIONS:
+    */
+   struct {
+      unsigned vsize;		/* vertex size plus header in urb registers */
+      unsigned csize;		/* constant buffer size in urb registers */
+      unsigned sfsize;		/* setup data size in urb registers */
+
+      boolean constrained;
+
+      unsigned nr_vs_entries;
+      unsigned nr_gs_entries;
+      unsigned nr_clip_entries;
+      unsigned nr_sf_entries;
+      unsigned nr_cs_entries;
+
+/*       unsigned vs_size; */
+/*       unsigned gs_size; */
+/*       unsigned clip_size; */
+/*       unsigned sf_size; */
+/*       unsigned cs_size; */
+
+      unsigned vs_start;
+      unsigned gs_start;
+      unsigned clip_start;
+      unsigned sf_start;
+      unsigned cs_start;
+   } urb;
+
+
+   /* BRW_NEW_CURBE_OFFSETS:
+    */
+   struct {
+      unsigned wm_start;
+      unsigned wm_size;
+      unsigned clip_start;
+      unsigned clip_size;
+      unsigned vs_start;
+      unsigned vs_size;
+      unsigned total_size;
+
+      unsigned gs_offset;
+
+      float *last_buf;
+      unsigned last_bufsz;
+   } curbe;
+
+   struct {
+      struct brw_vs_prog_data *prog_data;
+
+      unsigned prog_gs_offset;
+      unsigned state_gs_offset;
+   } vs;
+
+   struct {
+      struct brw_gs_prog_data *prog_data;
+
+      boolean prog_active;
+      unsigned prog_gs_offset;
+      unsigned state_gs_offset;
+   } gs;
+
+   struct {
+      struct brw_clip_prog_data *prog_data;
+
+      unsigned prog_gs_offset;
+      unsigned vp_gs_offset;
+      unsigned state_gs_offset;
+   } clip;
+
+
+   struct {
+      struct brw_sf_prog_data *prog_data;
+
+      struct pipe_setup_linkage linkage;
+
+      unsigned prog_gs_offset;
+      unsigned vp_gs_offset;
+      unsigned state_gs_offset;
+   } sf;
+
+   struct {
+      struct brw_wm_prog_data *prog_data;
+
+//      struct brw_wm_compiler *compile_data;
+
+
+      /**
+       * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER
+       * cache
+       */
+      struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT];
+
+      unsigned render_surf;
+      unsigned nr_surfaces;
+
+      unsigned max_threads;
+      struct pipe_buffer *scratch_buffer;
+      unsigned scratch_buffer_size;
+
+      unsigned sampler_count;
+      unsigned sampler_gs_offset;
+
+      struct brw_surface_binding_table bind;
+      unsigned bind_ss_offset;
+
+      unsigned prog_gs_offset;
+      unsigned state_gs_offset;
+   } wm;
+
+
+   struct {
+      unsigned vp_gs_offset;
+      unsigned state_gs_offset;
+   } cc;
+
+
+   /* Used to give every program string a unique id
+    */
+   unsigned program_id;
+};
+
+
+#define BRW_PACKCOLOR8888(r,g,b,a)  ((r<<24) | (g<<16) | (b<<8) | a)
+
+
+/*======================================================================
+ * brw_vtbl.c
+ */
+void brw_do_flush( struct brw_context *brw,
+		   unsigned flags );
+
+
+/*======================================================================
+ * brw_state.c
+ */
+void brw_validate_state(struct brw_context *brw);
+void brw_init_state(struct brw_context *brw);
+void brw_destroy_state(struct brw_context *brw);
+
+
+/*======================================================================
+ * brw_tex.c
+ */
+void brwUpdateTextureState( struct brw_context *brw );
+
+
+/* brw_urb.c
+ */
+void brw_upload_urb_fence(struct brw_context *brw);
+
+void brw_upload_constant_buffer_state(struct brw_context *brw);
+
+void brw_init_surface_functions(struct brw_context *brw);
+void brw_init_state_functions(struct brw_context *brw);
+void brw_init_flush_functions(struct brw_context *brw);
+void brw_init_string_functions(struct brw_context *brw);
+
+/*======================================================================
+ * Inline conversion functions.  These are better-typed than the
+ * macros used previously:
+ */
+static inline struct brw_context *
+brw_context( struct pipe_context *ctx )
+{
+   return (struct brw_context *)ctx;
+}
+
+#endif
+
diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c
index ac243b7e4f9..caeeba46300 100644
--- a/src/gallium/drivers/i965simple/brw_state.c
+++ b/src/gallium/drivers/i965simple/brw_state.c
@@ -1,469 +1,469 @@
-/**************************************************************************

- *

- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.

- * All Rights Reserved.

- *

- * Permission is hereby granted, free of charge, to any person obtaining a

- * copy of this software and associated documentation files (the

- * "Software"), to deal in the Software without restriction, including

- * without limitation the rights to use, copy, modify, merge, publish,

- * distribute, sub license, and/or sell copies of the Software, and to

- * permit persons to whom the Software is furnished to do so, subject to

- * the following conditions:

- *

- * The above copyright notice and this permission notice (including the

- * next paragraph) shall be included in all copies or substantial portions

- * of the Software.

- *

- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS

- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.

- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR

- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,

- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE

- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

- *

- **************************************************************************/

-

-/* Authors:  Zack Rusin <[email protected]>

- *           Keith Whitwell <[email protected]>

- */

-

-

-#include "pipe/p_winsys.h"

-#include "pipe/p_util.h"

-#include "pipe/p_inlines.h"

-#include "pipe/p_shader_tokens.h"

-#include "tgsi/util/tgsi_dump.h"

-#include "tgsi/util/tgsi_parse.h"

-

-#include "brw_context.h"

-#include "brw_defines.h"

-#include "brw_state.h"

-#include "brw_draw.h"

-

-

-#define DUP( TYPE, VAL )                        \

-do {                                            \

-   struct TYPE *x = malloc(sizeof(*x));         \

-   memcpy(x, VAL, sizeof(*x) );                 \

-   return x;                                    \

-} while (0)

-

-/************************************************************************

- * Blend 

- */

-static void *

-brw_create_blend_state(struct pipe_context *pipe,

-                        const struct pipe_blend_state *blend)

-{   

-   DUP( pipe_blend_state, blend );

-}

-

-static void brw_bind_blend_state(struct pipe_context *pipe,

-                                 void *blend)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.Blend = (struct pipe_blend_state*)blend;

-   brw->state.dirty.brw |= BRW_NEW_BLEND;

-}

-

-

-static void brw_delete_blend_state(struct pipe_context *pipe, void *blend)

-{

-   free(blend);

-}

-

-static void brw_set_blend_color( struct pipe_context *pipe,

-			     const struct pipe_blend_color *blend_color )

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.BlendColor = *blend_color;

-

-   brw->state.dirty.brw |= BRW_NEW_BLEND;

-}

-

-/************************************************************************

- * Sampler 

- */

-

-static void *

-brw_create_sampler_state(struct pipe_context *pipe,

-                          const struct pipe_sampler_state *sampler)

-{

-   DUP( pipe_sampler_state, sampler );

-}

-

-static void brw_bind_sampler_states(struct pipe_context *pipe,

-                                    unsigned num, void **sampler)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   assert(num <= PIPE_MAX_SAMPLERS);

-

-   /* Check for no-op */

-   if (num == brw->num_samplers &&

-       !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *)))

-      return;

-

-   memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *));

-   memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) *

-          sizeof(void *));

-

-   brw->num_samplers = num;

-

-   brw->state.dirty.brw |= BRW_NEW_SAMPLER;

-}

-

-static void brw_delete_sampler_state(struct pipe_context *pipe,

-                                      void *sampler)

-{

-   free(sampler);

-}

-

-

-/************************************************************************

- * Depth stencil 

- */

-

-static void *

-brw_create_depth_stencil_state(struct pipe_context *pipe,

-                           const struct pipe_depth_stencil_alpha_state *depth_stencil)

-{

-   DUP( pipe_depth_stencil_alpha_state, depth_stencil );

-}

-

-static void brw_bind_depth_stencil_state(struct pipe_context *pipe,

-                                         void *depth_stencil)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil;

-

-   brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL;

-}

-

-static void brw_delete_depth_stencil_state(struct pipe_context *pipe,

-                                           void *depth_stencil)

-{

-   free(depth_stencil);

-}

-

-/************************************************************************

- * Scissor

- */

-static void brw_set_scissor_state( struct pipe_context *pipe,

-                                 const struct pipe_scissor_state *scissor )

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) );

-   brw->state.dirty.brw |= BRW_NEW_SCISSOR;

-}

-

-

-/************************************************************************

- * Stipple

- */

-

-static void brw_set_polygon_stipple( struct pipe_context *pipe,

-                                   const struct pipe_poly_stipple *stipple )

-{

-}

-

-

-/************************************************************************

- * Fragment shader

- */

-

-static void * brw_create_fs_state(struct pipe_context *pipe,

-                                   const struct pipe_shader_state *shader)

-{

-   struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program);

-

-   brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens);

-   brw_fp->id = brw_context(pipe)->program_id++;

-

-   tgsi_scan_shader(shader->tokens, &brw_fp->info);

-

-#if 0

-   brw_shader_info(shader->tokens,

-		   &brw_fp->info2);

-#endif

-

-   tgsi_dump(shader->tokens, 0);

-

-

-   return (void *)brw_fp;

-}

-

-static void brw_bind_fs_state(struct pipe_context *pipe, void *shader)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader;

-   brw->state.dirty.brw |= BRW_NEW_FS;

-}

-

-static void brw_delete_fs_state(struct pipe_context *pipe, void *shader)

-{

-   struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader;

-

-   FREE((void *) brw_fp->program.tokens);

-   FREE(brw_fp);

-}

-

-

-/************************************************************************

- * Vertex shader and other TNL state 

- */

-

-static void *brw_create_vs_state(struct pipe_context *pipe,

-                                 const struct pipe_shader_state *shader)

-{

-   struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program);

-

-   brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens);

-   brw_vp->id = brw_context(pipe)->program_id++;

-

-   tgsi_scan_shader(shader->tokens, &brw_vp->info);

-

-#if 0

-   brw_shader_info(shader->tokens,

-		   &brw_vp->info2);

-#endif

-   tgsi_dump(shader->tokens, 0);

-

-   return (void *)brw_vp;

-}

-

-static void brw_bind_vs_state(struct pipe_context *pipe, void *vs)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.VertexProgram = (struct brw_vertex_program *)vs;

-   brw->state.dirty.brw |= BRW_NEW_VS;

-

-   debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n");

-}

-

-static void brw_delete_vs_state(struct pipe_context *pipe, void *shader)

-{

-   struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader;

-

-   FREE((void *) brw_vp->program.tokens);

-   FREE(brw_vp);

-}

-

-

-static void brw_set_clip_state( struct pipe_context *pipe,

-                                const struct pipe_clip_state *clip )

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.Clip = *clip;

-}

-

-

-static void brw_set_viewport_state( struct pipe_context *pipe,

-				     const struct pipe_viewport_state *viewport )

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.Viewport = *viewport; /* struct copy */

-   brw->state.dirty.brw |= BRW_NEW_VIEWPORT;

-

-   /* pass the viewport info to the draw module */

-   //draw_set_viewport_state(brw->draw, viewport);

-}

-

-

-static void brw_set_vertex_buffers(struct pipe_context *pipe,

-				   unsigned count,

-				   const struct pipe_vertex_buffer *buffers)

-{

-   struct brw_context *brw = brw_context(pipe);

-   memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0]));

-}

-

-static void brw_set_vertex_elements(struct pipe_context *pipe,

-                                    unsigned count,

-                                    const struct pipe_vertex_element *elements)

-{

-   /* flush ? */

-   struct brw_context *brw = brw_context(pipe);

-   uint i;

-

-   assert(count <= PIPE_MAX_ATTRIBS);

-

-   for (i = 0; i < count; i++) {

-      struct brw_vertex_element_state el;

-      memset(&el, 0, sizeof(el));

-

-      el.ve0.src_offset = elements[i].src_offset;

-      el.ve0.src_format = brw_translate_surface_format(elements[i].src_format);

-      el.ve0.valid = 1;

-      el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index;

-

-      el.ve1.dst_offset   = i * 4;

-

-      el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC;

-      el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC;

-      el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC;

-      el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC;

-

-      switch (elements[i].nr_components) {

-      case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0;

-      case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0;

-      case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT;

-         break;

-      }

-

-      brw->vb.inputs[i] = el;

-   }

-}

-

-

-

-/************************************************************************

- * Constant buffers

- */

-

-static void brw_set_constant_buffer(struct pipe_context *pipe,

-                                     uint shader, uint index,

-                                     const struct pipe_constant_buffer *buf)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   assert(buf == 0 || index == 0);

-

-   brw->attribs.Constants[shader] = buf;

-   brw->state.dirty.brw |= BRW_NEW_CONSTANTS;

-}

-

-

-/************************************************************************

- * Texture surfaces

- */

-

-

-static void brw_set_sampler_textures(struct pipe_context *pipe,

-                                     unsigned num,

-                                     struct pipe_texture **texture)

-{

-   struct brw_context *brw = brw_context(pipe);

-   uint i;

-

-   assert(num <= PIPE_MAX_SAMPLERS);

-

-   /* Check for no-op */

-   if (num == brw->num_textures &&

-       !memcmp(brw->attribs.Texture, texture, num *

-               sizeof(struct pipe_texture *)))

-      return;

-

-   for (i = 0; i < num; i++)

-      pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i],

-                             texture[i]);

-

-   for (i = num; i < brw->num_textures; i++)

-      pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i],

-                             NULL);

-

-   brw->num_textures = num;

-

-   brw->state.dirty.brw |= BRW_NEW_TEXTURE;

-}

-

-

-/************************************************************************

- * Render targets, etc

- */

-

-static void brw_set_framebuffer_state(struct pipe_context *pipe,

-				       const struct pipe_framebuffer_state *fb)

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.FrameBuffer = *fb; /* struct copy */

-

-   brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER;

-}

-

-

-

-/************************************************************************

- * Rasterizer state

- */

-

-static void *

-brw_create_rasterizer_state(struct pipe_context *pipe,

-                             const struct pipe_rasterizer_state *rasterizer)

-{

-   DUP(pipe_rasterizer_state, rasterizer);

-}

-

-static void brw_bind_rasterizer_state( struct pipe_context *pipe,

-                                        void *setup )

-{

-   struct brw_context *brw = brw_context(pipe);

-

-   brw->attribs.Raster = (struct pipe_rasterizer_state *)setup;

-

-   /* Also pass-through to draw module:

-    */

-   //draw_set_rasterizer_state(brw->draw, setup);

-

-   brw->state.dirty.brw |= BRW_NEW_RASTERIZER;

-}

-

-static void brw_delete_rasterizer_state(struct pipe_context *pipe,

-                                         void *setup)

-{

-   free(setup);

-}

-

-

-

-void

-brw_init_state_functions( struct brw_context *brw )

-{

-   brw->pipe.create_blend_state = brw_create_blend_state;

-   brw->pipe.bind_blend_state = brw_bind_blend_state;

-   brw->pipe.delete_blend_state = brw_delete_blend_state;

-

-   brw->pipe.create_sampler_state = brw_create_sampler_state;

-   brw->pipe.bind_sampler_states = brw_bind_sampler_states;

-   brw->pipe.delete_sampler_state = brw_delete_sampler_state;

-

-   brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state;

-   brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state;

-   brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state;

-

-   brw->pipe.create_rasterizer_state = brw_create_rasterizer_state;

-   brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state;

-   brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state;

-   brw->pipe.create_fs_state = brw_create_fs_state;

-   brw->pipe.bind_fs_state = brw_bind_fs_state;

-   brw->pipe.delete_fs_state = brw_delete_fs_state;

-   brw->pipe.create_vs_state = brw_create_vs_state;

-   brw->pipe.bind_vs_state = brw_bind_vs_state;

-   brw->pipe.delete_vs_state = brw_delete_vs_state;

-

-   brw->pipe.set_blend_color = brw_set_blend_color;

-   brw->pipe.set_clip_state = brw_set_clip_state;

-   brw->pipe.set_constant_buffer = brw_set_constant_buffer;

-   brw->pipe.set_framebuffer_state = brw_set_framebuffer_state;

-

-//   brw->pipe.set_feedback_state = brw_set_feedback_state;

-//   brw->pipe.set_feedback_buffer = brw_set_feedback_buffer;

-

-   brw->pipe.set_polygon_stipple = brw_set_polygon_stipple;

-   brw->pipe.set_scissor_state = brw_set_scissor_state;

-   brw->pipe.set_sampler_textures = brw_set_sampler_textures;

-   brw->pipe.set_viewport_state = brw_set_viewport_state;

-   brw->pipe.set_vertex_buffers = brw_set_vertex_buffers;

-   brw->pipe.set_vertex_elements = brw_set_vertex_elements;

-}

+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/* Authors:  Zack Rusin <[email protected]>
+ *           Keith Whitwell <[email protected]>
+ */
+
+
+#include "pipe/p_winsys.h"
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/util/tgsi_dump.h"
+#include "tgsi/util/tgsi_parse.h"
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_state.h"
+#include "brw_draw.h"
+
+
+#define DUP( TYPE, VAL )                        \
+do {                                            \
+   struct TYPE *x = malloc(sizeof(*x));         \
+   memcpy(x, VAL, sizeof(*x) );                 \
+   return x;                                    \
+} while (0)
+
+/************************************************************************
+ * Blend 
+ */
+static void *
+brw_create_blend_state(struct pipe_context *pipe,
+                        const struct pipe_blend_state *blend)
+{   
+   DUP( pipe_blend_state, blend );
+}
+
+static void brw_bind_blend_state(struct pipe_context *pipe,
+                                 void *blend)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Blend = (struct pipe_blend_state*)blend;
+   brw->state.dirty.brw |= BRW_NEW_BLEND;
+}
+
+
+static void brw_delete_blend_state(struct pipe_context *pipe, void *blend)
+{
+   free(blend);
+}
+
+static void brw_set_blend_color( struct pipe_context *pipe,
+			     const struct pipe_blend_color *blend_color )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.BlendColor = *blend_color;
+
+   brw->state.dirty.brw |= BRW_NEW_BLEND;
+}
+
+/************************************************************************
+ * Sampler 
+ */
+
+static void *
+brw_create_sampler_state(struct pipe_context *pipe,
+                          const struct pipe_sampler_state *sampler)
+{
+   DUP( pipe_sampler_state, sampler );
+}
+
+static void brw_bind_sampler_states(struct pipe_context *pipe,
+                                    unsigned num, void **sampler)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == brw->num_samplers &&
+       !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *)))
+      return;
+
+   memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *));
+   memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) *
+          sizeof(void *));
+
+   brw->num_samplers = num;
+
+   brw->state.dirty.brw |= BRW_NEW_SAMPLER;
+}
+
+static void brw_delete_sampler_state(struct pipe_context *pipe,
+                                      void *sampler)
+{
+   free(sampler);
+}
+
+
+/************************************************************************
+ * Depth stencil 
+ */
+
+static void *
+brw_create_depth_stencil_state(struct pipe_context *pipe,
+                           const struct pipe_depth_stencil_alpha_state *depth_stencil)
+{
+   DUP( pipe_depth_stencil_alpha_state, depth_stencil );
+}
+
+static void brw_bind_depth_stencil_state(struct pipe_context *pipe,
+                                         void *depth_stencil)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil;
+
+   brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL;
+}
+
+static void brw_delete_depth_stencil_state(struct pipe_context *pipe,
+                                           void *depth_stencil)
+{
+   free(depth_stencil);
+}
+
+/************************************************************************
+ * Scissor
+ */
+static void brw_set_scissor_state( struct pipe_context *pipe,
+                                 const struct pipe_scissor_state *scissor )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) );
+   brw->state.dirty.brw |= BRW_NEW_SCISSOR;
+}
+
+
+/************************************************************************
+ * Stipple
+ */
+
+static void brw_set_polygon_stipple( struct pipe_context *pipe,
+                                   const struct pipe_poly_stipple *stipple )
+{
+}
+
+
+/************************************************************************
+ * Fragment shader
+ */
+
+static void * brw_create_fs_state(struct pipe_context *pipe,
+                                   const struct pipe_shader_state *shader)
+{
+   struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program);
+
+   brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens);
+   brw_fp->id = brw_context(pipe)->program_id++;
+
+   tgsi_scan_shader(shader->tokens, &brw_fp->info);
+
+#if 0
+   brw_shader_info(shader->tokens,
+		   &brw_fp->info2);
+#endif
+
+   tgsi_dump(shader->tokens, 0);
+
+
+   return (void *)brw_fp;
+}
+
+static void brw_bind_fs_state(struct pipe_context *pipe, void *shader)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader;
+   brw->state.dirty.brw |= BRW_NEW_FS;
+}
+
+static void brw_delete_fs_state(struct pipe_context *pipe, void *shader)
+{
+   struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader;
+
+   FREE((void *) brw_fp->program.tokens);
+   FREE(brw_fp);
+}
+
+
+/************************************************************************
+ * Vertex shader and other TNL state 
+ */
+
+static void *brw_create_vs_state(struct pipe_context *pipe,
+                                 const struct pipe_shader_state *shader)
+{
+   struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program);
+
+   brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens);
+   brw_vp->id = brw_context(pipe)->program_id++;
+
+   tgsi_scan_shader(shader->tokens, &brw_vp->info);
+
+#if 0
+   brw_shader_info(shader->tokens,
+		   &brw_vp->info2);
+#endif
+   tgsi_dump(shader->tokens, 0);
+
+   return (void *)brw_vp;
+}
+
+static void brw_bind_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.VertexProgram = (struct brw_vertex_program *)vs;
+   brw->state.dirty.brw |= BRW_NEW_VS;
+
+   debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n");
+}
+
+static void brw_delete_vs_state(struct pipe_context *pipe, void *shader)
+{
+   struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader;
+
+   FREE((void *) brw_vp->program.tokens);
+   FREE(brw_vp);
+}
+
+
+static void brw_set_clip_state( struct pipe_context *pipe,
+                                const struct pipe_clip_state *clip )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Clip = *clip;
+}
+
+
+static void brw_set_viewport_state( struct pipe_context *pipe,
+				     const struct pipe_viewport_state *viewport )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Viewport = *viewport; /* struct copy */
+   brw->state.dirty.brw |= BRW_NEW_VIEWPORT;
+
+   /* pass the viewport info to the draw module */
+   //draw_set_viewport_state(brw->draw, viewport);
+}
+
+
+static void brw_set_vertex_buffers(struct pipe_context *pipe,
+				   unsigned count,
+				   const struct pipe_vertex_buffer *buffers)
+{
+   struct brw_context *brw = brw_context(pipe);
+   memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0]));
+}
+
+static void brw_set_vertex_elements(struct pipe_context *pipe,
+                                    unsigned count,
+                                    const struct pipe_vertex_element *elements)
+{
+   /* flush ? */
+   struct brw_context *brw = brw_context(pipe);
+   uint i;
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   for (i = 0; i < count; i++) {
+      struct brw_vertex_element_state el;
+      memset(&el, 0, sizeof(el));
+
+      el.ve0.src_offset = elements[i].src_offset;
+      el.ve0.src_format = brw_translate_surface_format(elements[i].src_format);
+      el.ve0.valid = 1;
+      el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index;
+
+      el.ve1.dst_offset   = i * 4;
+
+      el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC;
+      el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC;
+      el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC;
+      el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC;
+
+      switch (elements[i].nr_components) {
+      case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0;
+      case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0;
+      case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT;
+         break;
+      }
+
+      brw->vb.inputs[i] = el;
+   }
+}
+
+
+
+/************************************************************************
+ * Constant buffers
+ */
+
+static void brw_set_constant_buffer(struct pipe_context *pipe,
+                                     uint shader, uint index,
+                                     const struct pipe_constant_buffer *buf)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   assert(buf == 0 || index == 0);
+
+   brw->attribs.Constants[shader] = buf;
+   brw->state.dirty.brw |= BRW_NEW_CONSTANTS;
+}
+
+
+/************************************************************************
+ * Texture surfaces
+ */
+
+
+static void brw_set_sampler_textures(struct pipe_context *pipe,
+                                     unsigned num,
+                                     struct pipe_texture **texture)
+{
+   struct brw_context *brw = brw_context(pipe);
+   uint i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == brw->num_textures &&
+       !memcmp(brw->attribs.Texture, texture, num *
+               sizeof(struct pipe_texture *)))
+      return;
+
+   for (i = 0; i < num; i++)
+      pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i],
+                             texture[i]);
+
+   for (i = num; i < brw->num_textures; i++)
+      pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i],
+                             NULL);
+
+   brw->num_textures = num;
+
+   brw->state.dirty.brw |= BRW_NEW_TEXTURE;
+}
+
+
+/************************************************************************
+ * Render targets, etc
+ */
+
+static void brw_set_framebuffer_state(struct pipe_context *pipe,
+				       const struct pipe_framebuffer_state *fb)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.FrameBuffer = *fb; /* struct copy */
+
+   brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER;
+}
+
+
+
+/************************************************************************
+ * Rasterizer state
+ */
+
+static void *
+brw_create_rasterizer_state(struct pipe_context *pipe,
+                             const struct pipe_rasterizer_state *rasterizer)
+{
+   DUP(pipe_rasterizer_state, rasterizer);
+}
+
+static void brw_bind_rasterizer_state( struct pipe_context *pipe,
+                                        void *setup )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Raster = (struct pipe_rasterizer_state *)setup;
+
+   /* Also pass-through to draw module:
+    */
+   //draw_set_rasterizer_state(brw->draw, setup);
+
+   brw->state.dirty.brw |= BRW_NEW_RASTERIZER;
+}
+
+static void brw_delete_rasterizer_state(struct pipe_context *pipe,
+                                         void *setup)
+{
+   free(setup);
+}
+
+
+
+void
+brw_init_state_functions( struct brw_context *brw )
+{
+   brw->pipe.create_blend_state = brw_create_blend_state;
+   brw->pipe.bind_blend_state = brw_bind_blend_state;
+   brw->pipe.delete_blend_state = brw_delete_blend_state;
+
+   brw->pipe.create_sampler_state = brw_create_sampler_state;
+   brw->pipe.bind_sampler_states = brw_bind_sampler_states;
+   brw->pipe.delete_sampler_state = brw_delete_sampler_state;
+
+   brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state;
+   brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state;
+   brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state;
+
+   brw->pipe.create_rasterizer_state = brw_create_rasterizer_state;
+   brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state;
+   brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state;
+   brw->pipe.create_fs_state = brw_create_fs_state;
+   brw->pipe.bind_fs_state = brw_bind_fs_state;
+   brw->pipe.delete_fs_state = brw_delete_fs_state;
+   brw->pipe.create_vs_state = brw_create_vs_state;
+   brw->pipe.bind_vs_state = brw_bind_vs_state;
+   brw->pipe.delete_vs_state = brw_delete_vs_state;
+
+   brw->pipe.set_blend_color = brw_set_blend_color;
+   brw->pipe.set_clip_state = brw_set_clip_state;
+   brw->pipe.set_constant_buffer = brw_set_constant_buffer;
+   brw->pipe.set_framebuffer_state = brw_set_framebuffer_state;
+
+//   brw->pipe.set_feedback_state = brw_set_feedback_state;
+//   brw->pipe.set_feedback_buffer = brw_set_feedback_buffer;
+
+   brw->pipe.set_polygon_stipple = brw_set_polygon_stipple;
+   brw->pipe.set_scissor_state = brw_set_scissor_state;
+   brw->pipe.set_sampler_textures = brw_set_sampler_textures;
+   brw->pipe.set_viewport_state = brw_set_viewport_state;
+   brw->pipe.set_vertex_buffers = brw_set_vertex_buffers;
+   brw->pipe.set_vertex_elements = brw_set_vertex_elements;
+}
diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c
index c99a91dcf76..3e3736b2806 100644
--- a/src/gallium/drivers/i965simple/brw_surface.c
+++ b/src/gallium/drivers/i965simple/brw_surface.c
@@ -35,27 +35,6 @@
 #include "util/p_tile.h"
 
 
-/* Upload data to a rectangular sub-region.  Lots of choices how to do this:
- *
- * - memcpy by span to current destination
- * - upload data as new buffer and blit
- *
- * Currently always memcpy.
- */
-static void
-brw_surface_data(struct pipe_context *pipe,
-                 struct pipe_surface *dst,
-                 unsigned dstx, unsigned dsty,
-                 const void *src, unsigned src_pitch,
-                 unsigned srcx, unsigned srcy, unsigned width, unsigned height)
-{
-   pipe_copy_rect(pipe_surface_map(dst) + dst->offset,
-                  dst->cpp, dst->pitch,
-                  dstx, dsty, width, height, src, src_pitch, srcx, srcy);
-
-   pipe_surface_unmap(dst);
-}
-
 
 /* Assumes all values are within bounds -- no checking at this level -
  * do it higher up if required.
@@ -72,17 +51,25 @@ brw_surface_copy(struct pipe_context *pipe,
    assert(dst->cpp == src->cpp);
 
    if (0) {
-      pipe_copy_rect(pipe_surface_map(dst) + dst->offset,
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+      
+      const void *src_map = pipe->screen->surface_map( pipe->screen,
+                                                       src,
+                                                       PIPE_BUFFER_USAGE_CPU_READ );
+      
+      pipe_copy_rect(dst_map,
                      dst->cpp,
                      dst->pitch,
-                     dstx, dsty,
-                     width, height,
-                     pipe_surface_map(src) + src->offset,
-                     do_flip ? -src->pitch : src->pitch,
+                     dstx, dsty, 
+                     width, height, 
+                     src_map, 
+                     do_flip ? -(int) src->pitch : src->pitch, 
                      srcx, do_flip ? 1 - srcy - height : srcy);
 
-      pipe_surface_unmap(src);
-      pipe_surface_unmap(dst);
+      pipe->screen->surface_unmap(pipe->screen, src);
+      pipe->screen->surface_unmap(pipe->screen, dst);
    }
    else {
       brw_copy_blit(brw_context(pipe),
@@ -113,7 +100,10 @@ brw_surface_fill(struct pipe_context *pipe,
 {
    if (0) {
       unsigned i, j;
-      void *dst_map = pipe_surface_map(dst);
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+
 
       switch (dst->cpp) {
       case 1: {
@@ -147,7 +137,7 @@ brw_surface_fill(struct pipe_context *pipe,
 	 break;
       }
 
-      pipe_surface_unmap( dst );
+      pipe->screen->surface_unmap(pipe->screen, dst);
    }
    else {
       brw_fill_blit(brw_context(pipe),
@@ -164,7 +154,6 @@ brw_surface_fill(struct pipe_context *pipe,
 void
 brw_init_surface_functions(struct brw_context *brw)
 {
-   (void) brw_surface_data; /* silence warning */
    brw->pipe.surface_copy  = brw_surface_copy;
    brw->pipe.surface_fill  = brw_surface_fill;
 }
diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c
index b580f98204c..78ae0b1223e 100644
--- a/src/gallium/drivers/i965simple/brw_tex_layout.c
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.c
@@ -357,14 +357,6 @@ brw_texture_release_screen(struct pipe_screen *screen,
 }
 
 
-static void
-brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture,
-                   uint face, uint levelsMask)
-{
-   /* no-op? */
-}
-
-
 static struct pipe_surface *
 brw_get_tex_surface_screen(struct pipe_screen *screen,
                            struct pipe_texture *pt,
@@ -407,7 +399,7 @@ brw_get_tex_surface_screen(struct pipe_screen *screen,
 void
 brw_init_texture_functions(struct brw_context *brw)
 {
-   brw->pipe.texture_update = brw_texture_update;
+//   brw->pipe.texture_update = brw_texture_update;
 }
 
 
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index fe9cd8375e3..2af0db37143 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -192,11 +192,11 @@ softpipe_create( struct pipe_screen *screen,
     * Must be before quad stage setup!
     */
    for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++)
-      softpipe->cbuf_cache[i] = sp_create_tile_cache();
-   softpipe->zsbuf_cache = sp_create_tile_cache();
+      softpipe->cbuf_cache[i] = sp_create_tile_cache( screen );
+   softpipe->zsbuf_cache = sp_create_tile_cache( screen );
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
-      softpipe->tex_cache[i] = sp_create_tile_cache();
+      softpipe->tex_cache[i] = sp_create_tile_cache( screen );
 
 
    /* setup quad rendering stages */
diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c
index 0625b69099b..e03994b63b7 100644
--- a/src/gallium/drivers/softpipe/sp_flush.c
+++ b/src/gallium/drivers/softpipe/sp_flush.c
@@ -50,25 +50,28 @@ softpipe_flush( struct pipe_context *pipe,
 
    draw_flush(softpipe->draw);
 
-   /* - flush the quad pipeline
-    * - flush the texture cache
-    * - flush the render cache
-    */
+   if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
+      for (i = 0; i < softpipe->num_textures; i++) {
+         sp_flush_tile_cache(softpipe, softpipe->tex_cache[i]);
+      }
+   }
 
-   for (i = 0; i < softpipe->framebuffer.num_cbufs; i++)
-      if (softpipe->cbuf_cache[i])
-         sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]);
+   if (flags & PIPE_FLUSH_RENDER_CACHE) {
+      for (i = 0; i < softpipe->framebuffer.num_cbufs; i++)
+         if (softpipe->cbuf_cache[i])
+            sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]);
 
-   if (softpipe->zsbuf_cache)
-      sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache);
+      if (softpipe->zsbuf_cache)
+         sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache);
 
-   /* Need this call for hardware buffers before swapbuffers.
-    *
-    * there should probably be another/different flush-type function
-    * that's called before swapbuffers because we don't always want
-    * to unmap surfaces when flushing.
-    */
-   softpipe_unmap_surfaces(softpipe);
+      /* Need this call for hardware buffers before swapbuffers.
+       *
+       * there should probably be another/different flush-type function
+       * that's called before swapbuffers because we don't always want
+       * to unmap surfaces when flushing.
+       */
+      softpipe_unmap_surfaces(softpipe);
+   }
 
    if (fence)
       *fence = NULL;
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index 7dacb1c4613..e9926bf41f9 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -33,6 +33,7 @@
 
 #include "sp_texture.h"
 #include "sp_winsys.h"
+#include "sp_screen.h"
 
 
 static const char *
@@ -137,6 +138,7 @@ softpipe_destroy_screen( struct pipe_screen *screen )
 }
 
 
+
 /**
  * Create a new pipe_screen object
  * Note: we're not presently subclassing pipe_screen (no softpipe_screen).
@@ -144,22 +146,22 @@ softpipe_destroy_screen( struct pipe_screen *screen )
 struct pipe_screen *
 softpipe_create_screen(struct pipe_winsys *winsys)
 {
-   struct pipe_screen *screen = CALLOC_STRUCT(pipe_screen);
+   struct softpipe_screen *screen = CALLOC_STRUCT(softpipe_screen);
 
    if (!screen)
       return NULL;
 
-   screen->winsys = winsys;
+   screen->base.winsys = winsys;
 
-   screen->destroy = softpipe_destroy_screen;
+   screen->base.destroy = softpipe_destroy_screen;
 
-   screen->get_name = softpipe_get_name;
-   screen->get_vendor = softpipe_get_vendor;
-   screen->get_param = softpipe_get_param;
-   screen->get_paramf = softpipe_get_paramf;
-   screen->is_format_supported = softpipe_is_format_supported;
+   screen->base.get_name = softpipe_get_name;
+   screen->base.get_vendor = softpipe_get_vendor;
+   screen->base.get_param = softpipe_get_param;
+   screen->base.get_paramf = softpipe_get_paramf;
+   screen->base.is_format_supported = softpipe_is_format_supported;
 
-   softpipe_init_screen_texture_funcs(screen);
+   softpipe_init_screen_texture_funcs(&screen->base);
 
-   return screen;
+   return &screen->base;
 }
diff --git a/src/gallium/drivers/softpipe/sp_screen.h b/src/gallium/drivers/softpipe/sp_screen.h
new file mode 100644
index 00000000000..3d4bfd3e840
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_screen.h
@@ -0,0 +1,58 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <[email protected]>
+ */
+
+#ifndef SP_SCREEN_H
+#define SP_SCREEN_H
+
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+
+
+
+struct softpipe_screen {
+   struct pipe_screen base;
+
+   /* Increments whenever textures are modified.  Contexts can track
+    * this.
+    */
+   unsigned timestamp;          
+};
+
+
+
+
+static INLINE struct softpipe_screen *
+softpipe_screen( struct pipe_screen *pipe )
+{
+   return (struct softpipe_screen *)pipe;
+}
+
+
+#endif /* SP_SCREEN_H */
diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c
index 653449c4f18..29a1e92416e 100644
--- a/src/gallium/drivers/softpipe/sp_surface.c
+++ b/src/gallium/drivers/softpipe/sp_surface.c
@@ -46,19 +46,28 @@ sp_surface_copy(struct pipe_context *pipe,
 		struct pipe_surface *src,
 		unsigned srcx, unsigned srcy, unsigned width, unsigned height)
 {
+   void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                              dst,
+                                              PIPE_BUFFER_USAGE_CPU_WRITE );
+
+   const void *src_map = pipe->screen->surface_map( pipe->screen,
+                                                    src,
+                                                    PIPE_BUFFER_USAGE_CPU_READ );
+
    assert( dst->cpp == src->cpp );
+   assert(src_map && dst_map);
 
-   pipe_copy_rect(pipe_surface_map(dst),
+   pipe_copy_rect(dst_map,
                   dst->cpp,
                   dst->pitch,
                   dstx, dsty,
                   width, height,
-                  pipe_surface_map(src),
+                  src_map,
                   do_flip ? -(int) src->pitch : src->pitch,
                   srcx, do_flip ? 1 - srcy - height : srcy);
 
-   pipe_surface_unmap(src);
-   pipe_surface_unmap(dst);
+   pipe->screen->surface_unmap(pipe->screen, src);
+   pipe->screen->surface_unmap(pipe->screen, dst);
 }
 
 
@@ -83,7 +92,9 @@ sp_surface_fill(struct pipe_context *pipe,
 		unsigned width, unsigned height, unsigned value)
 {
    unsigned i, j;
-   void *dst_map = pipe_surface_map(dst);
+   void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                              dst,
+                                              PIPE_BUFFER_USAGE_CPU_WRITE );
 
    assert(dst->pitch > 0);
    assert(width <= dst->pitch);
@@ -147,7 +158,7 @@ sp_surface_fill(struct pipe_context *pipe,
       break;
    }
 
-   pipe_surface_unmap( dst );
+   pipe->screen->surface_unmap(pipe->screen, dst);
 }
 
 
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 256586ec886..1d7a1fffe4d 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -40,6 +40,7 @@
 #include "sp_state.h"
 #include "sp_texture.h"
 #include "sp_tile_cache.h"
+#include "sp_screen.h"
 
 
 /* Simple, maximally packed layout.
@@ -51,40 +52,88 @@ static unsigned minify( unsigned d )
 }
 
 
-static void
-softpipe_texture_layout(struct softpipe_texture * spt)
+/* Conventional allocation path for non-display textures:
+ */
+static boolean
+softpipe_texture_layout(struct pipe_screen *screen,
+                        struct softpipe_texture * spt)
 {
+   struct pipe_winsys *ws = screen->winsys;
    struct pipe_texture *pt = &spt->base;
    unsigned level;
    unsigned width = pt->width[0];
    unsigned height = pt->height[0];
    unsigned depth = pt->depth[0];
 
-   spt->buffer_size = 0;
+   unsigned buffer_size = 0;
 
    for (level = 0; level <= pt->last_level; level++) {
       pt->width[level] = width;
       pt->height[level] = height;
       pt->depth[level] = depth;
+      spt->pitch[level] = width;
 
-      spt->level_offset[level] = spt->buffer_size;
+      spt->level_offset[level] = buffer_size;
 
-      spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) *
-			  ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) *
-			  width * pt->cpp;
+      buffer_size += (((pt->compressed) ? MAX2(1, height/4) : height) *
+                      ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) *
+                      width * pt->cpp);
 
       width  = minify(width);
       height = minify(height);
       depth = minify(depth);
    }
+
+   spt->buffer = ws->buffer_create(ws, 32,
+                                   PIPE_BUFFER_USAGE_PIXEL,
+                                   buffer_size);
+
+   return spt->buffer != NULL;
+}
+
+
+
+/* Hack it up to use the old winsys->surface_alloc_storage()
+ * method for now:
+ */
+static boolean
+softpipe_displaytarget_layout(struct pipe_screen *screen,
+                              struct softpipe_texture * spt)
+{
+   struct pipe_winsys *ws = screen->winsys;
+   struct pipe_surface surf;
+   unsigned flags = (PIPE_BUFFER_USAGE_CPU_READ |
+                     PIPE_BUFFER_USAGE_CPU_WRITE |
+                     PIPE_BUFFER_USAGE_GPU_READ |
+                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+
+   memset(&surf, 0, sizeof(surf));
+
+   ws->surface_alloc_storage( ws, 
+                              &surf,
+                              spt->base.width[0], 
+                              spt->base.height[0],
+                              spt->base.format,
+                              flags,
+                              spt->base.tex_usage);
+      
+   /* Now extract the goodies: 
+    */
+   spt->buffer = surf.buffer;
+   spt->pitch[0] = surf.pitch;
+
+   return spt->buffer != NULL;
 }
 
 
+
+
+
 static struct pipe_texture *
 softpipe_texture_create(struct pipe_screen *screen,
                         const struct pipe_texture *templat)
 {
-   struct pipe_winsys *ws = screen->winsys;
    struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture);
    if (!spt)
       return NULL;
@@ -93,19 +142,21 @@ softpipe_texture_create(struct pipe_screen *screen,
    spt->base.refcount = 1;
    spt->base.screen = screen;
 
-   softpipe_texture_layout(spt);
-
-   spt->buffer = ws->buffer_create(ws, 32,
-                                   PIPE_BUFFER_USAGE_PIXEL,
-                                   spt->buffer_size);
-   if (!spt->buffer) {
-      FREE(spt);
-      return NULL;
+   if (spt->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
+      if (!softpipe_displaytarget_layout(screen, spt))
+         goto fail;
    }
-
+   else {
+      if (!softpipe_texture_layout(screen, spt))
+         goto fail;
+   }
+    
    assert(spt->base.refcount == 1);
-
    return &spt->base;
+
+ fail:
+   FREE(spt);
+   return NULL;
 }
 
 
@@ -116,19 +167,10 @@ softpipe_texture_release(struct pipe_screen *screen,
    if (!*pt)
       return;
 
-   /*
-   DBG("%s %p refcount will be %d\n",
-       __FUNCTION__, (void *) *pt, (*pt)->refcount - 1);
-   */
    if (--(*pt)->refcount <= 0) {
       struct softpipe_texture *spt = softpipe_texture(*pt);
 
-      /*
-      DBG("%s deleting %p\n", __FUNCTION__, (void *) spt);
-      */
-
       pipe_buffer_reference(screen->winsys, &spt->buffer, NULL);
-
       FREE(spt);
    }
    *pt = NULL;
@@ -138,7 +180,8 @@ softpipe_texture_release(struct pipe_screen *screen,
 static struct pipe_surface *
 softpipe_get_tex_surface(struct pipe_screen *screen,
                          struct pipe_texture *pt,
-                         unsigned face, unsigned level, unsigned zslice)
+                         unsigned face, unsigned level, unsigned zslice,
+                         unsigned usage)
 {
    struct pipe_winsys *ws = screen->winsys;
    struct softpipe_texture *spt = softpipe_texture(pt);
@@ -157,6 +200,24 @@ softpipe_get_tex_surface(struct pipe_screen *screen,
       ps->height = pt->height[level];
       ps->pitch = ps->width;
       ps->offset = spt->level_offset[level];
+      ps->usage = usage;
+      
+      /* Because we are softpipe, anything that the state tracker
+       * thought was going to be done with the GPU will actually get
+       * done with the CPU.  Let's adjust the flags to take that into
+       * account.
+       */
+      if (ps->usage & PIPE_BUFFER_USAGE_GPU_WRITE)
+         ps->usage |= PIPE_BUFFER_USAGE_CPU_WRITE;
+
+      if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ)
+         ps->usage |= PIPE_BUFFER_USAGE_CPU_READ;
+
+
+      pipe_texture_reference(&ps->texture, pt); 
+      ps->face = face;
+      ps->level = level;
+      ps->zslice = zslice;
 
       if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) {
 	 ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) *
@@ -172,25 +233,64 @@ softpipe_get_tex_surface(struct pipe_screen *screen,
 }
 
 
-static void
-softpipe_texture_update(struct pipe_context *pipe,
-                        struct pipe_texture *texture,
-                        uint face, uint levelsMask)
+static void 
+softpipe_tex_surface_release(struct pipe_screen *screen, 
+                             struct pipe_surface **s)
 {
-   struct softpipe_context *softpipe = softpipe_context(pipe);
-   uint unit;
-   for (unit = 0; unit < softpipe->num_textures; unit++) {
-      if (softpipe->texture[unit] == texture) {
-         sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]);
-      }
+   /* Effectively do the texture_update work here - if texture images
+    * needed post-processing to put them into hardware layout, this is
+    * where it would happen.  For softpipe, nothing to do.
+    */
+   assert ((*s)->texture);
+   pipe_texture_reference(&(*s)->texture, NULL); 
+
+   screen->winsys->surface_release(screen->winsys, s);
+}
+
+
+static void *
+softpipe_surface_map( struct pipe_screen *screen,
+                      struct pipe_surface *surface,
+                      unsigned flags )
+{
+   ubyte *map;
+
+   if (flags & ~surface->usage) {
+      assert(0);
+      return NULL;
+   }
+
+   map = screen->winsys->buffer_map( screen->winsys, surface->buffer, flags );
+   if (map == NULL)
+      return NULL;
+
+   /* May want to different things here depending on read/write nature
+    * of the map:
+    */
+   if (surface->texture &&
+       (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) 
+   {
+      /* Do something to notify sharing contexts of a texture change.
+       * In softpipe, that would mean flushing the texture cache.
+       */
+      softpipe_screen(screen)->timestamp++;
    }
+   
+   return map + surface->offset;
+}
+
+
+static void
+softpipe_surface_unmap(struct pipe_screen *screen,
+                       struct pipe_surface *surface)
+{
+   screen->winsys->buffer_unmap( screen->winsys, surface->buffer );
 }
 
 
 void
-softpipe_init_texture_funcs( struct softpipe_context *softpipe )
+softpipe_init_texture_funcs(struct softpipe_context *sp)
 {
-   softpipe->pipe.texture_update = softpipe_texture_update;
 }
 
 
@@ -199,5 +299,10 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen)
 {
    screen->texture_create = softpipe_texture_create;
    screen->texture_release = softpipe_texture_release;
+
    screen->get_tex_surface = softpipe_get_tex_surface;
+   screen->tex_surface_release = softpipe_tex_surface_release;
+
+   screen->surface_map = softpipe_surface_map;
+   screen->surface_unmap = softpipe_surface_unmap;
 }
diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h
index a7322144e6f..779a9d8fc97 100644
--- a/src/gallium/drivers/softpipe/sp_texture.h
+++ b/src/gallium/drivers/softpipe/sp_texture.h
@@ -42,11 +42,11 @@ struct softpipe_texture
    struct pipe_texture base;
 
    unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned long pitch[PIPE_MAX_TEXTURE_LEVELS];
 
    /* The data is held here:
     */
    struct pipe_buffer *buffer;
-   unsigned long buffer_size;
 };
 
 
@@ -61,7 +61,6 @@ softpipe_texture(struct pipe_texture *pt)
 extern void
 softpipe_init_texture_funcs( struct softpipe_context *softpipe );
 
-
 extern void
 softpipe_init_screen_texture_funcs(struct pipe_screen *screen);
 
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 1117c0ad4c7..28c29da87c9 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -49,6 +49,7 @@
 
 struct softpipe_tile_cache
 {
+   struct pipe_screen *screen;
    struct pipe_surface *surface;  /**< the surface we're caching */
    void *surface_map;
    struct pipe_texture *texture;  /**< if caching a texture */
@@ -109,13 +110,14 @@ clear_clear_flag(uint *bitvec, int x, int y)
    
 
 struct softpipe_tile_cache *
-sp_create_tile_cache(void)
+sp_create_tile_cache( struct pipe_screen *screen )
 {
    struct softpipe_tile_cache *tc;
    uint pos;
 
    tc = CALLOC_STRUCT( softpipe_tile_cache );
    if (tc) {
+      tc->screen = screen;
       for (pos = 0; pos < NUM_ENTRIES; pos++) {
          tc->entries[pos].x =
          tc->entries[pos].y = -1;
@@ -154,16 +156,17 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc,
    assert(!tc->texture);
 
    if (tc->surface_map) {
-      /*assert(tc->surface != ps);*/
-      pipe_surface_unmap(tc->surface);
+      tc->screen->surface_unmap(tc->screen, tc->surface);
       tc->surface_map = NULL;
    }
 
    pipe_surface_reference(&tc->surface, ps);
 
-   if (ps) {
-      if (tc->surface_map)
-	 tc->surface_map = pipe_surface_map(ps);
+   if (tc->surface) {
+      if (tc->surface_map) /* XXX: this is always NULL!? */
+	 tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface,
+                                                   PIPE_BUFFER_USAGE_CPU_READ | 
+                                                   PIPE_BUFFER_USAGE_CPU_WRITE);
 
       tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM ||
                            ps->format == PIPE_FORMAT_Z16_UNORM ||
@@ -187,10 +190,13 @@ void
 sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc)
 {
    if (tc->surface && !tc->surface_map)
-      tc->surface_map = pipe_surface_map(tc->surface);
+      tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface,
+                                                PIPE_BUFFER_USAGE_CPU_WRITE |
+                                                PIPE_BUFFER_USAGE_CPU_READ);
 
    if (tc->tex_surf && !tc->tex_surf_map)
-      tc->tex_surf_map = pipe_surface_map(tc->tex_surf);
+      tc->tex_surf_map = tc->screen->surface_map(tc->screen, tc->tex_surf,
+                                                 PIPE_BUFFER_USAGE_CPU_READ);
 }
 
 
@@ -198,12 +204,12 @@ void
 sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc)
 {
    if (tc->surface_map) {
-      pipe_surface_unmap(tc->surface);
+      tc->screen->surface_unmap(tc->screen, tc->surface);
       tc->surface_map = NULL;
    }
 
    if (tc->tex_surf_map) {
-      pipe_surface_unmap(tc->tex_surf);
+      tc->screen->surface_unmap(tc->screen, tc->tex_surf);
       tc->tex_surf_map = NULL;
    }
 }
@@ -224,7 +230,7 @@ sp_tile_cache_set_texture(struct pipe_context *pipe,
    pipe_texture_reference(&tc->texture, texture);
 
    if (tc->tex_surf_map) {
-      pipe_surface_unmap(tc->tex_surf);
+      tc->screen->surface_unmap(tc->screen, tc->tex_surf);
       tc->tex_surf_map = NULL;
    }
    pipe_surface_reference(&tc->tex_surf, NULL);
@@ -514,10 +520,12 @@ sp_get_cached_tile_tex(struct pipe_context *pipe,
          /* get new surface (view into texture) */
 
 	 if (tc->tex_surf_map)
-            pipe_surface_unmap(tc->tex_surf);
+            tc->screen->surface_unmap(tc->screen, tc->tex_surf);
 
-         tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z);
-         tc->tex_surf_map = pipe_surface_map(tc->tex_surf);
+         tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z, 
+                                                PIPE_BUFFER_USAGE_CPU_READ);
+         tc->tex_surf_map = screen->surface_map(screen, tc->tex_surf,
+                                                PIPE_BUFFER_USAGE_CPU_READ);
 
          tc->tex_face = face;
          tc->tex_level = level;
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h
index 2631e29a3a6..bc96c941f61 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.h
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.h
@@ -61,7 +61,7 @@ struct softpipe_cached_tile
 
 
 extern struct softpipe_tile_cache *
-sp_create_tile_cache(void);
+sp_create_tile_cache( struct pipe_screen *screen );
 
 extern void
 sp_destroy_tile_cache(struct softpipe_tile_cache *tc);
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index f3a9c2cd8b2..0f68f592f77 100644
--- a/src/gallium/include/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -198,12 +198,6 @@ struct pipe_context {
    /*@}*/
 
 
-   /** Called when texture data is changed */
-   void (*texture_update)(struct pipe_context *pipe,
-                          struct pipe_texture *texture,
-                          uint face, uint dirtyLevelsMask);
-
-
    /** Flush rendering (flags = bitmask of PIPE_FLUSH_x tokens) */
    void (*flush)( struct pipe_context *pipe,
                   unsigned flags,
diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h
index 7a7312ea122..0af635be573 100644
--- a/src/gallium/include/pipe/p_debug.h
+++ b/src/gallium/include/pipe/p_debug.h
@@ -313,6 +313,28 @@ void
 debug_memory_end(unsigned long beginning);
 
 
+#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+
+void
+debug_profile_start(void);
+
+void 
+debug_profile_stop(void);
+
+#endif
+
+
+#ifdef DEBUG
+void debug_dump_image(const char *prefix,
+                      unsigned format, unsigned cpp,
+                      unsigned width, unsigned height,
+                      unsigned pitch,
+                      const void *data);
+#else
+#define debug_dump_image(prefix, format, cpp, width, height, pitch, data) ((void)0)
+#endif
+
+
 #ifdef	__cplusplus
 }
 #endif
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 46a355daa21..83330ef22fe 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -174,6 +174,12 @@ enum pipe_texture_target {
 
 
 /**
+ * Surface layout
+ */
+#define PIPE_SURFACE_LAYOUT_LINEAR  0
+
+
+/**
  * Surface status
  */
 #define PIPE_SURFACE_STATUS_UNDEFINED  0
diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h
index a4bd23c302a..9ba00f8d7b6 100644
--- a/src/gallium/include/pipe/p_format.h
+++ b/src/gallium/include/pipe/p_format.h
@@ -307,6 +307,9 @@ enum pipe_format {
    PIPE_FORMAT_R8G8B8_SNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ),
    PIPE_FORMAT_R8G8B8A8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
    PIPE_FORMAT_R8G8B8X8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_B6G5R5_SNORM          = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_A8B8G8R8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_X8B8G8R8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
    PIPE_FORMAT_R8_SSCALED            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
    PIPE_FORMAT_R8G8_SSCALED          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
    PIPE_FORMAT_R8G8B8_SSCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ),
diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h
index 8eb604e73f1..1e4b98edb48 100644
--- a/src/gallium/include/pipe/p_inlines.h
+++ b/src/gallium/include/pipe/p_inlines.h
@@ -39,21 +39,40 @@ extern "C" {
 #endif
 
 
+/* XXX: these are a kludge.  will fix when all surfaces are views into
+ * textures, and free-floating winsys surfaces go away.
+ */
 static INLINE void *
-pipe_surface_map(struct pipe_surface *surface)
+pipe_surface_map( struct pipe_surface *surf, unsigned flags )
 {
-   return (char *)surface->winsys->buffer_map( surface->winsys, surface->buffer,
-					       PIPE_BUFFER_USAGE_CPU_WRITE |
-					       PIPE_BUFFER_USAGE_CPU_READ )
-      + surface->offset;
+   if (surf->texture) {
+      struct pipe_screen *screen = surf->texture->screen;
+      return surf->texture->screen->surface_map( screen, surf, flags );
+   }
+   else {
+      struct pipe_winsys *winsys = surf->winsys;
+      char *map = (char *)winsys->buffer_map( winsys, surf->buffer, flags );
+      if (map == NULL)
+         return NULL;
+      return (void *)(map + surf->offset);
+   }
 }
 
 static INLINE void
-pipe_surface_unmap(struct pipe_surface *surface)
+pipe_surface_unmap( struct pipe_surface *surf )
 {
-   surface->winsys->buffer_unmap( surface->winsys, surface->buffer );
+   if (surf->texture) {
+      struct pipe_screen *screen = surf->texture->screen;
+      surf->texture->screen->surface_unmap( screen, surf );
+   }
+   else {
+      struct pipe_winsys *winsys = surf->winsys;
+      winsys->buffer_unmap( winsys, surf->buffer );
+   }
 }
 
+
+
 /**
  * Set 'ptr' to point to 'surf' and update reference counting.
  * The old thing pointed to, if any, will be unreferenced first.
@@ -66,9 +85,20 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf)
    if (surf) 
       surf->refcount++;
 
-   if (*ptr /* && --(*ptr)->refcount == 0 */) {
-      struct pipe_winsys *winsys = (*ptr)->winsys;
-      winsys->surface_release(winsys, ptr);
+   if (*ptr) {
+
+      /* There are currently two sorts of surfaces... This needs to be
+       * fixed so that all surfaces are views into a texture.
+       */
+      if ((*ptr)->texture) {
+         struct pipe_screen *screen = (*ptr)->texture->screen;
+         screen->tex_surface_release( screen, ptr );
+      }
+      else {
+         struct pipe_winsys *winsys = (*ptr)->winsys;
+         winsys->surface_release(winsys, ptr);
+      }
+
       assert(!*ptr);
    }
 
diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h
index 26ac99d287d..cc8430dae16 100644
--- a/src/gallium/include/pipe/p_screen.h
+++ b/src/gallium/include/pipe/p_screen.h
@@ -89,6 +89,18 @@ struct pipe_screen {
    struct pipe_texture * (*texture_create)(struct pipe_screen *,
                                            const struct pipe_texture *templat);
 
+   /**
+    * Create a new texture object, using the given template info, but on top of 
+    * existing memory.
+    * 
+    * It is assumed that the buffer data is layed out according to the expected
+    * by the hardware. NULL will be returned if any inconsistency is found.  
+    */
+   struct pipe_texture * (*texture_blanket)(struct pipe_screen *,
+                                            const struct pipe_texture *templat,
+                                            const unsigned *pitch,
+                                            struct pipe_buffer *buffer);
+
    void (*texture_release)(struct pipe_screen *,
                            struct pipe_texture **pt);
 
@@ -96,7 +108,22 @@ struct pipe_screen {
    struct pipe_surface *(*get_tex_surface)(struct pipe_screen *,
                                            struct pipe_texture *texture,
                                            unsigned face, unsigned level,
-                                           unsigned zslice);
+                                           unsigned zslice,
+                                           unsigned usage );
+
+   /* Surfaces allocated by the above must be released here:
+    */
+   void (*tex_surface_release)( struct pipe_screen *,
+                                struct pipe_surface ** );
+   
+
+   void *(*surface_map)( struct pipe_screen *,
+                         struct pipe_surface *surface,
+                         unsigned flags );
+
+   void (*surface_unmap)( struct pipe_screen *,
+                          struct pipe_surface *surface );
+   
 };
 
 
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 912d84e7b99..e7ee8c97ed7 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -271,12 +271,26 @@ struct pipe_surface
    unsigned width;
    unsigned height;
    unsigned pitch;               /**< in pixels */
+   unsigned layout;              /**< PIPE_SURFACE_LAYOUT_x */
    unsigned offset;              /**< offset from start of buffer, in bytes */
    unsigned refcount;
+   unsigned usage;              /**< PIPE_BUFFER_USAGE_*  */
+
    struct pipe_winsys *winsys;   /**< winsys which owns/created the surface */
+
+   struct pipe_texture *texture; /**< optional texture into which this is a view  */
+   unsigned face;
+   unsigned level;
+   unsigned zslice;
 };
 
 
+#define PIPE_TEXTURE_USAGE_RENDER_TARGET   0x1
+#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2 /* ie a backbuffer */
+#define PIPE_TEXTURE_USAGE_PRIMARY         0x4 /* ie a frontbuffer */
+#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL   0x8
+#define PIPE_TEXTURE_USAGE_SAMPLER         0x10
+
 /**
  * Texture object.
  */
@@ -292,6 +306,8 @@ struct pipe_texture
    unsigned cpp:8;
    unsigned last_level:8;    /**< Index of last mipmap level present/defined */
    unsigned compressed:1;
+   
+   unsigned tex_usage;          /* PIPE_TEXTURE_USAGE_* */
 
    /* These are also refcounted:
     */
diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h
index a66ab765f72..597354ca96c 100644
--- a/src/gallium/include/pipe/p_util.h
+++ b/src/gallium/include/pipe/p_util.h
@@ -204,7 +204,6 @@ mem_dup(const void *src, uint size)
 #define MIN2( A, B )   ( (A)<(B) ? (A) : (B) )
 #define MAX2( A, B )   ( (A)>(B) ? (A) : (B) )
 
-/* JB stop warnings */
 #ifndef Elements
 #define Elements(x) (sizeof(x)/sizeof((x)[0]))
 #endif
diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h
index 3005ec2d941..7ebc2851928 100644
--- a/src/gallium/include/pipe/p_winsys.h
+++ b/src/gallium/include/pipe/p_winsys.h
@@ -86,11 +86,12 @@ struct pipe_winsys
                                 struct pipe_surface *surf,
                                 unsigned width, unsigned height,
                                 enum pipe_format format,
-                                unsigned flags);
+                                unsigned flags,
+                                unsigned tex_usage);
    
    void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s);
 
-   
+
    /**
     * Buffer management. Buffer attributes are mostly fixed over its lifetime.
     *
diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c
index 6a42c790337..10e1a84cc12 100644
--- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c
+++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c
@@ -220,7 +220,8 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys,
                                  struct pipe_surface *surf,
                                  unsigned width, unsigned height,
                                  enum pipe_format format, 
-                                 unsigned flags)
+                                 unsigned flags,
+                                 unsigned tex_usage)
 {
    const unsigned alignment = 64;
    //int ret;
diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h
index f1fd3939ab9..f84f4533093 100644
--- a/src/gallium/winsys/dri/intel/server/i830_common.h
+++ b/src/gallium/winsys/dri/intel/server/i830_common.h
@@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 **************************************************************************/
 
-/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */
 
 #ifndef _I830_COMMON_H_
 #define _I830_COMMON_H_
diff --git a/src/gallium/winsys/dri/intel/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h
index c2a3af8cbf7..685de4a551b 100644
--- a/src/gallium/winsys/dri/intel/server/i830_dri.h
+++ b/src/gallium/winsys/dri/intel/server/i830_dri.h
@@ -1,4 +1,3 @@
-/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */
 
 #ifndef _I830_DRI_H
 #define _I830_DRI_H
diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c
index 14c3892559b..b14758f3336 100644
--- a/src/gallium/winsys/xlib/xm_winsys.c
+++ b/src/gallium/winsys/xlib/xm_winsys.c
@@ -489,7 +489,8 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys,
                          struct pipe_surface *surf,
                          unsigned width, unsigned height,
                          enum pipe_format format, 
-                         unsigned flags)
+                         unsigned flags,
+                         unsigned tex_usage)
 {
    const unsigned alignment = 64;
 
@@ -498,6 +499,7 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys,
    surf->format = format;
    surf->cpp = pf_get_size(format);
    surf->pitch = round_up(width, alignment / surf->cpp);
+   surf->usage = flags;
 
 #ifdef GALLIUM_CELL /* XXX a bit of a hack */
    height = round_up(height, TILE_SIZE);
@@ -551,6 +553,7 @@ static void
 xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s)
 {
    struct pipe_surface *surf = *s;
+   assert(!surf->texture);
    surf->refcount--;
    if (surf->refcount == 0) {
       if (surf->buffer)
diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c
index f42f7fcc5f1..77376099f0d 100644
--- a/src/gallium/winsys/xlib/xm_winsys_aub.c
+++ b/src/gallium/winsys/xlib/xm_winsys_aub.c
@@ -276,7 +276,8 @@ aub_i915_surface_alloc_storage(struct pipe_winsys *winsys,
                                struct pipe_surface *surf,
                                unsigned width, unsigned height,
                                enum pipe_format format,
-                               unsigned flags)
+                               unsigned flags,
+                               unsigned tex_usage)
 {
     const unsigned alignment = 64;