Merge remote branch 'origin/master' into gallium_draw_llvm

136 files changed, 2833 insertions, 1483 deletions

diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h
index 28f80b82cd5..f7e2284445d 100644
--- a/src/gallium/drivers/cell/ppu/cell_context.h
+++ b/src/gallium/drivers/cell/ppu/cell_context.h
@@ -97,7 +97,7 @@ struct cell_velems_state
 {
    unsigned count;
    struct pipe_vertex_element velem[PIPE_MAX_ATTRIBS];
-}
+};
 
 /**
  * Per-context state, subclass of pipe_context.
@@ -127,6 +127,7 @@ struct cell_context
    struct pipe_poly_stipple poly_stipple;
    struct pipe_scissor_state scissor;
    struct cell_texture *texture[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
    uint num_textures;
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
index dce10ae7f86..059ce8597bc 100644
--- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c
+++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
@@ -257,8 +257,9 @@ cell_delete_sampler_state(struct pipe_context *pipe,
 
 
 static void
-cell_set_sampler_textures(struct pipe_context *pipe,
-                          unsigned num, struct pipe_texture **texture)
+cell_set_fragment_sampler_views(struct pipe_context *pipe,
+                                unsigned num,
+                                struct pipe_sampler_view **views)
 {
    struct cell_context *cell = cell_context(pipe);
    uint i, changed = 0x0;
@@ -266,10 +267,14 @@ cell_set_sampler_textures(struct pipe_context *pipe,
    assert(num <= CELL_MAX_SAMPLERS);
 
    for (i = 0; i < CELL_MAX_SAMPLERS; i++) {
-      struct cell_texture *new_tex = cell_texture(i < num ? texture[i] : NULL);
-      struct cell_texture *old_tex = cell->texture[i];
-      if (old_tex != new_tex) {
+      struct pipe_sampler_view *new_view = i < num ? views[i] : NULL;
+      struct pipe_sampler_view *old_view = cell->fragment_sampler_views[i];
 
+      if (old_view != new_view) {
+         struct pipe_texture *new_tex = new_view ? new_view->texture : NULL;
+
+         pipe_sampler_view_reference(&cell->fragment_sampler_views[i],
+                                     views[i]);
          pipe_texture_reference((struct pipe_texture **) &cell->texture[i],
                                 (struct pipe_texture *) new_tex);
 
@@ -286,6 +291,34 @@ cell_set_sampler_textures(struct pipe_context *pipe,
 }
 
 
+static struct pipe_sampler_view *
+cell_create_sampler_view(struct pipe_context *pipe,
+                         struct pipe_texture *texture,
+                         const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+
+static void
+cell_sampler_view_destroy(struct pipe_context *pipe,
+                          struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
+}
+
+
 /**
  * Map color and z/stencil framebuffer surfaces.
  */
@@ -409,7 +442,9 @@ cell_init_state_functions(struct cell_context *cell)
    cell->pipe.bind_fragment_sampler_states = cell_bind_sampler_states;
    cell->pipe.delete_sampler_state = cell_delete_sampler_state;
 
-   cell->pipe.set_fragment_sampler_textures = cell_set_sampler_textures;
+   cell->pipe.set_fragment_sampler_views = cell_set_fragment_sampler_views;
+   cell->pipe.create_sampler_view = cell_create_sampler_view;
+   cell->pipe.sampler_view_destroy = cell_sampler_view_destroy;
 
    cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state;
    cell->pipe.bind_depth_stencil_alpha_state   = cell_bind_depth_stencil_alpha_state;
diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c
index 31fd963d19a..42045744060 100644
--- a/src/gallium/drivers/cell/ppu/cell_screen.c
+++ b/src/gallium/drivers/cell/ppu/cell_screen.c
@@ -97,6 +97,8 @@ cell_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT:
    case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER:
       return 0;
+   case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+      return 1;
    default:
       return 0;
    }
@@ -136,20 +138,23 @@ cell_is_format_supported( struct pipe_screen *screen,
                           unsigned tex_usage, 
                           unsigned geom_flags )
 {
-   struct sw_winsys *winsys = cell_screen(screen)->winsys;
-
-   if (format == PIPE_FORMAT_DXT5_RGBA ||
-       format == PIPE_FORMAT_A8B8G8R8_SRGB)
-      return FALSE;
-
-   if (tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
-      if (!winsys->is_displaytarget_format_supported(winsys, format))
+   if (tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                    PIPE_TEXTURE_USAGE_SCANOUT |
+                    PIPE_TEXTURE_USAGE_SHARED)) {
+      if (!winsys->is_displaytarget_format_supported(winsys, tex_usage, format))
          return FALSE;
    }
 
-   /* This is often a lie.  Pull in logic from llvmpipe to fix.
-    */
-   return TRUE;
+   /* only a few formats are known to work at this time */
+   switch (format) {
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_I8_UNORM:
+      return TRUE;
+   default:
+      return FALSE;
+   }
 }
 
 
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index c65c3b4f885..5b169afaf88 100644
--- a/src/gallium/drivers/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -105,6 +105,7 @@ cell_displaytarget_layout(struct pipe_screen *screen,
    /* Round up the surface size to a multiple of the tile size?
     */
    ct->dt = winsys->displaytarget_create(winsys,
+                                          ct->base->tex_usage,
                                           ct->base.format,
                                           ct->base.width0, 
                                           ct->base.height0,
diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h
index 4a754465bbe..73031321d98 100644
--- a/src/gallium/drivers/failover/fo_context.h
+++ b/src/gallium/drivers/failover/fo_context.h
@@ -47,7 +47,7 @@
 #define FO_NEW_ALPHA_TEST      0x100
 #define FO_NEW_DEPTH_STENCIL   0x200
 #define FO_NEW_SAMPLER         0x400
-#define FO_NEW_TEXTURE         0x800
+#define FO_NEW_SAMPLER_VIEW    0x800
 #define FO_NEW_VERTEX          0x2000
 #define FO_NEW_VERTEX_SHADER   0x4000
 #define FO_NEW_BLEND_COLOR     0x8000
@@ -65,6 +65,13 @@ struct fo_state {
    void *sw_state;
    void *hw_state;
 };
+
+struct fo_sampler_view {
+   struct pipe_sampler_view base;
+   struct pipe_sampler_view *sw;
+   struct pipe_sampler_view *hw;
+};
+
 struct failover_context {
    struct pipe_context pipe;  /**< base class */
 
@@ -86,8 +93,6 @@ struct failover_context {
    struct pipe_framebuffer_state framebuffer;
    struct pipe_poly_stipple poly_stipple;
    struct pipe_scissor_state scissor;
-   struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
-   struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
 
@@ -98,12 +103,15 @@ struct failover_context {
    void *sw_vertex_sampler_state[PIPE_MAX_VERTEX_SAMPLERS];
    void *hw_vertex_sampler_state[PIPE_MAX_VERTEX_SAMPLERS];
 
+   struct fo_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
+   struct fo_sampler_view *vertex_sampler_views[PIPE_MAX_VERTEX_SAMPLERS];
+   unsigned num_fragment_sampler_views;
+   unsigned num_vertex_sampler_views;
+
    unsigned dirty;
 
    unsigned num_samplers;
    unsigned num_vertex_samplers;
-   unsigned num_textures;
-   unsigned num_vertex_textures;
 
    unsigned mode;
    struct pipe_context *hw;
diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c
index 0247fb803b2..25c62735705 100644
--- a/src/gallium/drivers/failover/fo_state.c
+++ b/src/gallium/drivers/failover/fo_state.c
@@ -447,60 +447,96 @@ failover_delete_sampler_state(struct pipe_context *pipe, void *sampler)
 }
 
 
+static struct pipe_sampler_view *
+failover_create_sampler_view(struct pipe_context *pipe,
+                             struct pipe_texture *texture,
+                             const struct pipe_sampler_view *templ)
+{
+   struct fo_sampler_view *view = malloc(sizeof(struct fo_sampler_view));
+   struct failover_context *failover = failover_context(pipe);
+
+   view->sw = failover->sw->create_sampler_view(failover->sw, texture, templ);
+   view->hw = failover->hw->create_sampler_view(failover->hw, texture, templ);
+
+   view->base = *templ;
+   view->base.reference.count = 1;
+   view->base.texture = NULL;
+   pipe_texture_reference(&view->base.texture, texture);
+   view->base.context = pipe;
+
+   return &view->base;
+}
+
+static void
+failover_sampler_view_destroy(struct pipe_context *pipe,
+                              struct pipe_sampler_view *view)
+{
+   struct fo_sampler_view *fo_view = (struct fo_sampler_view *)view;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->sampler_view_destroy(failover->sw, fo_view->sw);
+   failover->hw->sampler_view_destroy(failover->hw, fo_view->hw);
+
+   pipe_texture_reference(&fo_view->base.texture, NULL);
+   free(fo_view);
+}
+
 static void
-failover_set_fragment_sampler_textures(struct pipe_context *pipe,
-                                       unsigned num,
-                                       struct pipe_texture **texture)
+failover_set_fragment_sampler_views(struct pipe_context *pipe,
+                                    unsigned num,
+                                    struct pipe_sampler_view **views)
 {
    struct failover_context *failover = failover_context(pipe);
+   struct pipe_sampler_view *hw_views[PIPE_MAX_SAMPLERS];
    uint i;
 
    assert(num <= PIPE_MAX_SAMPLERS);
 
    /* Check for no-op */
-   if (num == failover->num_textures &&
-       !memcmp(failover->texture, texture, num * sizeof(struct pipe_texture *)))
+   if (num == failover->num_fragment_sampler_views &&
+       !memcmp(failover->fragment_sampler_views, views, num * sizeof(struct pipe_sampler_view *)))
       return;
-   for (i = 0; i < num; i++)
-      pipe_texture_reference((struct pipe_texture **) &failover->texture[i],
-                             texture[i]);
-   for (i = num; i < failover->num_textures; i++)
-      pipe_texture_reference((struct pipe_texture **) &failover->texture[i],
-                             NULL);
-   failover->dirty |= FO_NEW_TEXTURE;
-   failover->num_textures = num;
-   failover->sw->set_fragment_sampler_textures( failover->sw, num, texture );
-   failover->hw->set_fragment_sampler_textures( failover->hw, num, texture );
+   for (i = 0; i < num; i++) {
+      struct fo_sampler_view *fo_view = (struct fo_sampler_view *)views[i];
+
+      pipe_sampler_view_reference((struct pipe_sampler_view **)&failover->fragment_sampler_views[i], views[i]);
+      hw_views[i] = fo_view->hw;
+   }
+   for (i = num; i < failover->num_fragment_sampler_views; i++)
+      pipe_sampler_view_reference((struct pipe_sampler_view **)&failover->fragment_sampler_views[i], NULL);
+   failover->dirty |= FO_NEW_SAMPLER_VIEW;
+   failover->num_fragment_sampler_views = num;
+   failover->hw->set_fragment_sampler_views(failover->hw, num, hw_views);
 }
 
 
 static void
-failover_set_vertex_sampler_textures(struct pipe_context *pipe,
-                                     unsigned num_textures,
-                                     struct pipe_texture **textures)
+failover_set_vertex_sampler_views(struct pipe_context *pipe,
+                                  unsigned num,
+                                  struct pipe_sampler_view **views)
 {
    struct failover_context *failover = failover_context(pipe);
+   struct pipe_sampler_view *hw_views[PIPE_MAX_VERTEX_SAMPLERS];
    uint i;
 
-   assert(num_textures <= PIPE_MAX_VERTEX_SAMPLERS);
+   assert(num <= PIPE_MAX_VERTEX_SAMPLERS);
 
    /* Check for no-op */
-   if (num_textures == failover->num_vertex_textures &&
-       !memcmp(failover->vertex_textures, textures, num_textures * sizeof(struct pipe_texture *))) {
+   if (num == failover->num_vertex_sampler_views &&
+       !memcmp(failover->vertex_sampler_views, views, num * sizeof(struct pipe_sampler_view *))) {
       return;
    }
-   for (i = 0; i < num_textures; i++) {
-      pipe_texture_reference((struct pipe_texture **)&failover->vertex_textures[i],
-                             textures[i]);
-   }
-   for (i = num_textures; i < failover->num_vertex_textures; i++) {
-      pipe_texture_reference((struct pipe_texture **)&failover->vertex_textures[i],
-                             NULL);
+   for (i = 0; i < num; i++) {
+      struct fo_sampler_view *fo_view = (struct fo_sampler_view *)views[i];
+
+      pipe_sampler_view_reference((struct pipe_sampler_view **)&failover->vertex_sampler_views[i], views[i]);
+      hw_views[i] = fo_view->hw;
    }
-   failover->dirty |= FO_NEW_TEXTURE;
-   failover->num_vertex_textures = num_textures;
-   failover->sw->set_vertex_sampler_textures(failover->sw, num_textures, textures);
-   failover->hw->set_vertex_sampler_textures(failover->hw, num_textures, textures);
+   for (i = num; i < failover->num_vertex_sampler_views; i++)
+      pipe_sampler_view_reference((struct pipe_sampler_view **)&failover->vertex_sampler_views[i], NULL);
+   failover->dirty |= FO_NEW_SAMPLER_VIEW;
+   failover->num_vertex_sampler_views = num;
+   failover->hw->set_vertex_sampler_views(failover->hw, num, hw_views);
 }
 
 
@@ -580,9 +616,11 @@ failover_init_state_functions( struct failover_context *failover )
    failover->pipe.set_framebuffer_state = failover_set_framebuffer_state;
    failover->pipe.set_polygon_stipple = failover_set_polygon_stipple;
    failover->pipe.set_scissor_state = failover_set_scissor_state;
-   failover->pipe.set_fragment_sampler_textures = failover_set_fragment_sampler_textures;
-   failover->pipe.set_vertex_sampler_textures = failover_set_vertex_sampler_textures;
+   failover->pipe.set_fragment_sampler_views = failover_set_fragment_sampler_views;
+   failover->pipe.set_vertex_sampler_views = failover_set_vertex_sampler_views;
    failover->pipe.set_viewport_state = failover_set_viewport_state;
    failover->pipe.set_vertex_buffers = failover_set_vertex_buffers;
    failover->pipe.set_constant_buffer = failover_set_constant_buffer;
+   failover->pipe.create_sampler_view = failover_create_sampler_view;
+   failover->pipe.sampler_view_destroy = failover_sampler_view_destroy;
 }
diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c
index 09ca1944971..42bd6929a7f 100644
--- a/src/gallium/drivers/failover/fo_state_emit.c
+++ b/src/gallium/drivers/failover/fo_state_emit.c
@@ -106,12 +106,24 @@ failover_state_emit( struct failover_context *failover )
                                                failover->sw_vertex_sampler_state);
    }
 
-   if (failover->dirty & FO_NEW_TEXTURE) {
-      failover->sw->set_fragment_sampler_textures( failover->sw, failover->num_textures, 
-                                                   failover->texture );
-      failover->sw->set_vertex_sampler_textures(failover->sw,
-                                                failover->num_vertex_textures, 
-                                                failover->vertex_textures);
+   if (failover->dirty & FO_NEW_SAMPLER_VIEW) {
+      struct pipe_sampler_view *fragment_views[PIPE_MAX_SAMPLERS];
+      struct pipe_sampler_view *vertex_views[PIPE_MAX_VERTEX_SAMPLERS];
+      uint i;
+
+      for (i = 0; i < failover->num_fragment_sampler_views; i++) {
+         fragment_views[i] = failover->fragment_sampler_views[i]->sw;
+      }
+      failover->sw->set_fragment_sampler_views(failover->sw,
+                                               failover->num_fragment_sampler_views,
+                                               fragment_views);
+
+      for (i = 0; i < failover->num_vertex_sampler_views; i++) {
+         vertex_views[i] = failover->vertex_sampler_views[i]->sw;
+      }
+      failover->sw->set_vertex_sampler_views(failover->sw,
+                                             failover->num_vertex_sampler_views,
+                                             vertex_views);
    }
 
    if (failover->dirty & FO_NEW_VERTEX_BUFFER) {
diff --git a/src/gallium/drivers/i915/i915_batch.h b/src/gallium/drivers/i915/i915_batch.h
index b813784723f..f0086695d16 100644
--- a/src/gallium/drivers/i915/i915_batch.h
+++ b/src/gallium/drivers/i915/i915_batch.h
@@ -28,19 +28,19 @@
 #ifndef I915_BATCH_H
 #define I915_BATCH_H
 
-#include "intel_batchbuffer.h"
+#include "i915_batchbuffer.h"
 
 #define BEGIN_BATCH(dwords, relocs) \
-   (intel_batchbuffer_check(i915->batch, dwords, relocs))
+   (i915_winsys_batchbuffer_check(i915->batch, dwords, relocs))
 
 #define OUT_BATCH(dword) \
-   intel_batchbuffer_dword(i915->batch, dword)
+   i915_winsys_batchbuffer_dword(i915->batch, dword)
 
 #define OUT_RELOC(buf, usage, offset) \
-   intel_batchbuffer_reloc(i915->batch, buf, usage, offset)
+   i915_winsys_batchbuffer_reloc(i915->batch, buf, usage, offset)
 
 #define FLUSH_BATCH(fence) do {                 \
-   intel_batchbuffer_flush(i915->batch, fence); \
+   i915_winsys_batchbuffer_flush(i915->batch, fence); \
    i915->hardware_dirty = ~0;                   \
 } while (0)
 
diff --git a/src/gallium/drivers/i915/intel_batchbuffer.h b/src/gallium/drivers/i915/i915_batchbuffer.h
index db12dfd2ac2..27ccaa6b1fa 100644
--- a/src/gallium/drivers/i915/intel_batchbuffer.h
+++ b/src/gallium/drivers/i915/i915_batchbuffer.h
@@ -1,8 +1,8 @@
 /**************************************************************************
- * 
+ *
  * 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
@@ -10,11 +10,11 @@
  * 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.
@@ -22,34 +22,34 @@
  * 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 INTEL_BATCH_H
-#define INTEL_BATCH_H
+#ifndef I915_BATCHBUFFER_H
+#define I915_BATCHBUFFER_H
 
-#include "intel_winsys.h"
+#include "i915_winsys.h"
 
 static INLINE boolean
-intel_batchbuffer_check(struct intel_batchbuffer *batch,
-                        size_t dwords,
-                        size_t relocs)
+i915_winsys_batchbuffer_check(struct i915_winsys_batchbuffer *batch,
+                              size_t dwords,
+                              size_t relocs)
 {
    return dwords * 4 <= batch->size - (batch->ptr - batch->map) &&
           relocs <= (batch->max_relocs - batch->relocs);
 }
 
 static INLINE size_t
-intel_batchbuffer_space(struct intel_batchbuffer *batch)
+i915_winsys_batchbuffer_space(struct i915_winsys_batchbuffer *batch)
 {
    return batch->size - (batch->ptr - batch->map);
 }
 
 static INLINE void
-intel_batchbuffer_dword(struct intel_batchbuffer *batch,
-                        unsigned dword)
+i915_winsys_batchbuffer_dword(struct i915_winsys_batchbuffer *batch,
+                              unsigned dword)
 {
-   if (intel_batchbuffer_space(batch) < 4)
+   if (i915_winsys_batchbuffer_space(batch) < 4)
       return;
 
    *(unsigned *)batch->ptr = dword;
@@ -57,11 +57,11 @@ intel_batchbuffer_dword(struct intel_batchbuffer *batch,
 }
 
 static INLINE void
-intel_batchbuffer_write(struct intel_batchbuffer *batch,
-                        void *data,
-                        size_t size)
+i915_winsys_batchbuffer_write(struct i915_winsys_batchbuffer *batch,
+                       void *data,
+                       size_t size)
 {
-   if (intel_batchbuffer_space(batch) < size)
+   if (i915_winsys_batchbuffer_space(batch) < size)
       return;
 
    memcpy(data, batch->ptr, size);
@@ -69,17 +69,17 @@ intel_batchbuffer_write(struct intel_batchbuffer *batch,
 }
 
 static INLINE int
-intel_batchbuffer_reloc(struct intel_batchbuffer *batch,
-                        struct intel_buffer *buffer,
-                        enum intel_buffer_usage usage,
-                        size_t offset)
+i915_winsys_batchbuffer_reloc(struct i915_winsys_batchbuffer *batch,
+                              struct i915_winsys_buffer *buffer,
+                              enum i915_winsys_buffer_usage usage,
+                              size_t offset)
 {
    return batch->iws->batchbuffer_reloc(batch, buffer, usage, offset);
 }
 
 static INLINE void
-intel_batchbuffer_flush(struct intel_batchbuffer *batch,
-                        struct pipe_fence_handle **fence)
+i915_winsys_batchbuffer_flush(struct i915_winsys_batchbuffer *batch,
+                              struct pipe_fence_handle **fence)
 {
    batch->iws->batchbuffer_flush(batch, fence);
 }
diff --git a/src/gallium/drivers/i915/i915_blit.c b/src/gallium/drivers/i915/i915_blit.c
index 83dfc335288..533fa81219b 100644
--- a/src/gallium/drivers/i915/i915_blit.c
+++ b/src/gallium/drivers/i915/i915_blit.c
@@ -37,7 +37,7 @@ void
 i915_fill_blit(struct i915_context *i915,
                unsigned cpp,
                unsigned short dst_pitch,
-               struct intel_buffer *dst_buffer,
+               struct i915_winsys_buffer *dst_buffer,
                unsigned dst_offset,
                short x, short y, 
                short w, short h, 
@@ -77,7 +77,7 @@ i915_fill_blit(struct i915_context *i915,
    OUT_BATCH(BR13);
    OUT_BATCH((y << 16) | x);
    OUT_BATCH(((y + h) << 16) | (x + w));
-   OUT_RELOC(dst_buffer, INTEL_USAGE_2D_TARGET, dst_offset);
+   OUT_RELOC(dst_buffer, I915_USAGE_2D_TARGET, dst_offset);
    OUT_BATCH(color);
    FLUSH_BATCH(NULL);
 }
@@ -87,10 +87,10 @@ i915_copy_blit(struct i915_context *i915,
                unsigned do_flip,
                unsigned cpp,
                unsigned short src_pitch,
-               struct intel_buffer *src_buffer,
+               struct i915_winsys_buffer *src_buffer,
                unsigned src_offset,
                unsigned short dst_pitch,
-               struct intel_buffer *dst_buffer,
+               struct i915_winsys_buffer *dst_buffer,
                unsigned dst_offset,
                short src_x, short src_y,
                short dst_x, short dst_y, 
@@ -143,9 +143,9 @@ i915_copy_blit(struct i915_context *i915,
    OUT_BATCH(BR13);
    OUT_BATCH((dst_y << 16) | dst_x);
    OUT_BATCH((dst_y2 << 16) | dst_x2);
-   OUT_RELOC(dst_buffer, INTEL_USAGE_2D_TARGET, dst_offset);
+   OUT_RELOC(dst_buffer, I915_USAGE_2D_TARGET, dst_offset);
    OUT_BATCH((src_y << 16) | src_x);
    OUT_BATCH(((int) src_pitch & 0xffff));
-   OUT_RELOC(src_buffer, INTEL_USAGE_2D_SOURCE, src_offset);
+   OUT_RELOC(src_buffer, I915_USAGE_2D_SOURCE, src_offset);
    FLUSH_BATCH(NULL);
 }
diff --git a/src/gallium/drivers/i915/i915_blit.h b/src/gallium/drivers/i915/i915_blit.h
index 8ce3220cfd9..db576ed4c90 100644
--- a/src/gallium/drivers/i915/i915_blit.h
+++ b/src/gallium/drivers/i915/i915_blit.h
@@ -34,10 +34,10 @@ extern void i915_copy_blit(struct i915_context *i915,
                            unsigned do_flip,
                            unsigned cpp,
                            unsigned short src_pitch,
-                           struct intel_buffer *src_buffer,
+                           struct i915_winsys_buffer *src_buffer,
                            unsigned src_offset,
                            unsigned short dst_pitch,
-                           struct intel_buffer *dst_buffer,
+                           struct i915_winsys_buffer *dst_buffer,
                            unsigned dst_offset,
                            short srcx, short srcy,
                            short dstx, short dsty,
@@ -46,7 +46,7 @@ extern void i915_copy_blit(struct i915_context *i915,
 extern void i915_fill_blit(struct i915_context *i915,
                            unsigned cpp,
                            unsigned short dst_pitch,
-                           struct intel_buffer *dst_buffer,
+                           struct i915_winsys_buffer *dst_buffer,
                            unsigned dst_offset,
                            short x, short y,
                            short w, short h, unsigned color);
diff --git a/src/gallium/drivers/i915/i915_buffer.c b/src/gallium/drivers/i915/i915_buffer.c
index 0f76a59e93a..1247de320d4 100644
--- a/src/gallium/drivers/i915/i915_buffer.c
+++ b/src/gallium/drivers/i915/i915_buffer.c
@@ -28,13 +28,13 @@
 #include "i915_screen.h"
 #include "i915_buffer.h"
 
-struct intel_buffer;
+struct i915_winsys_buffer;
 
 struct i915_buffer
 {
    struct pipe_buffer base;
 
-   struct intel_buffer *ibuf; /** hw buffer */
+   struct i915_winsys_buffer *ibuf; /** hw buffer */
 
    void *data; /**< user and malloc data */
    boolean own; /**< we own the data incase of malloc */
diff --git a/src/gallium/drivers/i915/i915_context.h b/src/gallium/drivers/i915/i915_context.h
index 039165b63f5..8df4dce8653 100644
--- a/src/gallium/drivers/i915/i915_context.h
+++ b/src/gallium/drivers/i915/i915_context.h
@@ -38,9 +38,9 @@
 #include "tgsi/tgsi_scan.h"
 
 
-struct intel_winsys;
-struct intel_buffer;
-struct intel_batchbuffer;
+struct i915_winsys;
+struct i915_winsys_buffer;
+struct i915_winsys_batchbuffer;
 
 
 #define I915_TEX_UNITS 8
@@ -219,14 +219,14 @@ struct i915_texture {
 
    /* The data is held here:
     */
-   struct intel_buffer *buffer;
+   struct i915_winsys_buffer *buffer;
 };
 
 struct i915_context
 {
    struct pipe_context base;
 
-   struct intel_winsys *iws;
+   struct i915_winsys *iws;
 
    struct draw_context *draw;
 
@@ -247,20 +247,20 @@ struct i915_context
    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_sampler_view *fragment_sampler_views[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_fragment_sampler_views;
    unsigned num_vertex_buffers;
 
-   struct intel_batchbuffer *batch;
+   struct i915_winsys_batchbuffer *batch;
 
    /** Vertex buffer */
-   struct intel_buffer *vbo;
+   struct i915_winsys_buffer *vbo;
    size_t vbo_offset;
    unsigned vbo_flushed;
 
@@ -283,7 +283,7 @@ struct i915_context
 #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_SAMPLER_VIEW  0x800
 #define I915_NEW_CONSTANTS     0x1000
 #define I915_NEW_VBO           0x2000
 #define I915_NEW_VS            0x4000
diff --git a/src/gallium/drivers/i915/i915_debug.c b/src/gallium/drivers/i915/i915_debug.c
index 237654d26b2..663fac3055c 100644
--- a/src/gallium/drivers/i915/i915_debug.c
+++ b/src/gallium/drivers/i915/i915_debug.c
@@ -863,7 +863,7 @@ static boolean i915_debug_packet( struct debug_stream *stream )
 
 
 void
-i915_dump_batchbuffer( struct intel_batchbuffer *batch )
+i915_dump_batchbuffer( struct i915_winsys_batchbuffer *batch )
 {
    struct debug_stream stream;
    unsigned *start = (unsigned*)batch->map;
diff --git a/src/gallium/drivers/i915/i915_debug.h b/src/gallium/drivers/i915/i915_debug.h
index 8f7484797de..67b8d9c2f63 100644
--- a/src/gallium/drivers/i915/i915_debug.h
+++ b/src/gallium/drivers/i915/i915_debug.h
@@ -104,9 +104,9 @@ I915_DBG(
 #endif
 
 
-struct intel_batchbuffer;
+struct i915_winsys_batchbuffer;
 
-void i915_dump_batchbuffer( struct intel_batchbuffer *i915 );
+void i915_dump_batchbuffer( struct i915_winsys_batchbuffer *i915 );
 
 void i915_debug_init( struct i915_context *i915 );
 
diff --git a/src/gallium/drivers/i915/i915_prim_emit.c b/src/gallium/drivers/i915/i915_prim_emit.c
index d9a5c40ab97..dd997e2cf48 100644
--- a/src/gallium/drivers/i915/i915_prim_emit.c
+++ b/src/gallium/drivers/i915/i915_prim_emit.c
@@ -102,6 +102,13 @@ emit_hw_vertex( struct i915_context *i915,
          count += 4;
          break;
       case EMIT_4UB:
+         OUT_BATCH( pack_ub4(float_to_ubyte( attrib[0] ),
+                             float_to_ubyte( attrib[1] ),
+                             float_to_ubyte( attrib[2] ),
+                             float_to_ubyte( attrib[3] )) );
+         count += 1;
+         break;
+      case EMIT_4UB_BGRA:
          OUT_BATCH( pack_ub4(float_to_ubyte( attrib[2] ),
                              float_to_ubyte( attrib[1] ),
                              float_to_ubyte( attrib[0] ),
diff --git a/src/gallium/drivers/i915/i915_prim_vbuf.c b/src/gallium/drivers/i915/i915_prim_vbuf.c
index cad4109ee6b..df9e68af4fc 100644
--- a/src/gallium/drivers/i915/i915_prim_vbuf.c
+++ b/src/gallium/drivers/i915/i915_prim_vbuf.c
@@ -76,7 +76,7 @@ struct i915_vbuf_render {
    unsigned fallback;
 
    /* Stuff for the vbo */
-   struct intel_buffer *vbo;
+   struct i915_winsys_buffer *vbo;
    size_t vbo_size; /**< current size of allocated buffer */
    size_t vbo_alloc_size; /**< minimum buffer size to allocate */
    size_t vbo_offset;
@@ -141,7 +141,7 @@ static void
 i915_vbuf_render_new_buf(struct i915_vbuf_render *i915_render, size_t size)
 {
    struct i915_context *i915 = i915_render->i915;
-   struct intel_winsys *iws = i915->iws;
+   struct i915_winsys *iws = i915->iws;
 
    if (i915_render->vbo) {
 #ifdef VBUF_USE_FIFO
@@ -172,7 +172,7 @@ i915_vbuf_render_new_buf(struct i915_vbuf_render *i915_render, size_t size)
    if (i915_render->vbo_size != i915_render->pool_buffer_size) {
       i915_render->pool_not_used = TRUE;
       i915_render->vbo = iws->buffer_create(iws, i915_render->vbo_size, 64,
-            INTEL_NEW_VERTEX);
+            I915_NEW_VERTEX);
    } else {
       i915_render->pool_not_used = FALSE;
 
@@ -185,7 +185,7 @@ i915_vbuf_render_new_buf(struct i915_vbuf_render *i915_render, size_t size)
    }
 #else
    i915_render->vbo = iws->buffer_create(iws, i915_render->vbo_size,
-                                         64, INTEL_NEW_VERTEX);
+                                         64, I915_NEW_VERTEX);
 #endif
 }
 
@@ -225,7 +225,7 @@ i915_vbuf_render_map_vertices(struct vbuf_render *render)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
-   struct intel_winsys *iws = i915->iws;
+   struct i915_winsys *iws = i915->iws;
 
    if (i915->vbo_flushed)
       debug_printf("%s bad vbo flush occured stalling on hw\n", __FUNCTION__);
@@ -246,7 +246,7 @@ i915_vbuf_render_unmap_vertices(struct vbuf_render *render,
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
-   struct intel_winsys *iws = i915->iws;
+   struct i915_winsys *iws = i915->iws;
 
    i915_render->vbo_max_used = MAX2(i915_render->vbo_max_used, i915_render->vertex_size * (max_index + 1));
 #ifdef VBUF_MAP_BUFFER
@@ -621,7 +621,7 @@ static struct vbuf_render *
 i915_vbuf_render_create(struct i915_context *i915)
 {
    struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render);
-   struct intel_winsys *iws = i915->iws;
+   struct i915_winsys *iws = i915->iws;
    int i;
 
    i915_render->i915 = i915;
@@ -662,7 +662,7 @@ i915_vbuf_render_create(struct i915_context *i915)
    for (i = 0; i < 6; i++)
       u_fifo_add(i915_render->pool_fifo,
                  iws->buffer_create(iws, i915_render->pool_buffer_size, 64,
-                                    INTEL_NEW_VERTEX));
+                                    I915_NEW_VERTEX));
 #else
    (void)i;
    (void)iws;
diff --git a/src/gallium/drivers/i915/i915_screen.c b/src/gallium/drivers/i915/i915_screen.c
index e5bf4a20bd0..53ef6e50f6c 100644
--- a/src/gallium/drivers/i915/i915_screen.c
+++ b/src/gallium/drivers/i915/i915_screen.c
@@ -35,7 +35,7 @@
 #include "i915_screen.h"
 #include "i915_buffer.h"
 #include "i915_texture.h"
-#include "intel_winsys.h"
+#include "i915_winsys.h"
 
 
 /*
@@ -165,8 +165,12 @@ i915_is_format_supported(struct pipe_screen *screen,
                          unsigned geom_flags)
 {
    static const enum pipe_format tex_supported[] = {
-      PIPE_FORMAT_A8B8G8R8_UNORM,
       PIPE_FORMAT_B8G8R8A8_UNORM,
+      PIPE_FORMAT_B8G8R8X8_UNORM,
+      PIPE_FORMAT_R8G8B8A8_UNORM,
+#if 0
+      PIPE_FORMAT_R8G8B8X8_UNORM,
+#endif
       PIPE_FORMAT_B5G6R5_UNORM,
       PIPE_FORMAT_L8_UNORM,
       PIPE_FORMAT_A8_UNORM,
@@ -256,7 +260,7 @@ i915_destroy_screen(struct pipe_screen *screen)
  * Create a new i915_screen object
  */
 struct pipe_screen *
-i915_create_screen(struct intel_winsys *iws, uint pci_id)
+i915_create_screen(struct i915_winsys *iws, uint pci_id)
 {
    struct i915_screen *is = CALLOC_STRUCT(i915_screen);
 
diff --git a/src/gallium/drivers/i915/i915_screen.h b/src/gallium/drivers/i915/i915_screen.h
index 5126485caa7..7f9e02fc0f6 100644
--- a/src/gallium/drivers/i915/i915_screen.h
+++ b/src/gallium/drivers/i915/i915_screen.h
@@ -32,7 +32,7 @@
 #include "pipe/p_screen.h"
 
 
-struct intel_winsys;
+struct i915_winsys;
 
 
 /**
@@ -42,7 +42,7 @@ struct i915_screen
 {
    struct pipe_screen base;
 
-   struct intel_winsys *iws;
+   struct i915_winsys *iws;
 
    boolean is_i945;
    uint pci_id;
diff --git a/src/gallium/drivers/i915/i915_state.c b/src/gallium/drivers/i915/i915_state.c
index 377d8425a5c..0f7395246cc 100644
--- a/src/gallium/drivers/i915/i915_state.c
+++ b/src/gallium/drivers/i915/i915_state.c
@@ -560,9 +560,9 @@ static void i915_set_constant_buffer(struct pipe_context *pipe,
 }
 
 
-static void i915_set_sampler_textures(struct pipe_context *pipe,
-                                      unsigned num,
-                                      struct pipe_texture **texture)
+static void i915_set_fragment_sampler_views(struct pipe_context *pipe,
+                                            unsigned num,
+                                            struct pipe_sampler_view **views)
 {
    struct i915_context *i915 = i915_context(pipe);
    uint i;
@@ -570,27 +570,54 @@ static void i915_set_sampler_textures(struct pipe_context *pipe,
    assert(num <= PIPE_MAX_SAMPLERS);
 
    /* Check for no-op */
-   if (num == i915->num_textures &&
-       !memcmp(i915->texture, texture, num * sizeof(struct pipe_texture *)))
+   if (num == i915->num_fragment_sampler_views &&
+       !memcmp(i915->fragment_sampler_views, views, num * sizeof(struct pipe_sampler_view *)))
       return;
 
    /* Fixes wrong texture in texobj with VBUF */
    draw_flush(i915->draw);
 
    for (i = 0; i < num; i++)
-      pipe_texture_reference((struct pipe_texture **) &i915->texture[i],
-                             texture[i]);
+      pipe_sampler_view_reference(&i915->fragment_sampler_views[i],
+                                  views[i]);
 
-   for (i = num; i < i915->num_textures; i++)
-      pipe_texture_reference((struct pipe_texture **) &i915->texture[i],
-                             NULL);
+   for (i = num; i < i915->num_fragment_sampler_views; i++)
+      pipe_sampler_view_reference(&i915->fragment_sampler_views[i],
+                                  NULL);
 
-   i915->num_textures = num;
+   i915->num_fragment_sampler_views = num;
 
-   i915->dirty |= I915_NEW_TEXTURE;
+   i915->dirty |= I915_NEW_SAMPLER_VIEW;
 }
 
 
+static struct pipe_sampler_view *
+i915_create_sampler_view(struct pipe_context *pipe,
+                         struct pipe_texture *texture,
+                         const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+
+static void
+i915_sampler_view_destroy(struct pipe_context *pipe,
+                          struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
+}
+
 
 static void i915_set_framebuffer_state(struct pipe_context *pipe,
 				       const struct pipe_framebuffer_state *fb)
@@ -818,7 +845,9 @@ i915_init_state_functions( struct i915_context *i915 )
 
    i915->base.set_polygon_stipple = i915_set_polygon_stipple;
    i915->base.set_scissor_state = i915_set_scissor_state;
-   i915->base.set_fragment_sampler_textures = i915_set_sampler_textures;
+   i915->base.set_fragment_sampler_views = i915_set_fragment_sampler_views;
+   i915->base.create_sampler_view = i915_create_sampler_view;
+   i915->base.sampler_view_destroy = i915_sampler_view_destroy;
    i915->base.set_viewport_state = i915_set_viewport_state;
    i915->base.set_vertex_buffers = i915_set_vertex_buffers;
 }
diff --git a/src/gallium/drivers/i915/i915_state_derived.c b/src/gallium/drivers/i915/i915_state_derived.c
index f5b0e9f011e..4da46772b5d 100644
--- a/src/gallium/drivers/i915/i915_state_derived.c
+++ b/src/gallium/drivers/i915/i915_state_derived.c
@@ -101,14 +101,14 @@ static void calculate_vertex_layout( struct i915_context *i915 )
    /* primary color */
    if (colors[0]) {
       src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_COLOR, 0);
-      draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src);
+      draw_emit_vertex_attr(&vinfo, EMIT_4UB_BGRA, colorInterp, src);
       vinfo.hwfmt[0] |= S4_VFMT_COLOR;
    }
 
    /* secondary color */
    if (colors[1]) {
       src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_COLOR, 1);
-      draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src);
+      draw_emit_vertex_attr(&vinfo, EMIT_4UB_BGRA, colorInterp, src);
       vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG;
    }
 
@@ -157,10 +157,10 @@ void i915_update_derived( struct i915_context *i915 )
    if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS | I915_NEW_VS))
       calculate_vertex_layout( i915 );
 
-   if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE))
+   if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_SAMPLER_VIEW))
       i915_update_samplers(i915);
 
-   if (i915->dirty & I915_NEW_TEXTURE)
+   if (i915->dirty & I915_NEW_SAMPLER_VIEW)
       i915_update_textures(i915);
 
    if (i915->dirty)
diff --git a/src/gallium/drivers/i915/i915_state_emit.c b/src/gallium/drivers/i915/i915_state_emit.c
index 51f0ef12baf..04a3924aaf7 100644
--- a/src/gallium/drivers/i915/i915_state_emit.c
+++ b/src/gallium/drivers/i915/i915_state_emit.c
@@ -182,7 +182,7 @@ i915_emit_hardware_state(struct i915_context *i915 )
       
       if(i915->vbo)
          OUT_RELOC(i915->vbo,
-                   INTEL_USAGE_VERTEX,
+                   I915_USAGE_VERTEX,
                    i915->current.immediate[I915_IMMEDIATE_S0]);
       else
          /* FIXME: we should not do this */
@@ -226,7 +226,7 @@ i915_emit_hardware_state(struct i915_context *i915 )
                    ctile);
 
          OUT_RELOC(tex->buffer,
-                   INTEL_USAGE_RENDER,
+                   I915_USAGE_RENDER,
                    cbuf_surface->offset);
       }
 
@@ -250,7 +250,7 @@ i915_emit_hardware_state(struct i915_context *i915 )
                    ztile);
 
          OUT_RELOC(tex->buffer,
-                   INTEL_USAGE_RENDER,
+                   I915_USAGE_RENDER,
                    depth_surface->offset);
       }
    
@@ -290,13 +290,14 @@ i915_emit_hardware_state(struct i915_context *i915 )
             OUT_BATCH(enabled);
             for (unit = 0; unit < I915_TEX_UNITS; unit++) {
                if (enabled & (1 << unit)) {
-                  struct intel_buffer *buf = i915->texture[unit]->buffer;
+                  struct i915_texture *texture = (struct i915_texture *)i915->fragment_sampler_views[unit]->texture;
+                  struct i915_winsys_buffer *buf = texture->buffer;
                   uint offset = 0;
                   assert(buf);
 
                   count++;
 
-                  OUT_RELOC(buf, INTEL_USAGE_SAMPLER, offset);
+                  OUT_RELOC(buf, I915_USAGE_SAMPLER, offset);
                   OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */
                   OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */
                }
diff --git a/src/gallium/drivers/i915/i915_state_immediate.c b/src/gallium/drivers/i915/i915_state_immediate.c
index d2c6f151434..8cec699285c 100644
--- a/src/gallium/drivers/i915/i915_state_immediate.c
+++ b/src/gallium/drivers/i915/i915_state_immediate.c
@@ -52,11 +52,11 @@ static void upload_S0S1(struct i915_context *i915)
 {
    unsigned LIS0, LIS1;
 
-   /* INTEL_NEW_VBO */
+   /* I915_NEW_VBO */
    /* TODO: re-use vertex buffers here? */
    LIS0 = i915->vbo_offset;
 
-   /* INTEL_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! 
+   /* I915_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! 
     */
    {
       unsigned vertex_size = i915->current.vertex_info.size;
@@ -65,7 +65,7 @@ static void upload_S0S1(struct i915_context *i915)
 	      (vertex_size << 16));
    }
 
-   /* INTEL_NEW_VBO */
+   /* I915_NEW_VBO */
    /* TODO: use a vertex generation number to track vbo changes */
    if (1 ||
        i915->current.immediate[I915_IMMEDIATE_S0] != LIS0 ||
diff --git a/src/gallium/drivers/i915/i915_state_sampler.c b/src/gallium/drivers/i915/i915_state_sampler.c
index 9813290b51b..4c326561cb9 100644
--- a/src/gallium/drivers/i915/i915_state_sampler.c
+++ b/src/gallium/drivers/i915/i915_state_sampler.c
@@ -144,20 +144,22 @@ void i915_update_samplers( struct i915_context *i915 )
    i915->current.sampler_enable_nr = 0;
    i915->current.sampler_enable_flags = 0x0;
 
-   for (unit = 0; unit < i915->num_textures && unit < i915->num_samplers;
+   for (unit = 0; unit < i915->num_fragment_sampler_views && unit < i915->num_samplers;
         unit++) {
       /* determine unit enable/disable by looking for a bound texture */
       /* could also examine the fragment program? */
-      if (i915->texture[unit]) {
+      if (i915->fragment_sampler_views[unit]) {
+         struct i915_texture *texture = (struct i915_texture *)i915->fragment_sampler_views[unit]->texture;
+
 	 update_sampler( i915,
 	                 unit,
 	                 i915->sampler[unit],       /* sampler state */
-	                 i915->texture[unit],        /* texture */
+	                 texture,                    /* texture */
 	                 i915->current.sampler[unit] /* the result */
 	                 );
 	 i915_update_texture( i915,
 	                      unit,
-	                      i915->texture[unit],          /* texture */
+	                      texture,                      /* texture */
 	                      i915->sampler[unit],          /* sampler state */
 	                      i915->current.texbuffer[unit] );
 
@@ -190,6 +192,14 @@ translate_texture_format(enum pipe_format pipeFormat)
       return MAPSURF_16BIT | MT_16BIT_ARGB4444;
    case PIPE_FORMAT_B8G8R8A8_UNORM:
       return MAPSURF_32BIT | MT_32BIT_ARGB8888;
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+      return MAPSURF_32BIT | MT_32BIT_XRGB8888;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return MAPSURF_32BIT | MT_32BIT_ABGR8888;
+#if 0
+   case PIPE_FORMAT_R8G8B8X8_UNORM:
+      return MAPSURF_32BIT | MT_32BIT_XBGR8888;
+#endif
    case PIPE_FORMAT_YUYV:
       return (MAPSURF_422 | MT_422_YCRCB_NORMAL);
    case PIPE_FORMAT_UYVY:
@@ -281,14 +291,16 @@ i915_update_textures(struct i915_context *i915)
 {
    uint unit;
 
-   for (unit = 0; unit < i915->num_textures && unit < i915->num_samplers;
+   for (unit = 0; unit < i915->num_fragment_sampler_views && unit < i915->num_samplers;
         unit++) {
       /* determine unit enable/disable by looking for a bound texture */
       /* could also examine the fragment program? */
-      if (i915->texture[unit]) {
+      if (i915->fragment_sampler_views[unit]) {
+         struct i915_texture *texture = (struct i915_texture *)i915->fragment_sampler_views[unit]->texture;
+
 	 i915_update_texture( i915,
 	                      unit,
-	                      i915->texture[unit],          /* texture */
+	                      texture,                      /* texture */
 	                      i915->sampler[unit],          /* sampler state */
 	                      i915->current.texbuffer[unit] );
       }
diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c
index b252fb5330c..8c405c2443e 100644
--- a/src/gallium/drivers/i915/i915_texture.c
+++ b/src/gallium/drivers/i915/i915_texture.c
@@ -41,7 +41,7 @@
 #include "i915_context.h"
 #include "i915_texture.h"
 #include "i915_screen.h"
-#include "intel_winsys.h"
+#include "i915_winsys.h"
 
 
 /*
@@ -162,7 +162,7 @@ i915_scanout_layout(struct i915_texture *tex)
    if (pt->width0 >= 240) {
       tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0));
       tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8);
-      tex->hw_tiled = INTEL_TILE_X;
+      tex->hw_tiled = I915_TILE_X;
    } else if (pt->width0 == 64 && pt->height0 == 64) {
       tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0));
       tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8);
@@ -200,7 +200,7 @@ i915_display_target_layout(struct i915_texture *tex)
 
    tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0));
    tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8);
-   tex->hw_tiled = INTEL_TILE_X;
+   tex->hw_tiled = I915_TILE_X;
 
    debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
       pt->width0, pt->height0, util_format_get_blocksize(pt->format),
@@ -617,7 +617,7 @@ i915_texture_create(struct pipe_screen *screen,
                     const struct pipe_texture *templat)
 {
    struct i915_screen *is = i915_screen(screen);
-   struct intel_winsys *iws = is->iws;
+   struct i915_winsys *iws = is->iws;
    struct i915_texture *tex = CALLOC_STRUCT(i915_texture);
    size_t tex_size;
    unsigned buf_usage = 0;
@@ -643,9 +643,9 @@ i915_texture_create(struct pipe_screen *screen,
 
    /* for scanouts and cursors, cursors arn't scanouts */
    if (templat->tex_usage & PIPE_TEXTURE_USAGE_SCANOUT && templat->width0 != 64)
-      buf_usage = INTEL_NEW_SCANOUT;
+      buf_usage = I915_NEW_SCANOUT;
    else
-      buf_usage = INTEL_NEW_TEXTURE;
+      buf_usage = I915_NEW_TEXTURE;
 
    tex->buffer = iws->buffer_create(iws, tex_size, 64, buf_usage);
    if (!tex->buffer)
@@ -653,7 +653,7 @@ i915_texture_create(struct pipe_screen *screen,
 
    /* setup any hw fences */
    if (tex->hw_tiled) {
-      assert(tex->sw_tiled == INTEL_TILE_NONE);
+      assert(tex->sw_tiled == I915_TILE_NONE);
       iws->buffer_set_fence_reg(iws, tex->buffer, tex->stride, tex->hw_tiled);
    }
 
@@ -679,8 +679,8 @@ i915_texture_from_handle(struct pipe_screen * screen,
 {
    struct i915_screen *is = i915_screen(screen);
    struct i915_texture *tex;
-   struct intel_winsys *iws = is->iws;
-   struct intel_buffer *buffer;
+   struct i915_winsys *iws = is->iws;
+   struct i915_winsys_buffer *buffer;
    unsigned stride;
 
    assert(screen);
@@ -719,7 +719,7 @@ i915_texture_get_handle(struct pipe_screen * screen,
 {
    struct i915_screen *is = i915_screen(screen);
    struct i915_texture *tex = (struct i915_texture *)texture;
-   struct intel_winsys *iws = is->iws;
+   struct i915_winsys *iws = is->iws;
 
    return iws->buffer_get_handle(iws, tex->buffer, whandle, tex->stride);
 }
@@ -729,7 +729,7 @@ static void
 i915_texture_destroy(struct pipe_texture *pt)
 {
    struct i915_texture *tex = (struct i915_texture *)pt;
-   struct intel_winsys *iws = i915_screen(pt->screen)->iws;
+   struct i915_winsys *iws = i915_screen(pt->screen)->iws;
    uint i;
 
    /*
@@ -841,7 +841,7 @@ i915_transfer_map(struct pipe_context *pipe,
                   struct pipe_transfer *transfer)
 {
    struct i915_texture *tex = (struct i915_texture *)transfer->texture;
-   struct intel_winsys *iws = i915_screen(tex->base.screen)->iws;
+   struct i915_winsys *iws = i915_screen(tex->base.screen)->iws;
    char *map;
    boolean write = FALSE;
    enum pipe_format format = tex->base.format;
@@ -863,7 +863,7 @@ i915_transfer_unmap(struct pipe_context *pipe,
                     struct pipe_transfer *transfer)
 {
    struct i915_texture *tex = (struct i915_texture *)transfer->texture;
-   struct intel_winsys *iws = i915_screen(tex->base.screen)->iws;
+   struct i915_winsys *iws = i915_screen(tex->base.screen)->iws;
    iws->buffer_unmap(iws, tex->buffer);
 }
 
diff --git a/src/gallium/drivers/i915/intel_winsys.h b/src/gallium/drivers/i915/i915_winsys.h
index 00fd0c1efea..246e95b00d6 100644
--- a/src/gallium/drivers/i915/intel_winsys.h
+++ b/src/gallium/drivers/i915/i915_winsys.h
@@ -23,46 +23,46 @@
  *
  **************************************************************************/
 
-#ifndef INTEL_WINSYS_H
-#define INTEL_WINSYS_H
+#ifndef I915_WINSYS_H
+#define I915_WINSYS_H
 
 #include "pipe/p_compiler.h"
 
-struct intel_winsys;
-struct intel_buffer;
-struct intel_batchbuffer;
+struct i915_winsys;
+struct i915_winsys_buffer;
+struct i915_winsys_batchbuffer;
 struct pipe_texture;
 struct pipe_fence_handle;
 struct winsys_handle;
 
-enum intel_buffer_usage
+enum i915_winsys_buffer_usage
 {
    /* use on textures */
-   INTEL_USAGE_RENDER    = 0x01,
-   INTEL_USAGE_SAMPLER   = 0x02,
-   INTEL_USAGE_2D_TARGET = 0x04,
-   INTEL_USAGE_2D_SOURCE = 0x08,
+   I915_USAGE_RENDER    = 0x01,
+   I915_USAGE_SAMPLER   = 0x02,
+   I915_USAGE_2D_TARGET = 0x04,
+   I915_USAGE_2D_SOURCE = 0x08,
    /* use on vertex */
-   INTEL_USAGE_VERTEX    = 0x10
+   I915_USAGE_VERTEX    = 0x10
 };
 
-enum intel_buffer_type
+enum i915_winsys_buffer_type
 {
-   INTEL_NEW_TEXTURE,
-   INTEL_NEW_SCANOUT, /**< a texture used for scanning out from */
-   INTEL_NEW_VERTEX
+   I915_NEW_TEXTURE,
+   I915_NEW_SCANOUT, /**< a texture used for scanning out from */
+   I915_NEW_VERTEX
 };
 
-enum intel_buffer_tile
+enum i915_winsys_buffer_tile
 {
-   INTEL_TILE_NONE,
-   INTEL_TILE_X,
-   INTEL_TILE_Y
+   I915_TILE_NONE,
+   I915_TILE_X,
+   I915_TILE_Y
 };
 
-struct intel_batchbuffer {
+struct i915_winsys_batchbuffer {
 
-   struct intel_winsys *iws;
+   struct i915_winsys *iws;
 
    /**
     * Values exported to speed up the writing the batchbuffer,
@@ -79,7 +79,7 @@ struct intel_batchbuffer {
    /*@}*/
 };
 
-struct intel_winsys {
+struct i915_winsys {
 
    /**
     * Batchbuffer functions.
@@ -88,7 +88,8 @@ struct intel_winsys {
    /**
     * Create a new batchbuffer.
     */
-   struct intel_batchbuffer *(*batchbuffer_create)(struct intel_winsys *iws);
+   struct i915_winsys_batchbuffer *
+      (*batchbuffer_create)(struct i915_winsys *iws);
 
    /**
     * Emit a relocation to a buffer.
@@ -100,21 +101,21 @@ struct intel_winsys {
     * @offset add this to the reloc buffers address
     * @target buffer where to write the address, null for batchbuffer.
     */
-   int (*batchbuffer_reloc)(struct intel_batchbuffer *batch,
-                            struct intel_buffer *reloc,
-                            enum intel_buffer_usage usage,
+   int (*batchbuffer_reloc)(struct i915_winsys_batchbuffer *batch,
+                            struct i915_winsys_buffer *reloc,
+                            enum i915_winsys_buffer_usage usage,
                             unsigned offset);
 
    /**
     * Flush a bufferbatch.
     */
-   void (*batchbuffer_flush)(struct intel_batchbuffer *batch,
+   void (*batchbuffer_flush)(struct i915_winsys_batchbuffer *batch,
                              struct pipe_fence_handle **fence);
 
    /**
     * Destroy a batchbuffer.
     */
-   void (*batchbuffer_destroy)(struct intel_batchbuffer *batch);
+   void (*batchbuffer_destroy)(struct i915_winsys_batchbuffer *batch);
    /*@}*/
 
 
@@ -125,9 +126,10 @@ struct intel_winsys {
    /**
     * Create a buffer.
     */
-   struct intel_buffer *(*buffer_create)(struct intel_winsys *iws,
-                                         unsigned size, unsigned alignment,
-                                         enum intel_buffer_type type);
+   struct i915_winsys_buffer *
+      (*buffer_create)(struct i915_winsys *iws,
+                       unsigned size, unsigned alignment,
+                       enum i915_winsys_buffer_type type);
 
    /**
     * Creates a buffer from a handle.
@@ -135,16 +137,17 @@ struct intel_winsys {
     * Also provides the stride information needed for the
     * texture via the stride argument.
     */
-   struct intel_buffer *(*buffer_from_handle)(struct intel_winsys *iws,
-                                              struct winsys_handle *whandle,
-                                              unsigned *stride);
+   struct i915_winsys_buffer *
+      (*buffer_from_handle)(struct i915_winsys *iws,
+                            struct winsys_handle *whandle,
+                            unsigned *stride);
 
    /**
     * Used to implement pipe_screen::texture_get_handle.
     * The winsys might need the stride information.
     */
-   boolean (*buffer_get_handle)(struct intel_winsys *iws,
-                                struct intel_buffer *buffer,
+   boolean (*buffer_get_handle)(struct i915_winsys *iws,
+                                struct i915_winsys_buffer *buffer,
                                 struct winsys_handle *whandle,
                                 unsigned stride);
 
@@ -152,37 +155,37 @@ struct intel_winsys {
     * Fence a buffer with a fence reg.
     * Not to be confused with pipe_fence_handle.
     */
-   int (*buffer_set_fence_reg)(struct intel_winsys *iws,
-                               struct intel_buffer *buffer,
+   int (*buffer_set_fence_reg)(struct i915_winsys *iws,
+                               struct i915_winsys_buffer *buffer,
                                unsigned stride,
-                               enum intel_buffer_tile tile);
+                               enum i915_winsys_buffer_tile tile);
 
    /**
     * Map a buffer.
     */
-   void *(*buffer_map)(struct intel_winsys *iws,
-                       struct intel_buffer *buffer,
+   void *(*buffer_map)(struct i915_winsys *iws,
+                       struct i915_winsys_buffer *buffer,
                        boolean write);
 
    /**
     * Unmap a buffer.
     */
-   void (*buffer_unmap)(struct intel_winsys *iws,
-                        struct intel_buffer *buffer);
+   void (*buffer_unmap)(struct i915_winsys *iws,
+                        struct i915_winsys_buffer *buffer);
 
    /**
     * Write to a buffer.
     *
     * Arguments follows pipe_buffer_write.
     */
-   int (*buffer_write)(struct intel_winsys *iws,
-                       struct intel_buffer *dst,
+   int (*buffer_write)(struct i915_winsys *iws,
+                       struct i915_winsys_buffer *dst,
                        size_t offset,
                        size_t size,
                        const void *data);
 
-   void (*buffer_destroy)(struct intel_winsys *iws,
-                          struct intel_buffer *buffer);
+   void (*buffer_destroy)(struct i915_winsys *iws,
+                          struct i915_winsys_buffer *buffer);
    /*@}*/
 
 
@@ -193,20 +196,20 @@ struct intel_winsys {
    /**
     * Reference fence and set ptr to fence.
     */
-   void (*fence_reference)(struct intel_winsys *iws,
+   void (*fence_reference)(struct i915_winsys *iws,
                            struct pipe_fence_handle **ptr,
                            struct pipe_fence_handle *fence);
 
    /**
     * Check if a fence has finished.
     */
-   int (*fence_signalled)(struct intel_winsys *iws,
+   int (*fence_signalled)(struct i915_winsys *iws,
                           struct pipe_fence_handle *fence);
 
    /**
     * Wait on a fence to finish.
     */
-   int (*fence_finish)(struct intel_winsys *iws,
+   int (*fence_finish)(struct i915_winsys *iws,
                        struct pipe_fence_handle *fence);
    /*@}*/
 
@@ -214,14 +217,14 @@ struct intel_winsys {
    /**
     * Destroy the winsys.
     */
-   void (*destroy)(struct intel_winsys *iws);
+   void (*destroy)(struct i915_winsys *iws);
 };
 
 
 /**
  * Create i915 pipe_screen.
  */
-struct pipe_screen *i915_create_screen(struct intel_winsys *iws, unsigned pci_id);
+struct pipe_screen *i915_create_screen(struct i915_winsys *iws, unsigned pci_id);
 
 
 #endif
diff --git a/src/gallium/drivers/i965/SConscript b/src/gallium/drivers/i965/SConscript
index 9c2faaf4b49..d900ea25968 100644
--- a/src/gallium/drivers/i965/SConscript
+++ b/src/gallium/drivers/i965/SConscript
@@ -58,6 +58,7 @@ i965 = env.ConvenienceLibrary(
 		'brw_vs_emit.c',
 		'brw_vs_state.c',
 		'brw_vs_surface_state.c',
+		'brw_winsys_debug.c',
 		'brw_wm.c',
 #		'brw_wm_constant_buffer.c',
 		'brw_wm_debug.c',
diff --git a/src/gallium/drivers/i965/brw_context.c b/src/gallium/drivers/i965/brw_context.c
index 3dbe2b91308..4bcdcdd17eb 100644
--- a/src/gallium/drivers/i965/brw_context.c
+++ b/src/gallium/drivers/i965/brw_context.c
@@ -118,6 +118,7 @@ struct pipe_context *brw_create_context(struct pipe_screen *screen,
    brw->sws = brw_screen(screen)->sws;
    brw->chipset = brw_screen(screen)->chipset;
 
+   brw_tex_init( brw );
    brw_pipe_blend_init( brw );
    brw_pipe_depth_stencil_init( brw );
    brw_pipe_framebuffer_init( brw );
diff --git a/src/gallium/drivers/i965/brw_context.h b/src/gallium/drivers/i965/brw_context.h
index f5b1a06576b..dab881fea24 100644
--- a/src/gallium/drivers/i965/brw_context.h
+++ b/src/gallium/drivers/i965/brw_context.h
@@ -551,9 +551,9 @@ struct brw_context
       const struct brw_sampler *sampler[PIPE_MAX_SAMPLERS];
       unsigned num_samplers;
 
-      struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+      struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
       struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
-      unsigned num_textures;
+      unsigned num_fragment_sampler_views;
       unsigned num_vertex_buffers;
 
       struct pipe_scissor_state scissor;
diff --git a/src/gallium/drivers/i965/brw_pipe_sampler.c b/src/gallium/drivers/i965/brw_pipe_sampler.c
index c7c0e2ae95e..d2aa2bc9f35 100644
--- a/src/gallium/drivers/i965/brw_pipe_sampler.c
+++ b/src/gallium/drivers/i965/brw_pipe_sampler.c
@@ -183,26 +183,26 @@ static void brw_delete_sampler_state(struct pipe_context *pipe,
    FREE(cso);
 }
 
-static void brw_set_sampler_textures(struct pipe_context *pipe,
-				     unsigned num,
-				     struct pipe_texture **texture)
+static void brw_set_fragment_sampler_views(struct pipe_context *pipe,
+                                           unsigned num,
+                                           struct pipe_sampler_view **views)
 {
    struct brw_context *brw = brw_context(pipe);
    int i;
 
    for (i = 0; i < num; i++)
-      pipe_texture_reference(&brw->curr.texture[i], texture[i]);
+      pipe_sampler_view_reference(&brw->curr.fragment_sampler_views[i], views[i]);
 
-   for (i = num; i < brw->curr.num_textures; i++)
-      pipe_texture_reference(&brw->curr.texture[i], NULL);
+   for (i = num; i < brw->curr.num_fragment_sampler_views; i++)
+      pipe_sampler_view_reference(&brw->curr.fragment_sampler_views[i], NULL);
 
-   brw->curr.num_textures = num;
+   brw->curr.num_fragment_sampler_views = num;
    brw->state.dirty.mesa |= PIPE_NEW_BOUND_TEXTURES;
 }
 
-static void brw_set_vertex_sampler_textures(struct pipe_context *pipe,
-                                            unsigned num,
-                                            struct pipe_texture **texture)
+static void brw_set_vertex_sampler_views(struct pipe_context *pipe,
+                                         unsigned num,
+                                         struct pipe_sampler_view **views)
 {
 }
 
@@ -212,17 +212,47 @@ static void brw_bind_vertex_sampler_state(struct pipe_context *pipe,
 }
 
 
+static struct pipe_sampler_view *
+brw_create_sampler_view(struct pipe_context *pipe,
+                        struct pipe_texture *texture,
+                        const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+
+static void
+brw_sampler_view_destroy(struct pipe_context *pipe,
+                         struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
+}
+
+
 void brw_pipe_sampler_init( struct brw_context *brw )
 {
    brw->base.create_sampler_state = brw_create_sampler_state;
    brw->base.delete_sampler_state = brw_delete_sampler_state;
 
-   brw->base.set_fragment_sampler_textures = brw_set_sampler_textures;
+   brw->base.set_fragment_sampler_views = brw_set_fragment_sampler_views;
    brw->base.bind_fragment_sampler_states = brw_bind_sampler_state;
 
-   brw->base.set_vertex_sampler_textures = brw_set_vertex_sampler_textures;
+   brw->base.set_vertex_sampler_views = brw_set_vertex_sampler_views;
    brw->base.bind_vertex_sampler_states = brw_bind_vertex_sampler_state;
 
+   brw->base.create_sampler_view = brw_create_sampler_view;
+   brw->base.sampler_view_destroy = brw_sampler_view_destroy;
 }
 void brw_pipe_sampler_cleanup( struct brw_context *brw )
 {
diff --git a/src/gallium/drivers/i965/brw_screen_texture.c b/src/gallium/drivers/i965/brw_screen_texture.c
index cadcb7cee2a..f9f17bdabac 100644
--- a/src/gallium/drivers/i965/brw_screen_texture.c
+++ b/src/gallium/drivers/i965/brw_screen_texture.c
@@ -206,7 +206,7 @@ static struct pipe_texture *brw_texture_create( struct pipe_screen *screen,
    /* XXX: compressed textures need special treatment here
     */
    tex->cpp = util_format_get_blocksize(tex->base.format);
-   tex->compressed = util_format_is_compressed(tex->base.format);
+   tex->compressed = util_format_is_s3tc(tex->base.format);
 
    make_empty_list(&tex->views[0]);
    make_empty_list(&tex->views[1]);
@@ -321,7 +321,7 @@ brw_texture_from_handle(struct pipe_screen *screen,
        templ->depth0 != 1)
       return NULL;
 
-   if (util_format_is_compressed(templ->format))
+   if (util_format_is_s3tc(templ->format))
       return NULL;
 
    tex = CALLOC_STRUCT(brw_texture);
diff --git a/src/gallium/drivers/i965/brw_wm.c b/src/gallium/drivers/i965/brw_wm.c
index dfb718e64fe..7ed2378ec04 100644
--- a/src/gallium/drivers/i965/brw_wm.c
+++ b/src/gallium/drivers/i965/brw_wm.c
@@ -251,8 +251,8 @@ static void brw_wm_populate_key( struct brw_context *brw,
 
 
    /* PIPE_NEW_BOUND_TEXTURES */
-   for (i = 0; i < brw->curr.num_textures; i++) {
-      const struct brw_texture *tex = brw_texture(brw->curr.texture[i]);
+   for (i = 0; i < brw->curr.num_fragment_sampler_views; i++) {
+      const struct brw_texture *tex = brw_texture(brw->curr.fragment_sampler_views[i]->texture);
 	 
       if (tex->base.format == PIPE_FORMAT_UYVY)
 	 key->yuvtex_mask |= 1 << i;
diff --git a/src/gallium/drivers/i965/brw_wm_sampler_state.c b/src/gallium/drivers/i965/brw_wm_sampler_state.c
index 6a6086fc51b..3f18062c584 100644
--- a/src/gallium/drivers/i965/brw_wm_sampler_state.c
+++ b/src/gallium/drivers/i965/brw_wm_sampler_state.c
@@ -78,11 +78,11 @@ brw_wm_sampler_populate_key(struct brw_context *brw,
 
    memset(key, 0, sizeof(*key));
 
-   key->sampler_count = MIN2(brw->curr.num_textures,
+   key->sampler_count = MIN2(brw->curr.num_fragment_sampler_views,
 			    brw->curr.num_samplers);
 
    for (i = 0; i < key->sampler_count; i++) {
-      const struct brw_texture *tex = brw_texture(brw->curr.texture[i]);
+      const struct brw_texture *tex = brw_texture(brw->curr.fragment_sampler_views[i]->texture);
       const struct brw_sampler *sampler = brw->curr.sampler[i];
       struct brw_sampler_state *entry = &key->sampler[i];
 
@@ -122,12 +122,12 @@ static enum pipe_error
 brw_wm_sampler_update_default_colors(struct brw_context *brw)
 {
    enum pipe_error ret;
-   int nr = MIN2(brw->curr.num_textures,
+   int nr = MIN2(brw->curr.num_fragment_sampler_views,
 		 brw->curr.num_samplers);
    int i;
 
    for (i = 0; i < nr; i++) {
-      const struct brw_texture *tex = brw_texture(brw->curr.texture[i]);
+      const struct brw_texture *tex = brw_texture(brw->curr.fragment_sampler_views[i]->texture);
       const struct brw_sampler *sampler = brw->curr.sampler[i];
       const float *bc;
       float bordercolor[4] = {
diff --git a/src/gallium/drivers/i965/brw_wm_surface_state.c b/src/gallium/drivers/i965/brw_wm_surface_state.c
index b01a7f194b7..2368ae3f808 100644
--- a/src/gallium/drivers/i965/brw_wm_surface_state.c
+++ b/src/gallium/drivers/i965/brw_wm_surface_state.c
@@ -242,9 +242,9 @@ static enum pipe_error prepare_wm_surfaces(struct brw_context *brw )
 
    /* PIPE_NEW_TEXTURE 
     */
-   for (i = 0; i < brw->curr.num_textures; i++) {
+   for (i = 0; i < brw->curr.num_fragment_sampler_views; i++) {
       ret = brw_update_texture_surface(brw, 
-                                       brw_texture(brw->curr.texture[i]),
+                                       brw_texture(brw->curr.fragment_sampler_views[i]->texture),
                                        &brw->wm.surf_bo[BTI_TEXTURE(i)]);
       if (ret)
          return ret;
@@ -261,7 +261,7 @@ static enum pipe_error prepare_wm_surfaces(struct brw_context *brw )
       bo_reference(&brw->wm.surf_bo[BTI_FRAGMENT_CONSTANTS], NULL);      
 
    /* XXX: no pipe_max_textures define?? */
-   for (i = brw->curr.num_textures; i < PIPE_MAX_SAMPLERS; i++)
+   for (i = brw->curr.num_fragment_sampler_views; i < PIPE_MAX_SAMPLERS; i++)
       bo_reference(&brw->wm.surf_bo[BTI_TEXTURE(i)], NULL);
 
    if (brw->wm.nr_surfaces != nr_surfaces) {
diff --git a/src/gallium/drivers/identity/SConscript b/src/gallium/drivers/identity/SConscript
index 7f079dd0a8b..2a68891c284 100644
--- a/src/gallium/drivers/identity/SConscript
+++ b/src/gallium/drivers/identity/SConscript
@@ -5,9 +5,10 @@ env = env.Clone()
 identity = env.ConvenienceLibrary(
 	target = 'identity',
 	source = [
-		'id_screen.c',
 		'id_context.c',
+		'id_drm.c',
 		'id_objects.c',
+		'id_screen.c',
 	])
 
 Export('identity')
diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c
index 26770d6b1e9..00a542215ad 100644
--- a/src/gallium/drivers/identity/id_context.c
+++ b/src/gallium/drivers/identity/id_context.c
@@ -528,53 +528,49 @@ identity_set_viewport_state(struct pipe_context *_pipe,
 }
 
 static void
-identity_set_fragment_sampler_textures(struct pipe_context *_pipe,
-                                       unsigned num_textures,
-                                       struct pipe_texture **_textures)
+identity_set_fragment_sampler_views(struct pipe_context *_pipe,
+                                    unsigned num,
+                                    struct pipe_sampler_view **_views)
 {
    struct identity_context *id_pipe = identity_context(_pipe);
    struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
-   struct pipe_texture **textures = NULL;
+   struct pipe_sampler_view *unwrapped_views[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_view **views = NULL;
    unsigned i;
 
-   if (_textures) {
-      for (i = 0; i < num_textures; i++)
-         unwrapped_textures[i] = identity_texture_unwrap(_textures[i]);
+   if (_views) {
+      for (i = 0; i < num; i++)
+         unwrapped_views[i] = identity_sampler_view_unwrap(_views[i]);
       for (; i < PIPE_MAX_SAMPLERS; i++)
-         unwrapped_textures[i] = NULL;
+         unwrapped_views[i] = NULL;
 
-      textures = unwrapped_textures;
+      views = unwrapped_views;
    }
 
-   pipe->set_fragment_sampler_textures(pipe,
-                                       num_textures,
-                                       textures);
+   pipe->set_fragment_sampler_views(pipe, num, views);
 }
 
 static void
-identity_set_vertex_sampler_textures(struct pipe_context *_pipe,
-                                     unsigned num_textures,
-                                     struct pipe_texture **_textures)
+identity_set_vertex_sampler_views(struct pipe_context *_pipe,
+                                  unsigned num,
+                                  struct pipe_sampler_view **_views)
 {
    struct identity_context *id_pipe = identity_context(_pipe);
    struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_texture *unwrapped_textures[PIPE_MAX_VERTEX_SAMPLERS];
-   struct pipe_texture **textures = NULL;
+   struct pipe_sampler_view *unwrapped_views[PIPE_MAX_VERTEX_SAMPLERS];
+   struct pipe_sampler_view **views = NULL;
    unsigned i;
 
-   if (_textures) {
-      for (i = 0; i < num_textures; i++)
-         unwrapped_textures[i] = identity_texture_unwrap(_textures[i]);
+   if (_views) {
+      for (i = 0; i < num; i++)
+         unwrapped_views[i] = identity_sampler_view_unwrap(_views[i]);
       for (; i < PIPE_MAX_VERTEX_SAMPLERS; i++)
-         unwrapped_textures[i] = NULL;
+         unwrapped_views[i] = NULL;
 
-      textures = unwrapped_textures;
+      views = unwrapped_views;
    }
 
-   pipe->set_vertex_sampler_textures(pipe,
-                                     num_textures,
-                                     textures);
+   pipe->set_vertex_sampler_views(pipe, num, views);
 }
 
 static void
@@ -711,6 +707,45 @@ identity_is_buffer_referenced(struct pipe_context *_pipe,
                                      buffer);
 }
 
+static struct pipe_sampler_view *
+identity_create_sampler_view(struct pipe_context *pipe,
+                             struct pipe_texture *texture,
+                             const struct pipe_sampler_view *templ)
+{
+   struct identity_context *id_pipe = identity_context(pipe);
+   struct identity_texture *id_texture = identity_texture(texture);
+   struct pipe_context *pipe_unwrapped = id_pipe->pipe;
+   struct pipe_texture *texture_unwrapped = id_texture->texture;
+   struct identity_sampler_view *view = malloc(sizeof(struct identity_sampler_view));
+
+   view->sampler_view = pipe_unwrapped->create_sampler_view(pipe_unwrapped,
+                                                            texture_unwrapped,
+                                                            templ);
+
+   view->base = *templ;
+   view->base.reference.count = 1;
+   view->base.texture = NULL;
+   pipe_texture_reference(&view->base.texture, texture);
+   view->base.context = pipe;
+
+   return &view->base;
+}
+
+static void
+identity_sampler_view_destroy(struct pipe_context *pipe,
+                              struct pipe_sampler_view *view)
+{
+   struct identity_context *id_pipe = identity_context(pipe);
+   struct identity_sampler_view *id_view = identity_sampler_view(view);
+   struct pipe_context *pipe_unwrapped = id_pipe->pipe;
+   struct pipe_sampler_view *view_unwrapped = id_view->sampler_view;
+
+   pipe_unwrapped->sampler_view_destroy(pipe_unwrapped,
+                                        view_unwrapped);
+
+   pipe_texture_reference(&view->texture, NULL);
+   free(view);
+}
 
 
 static struct pipe_transfer *
@@ -836,8 +871,8 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.set_polygon_stipple = identity_set_polygon_stipple;
    id_pipe->base.set_scissor_state = identity_set_scissor_state;
    id_pipe->base.set_viewport_state = identity_set_viewport_state;
-   id_pipe->base.set_fragment_sampler_textures = identity_set_fragment_sampler_textures;
-   id_pipe->base.set_vertex_sampler_textures = identity_set_vertex_sampler_textures;
+   id_pipe->base.set_fragment_sampler_views = identity_set_fragment_sampler_views;
+   id_pipe->base.set_vertex_sampler_views = identity_set_vertex_sampler_views;
    id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers;
    id_pipe->base.surface_copy = identity_surface_copy;
    id_pipe->base.surface_fill = identity_surface_fill;
@@ -845,6 +880,8 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.flush = identity_flush;
    id_pipe->base.is_texture_referenced = identity_is_texture_referenced;
    id_pipe->base.is_buffer_referenced = identity_is_buffer_referenced;
+   id_pipe->base.create_sampler_view = identity_create_sampler_view;
+   id_pipe->base.sampler_view_destroy = identity_sampler_view_destroy;
    id_pipe->base.get_tex_transfer = identity_context_get_tex_transfer;
    id_pipe->base.tex_transfer_destroy = identity_context_tex_transfer_destroy;
    id_pipe->base.transfer_map = identity_context_transfer_map;
diff --git a/src/gallium/drivers/identity/id_drm.c b/src/gallium/drivers/identity/id_drm.c
index 936ccc444a8..26920fed08d 100644
--- a/src/gallium/drivers/identity/id_drm.c
+++ b/src/gallium/drivers/identity/id_drm.c
@@ -31,7 +31,6 @@
 #include "id_drm.h"
 #include "id_screen.h"
 #include "id_public.h"
-#include "id_screen.h"
 #include "id_objects.h"
 
 struct identity_drm_api
diff --git a/src/gallium/drivers/identity/id_objects.h b/src/gallium/drivers/identity/id_objects.h
index 7333ecfb7fb..9a07ebe8d72 100644
--- a/src/gallium/drivers/identity/id_objects.h
+++ b/src/gallium/drivers/identity/id_objects.h
@@ -53,6 +53,14 @@ struct identity_texture
 };
 
 
+struct identity_sampler_view
+{
+   struct pipe_sampler_view base;
+
+   struct pipe_sampler_view *sampler_view;
+};
+
+
 struct identity_surface
 {
    struct pipe_surface base;
@@ -96,6 +104,15 @@ identity_texture(struct pipe_texture *_texture)
    return (struct identity_texture *)_texture;
 }
 
+static INLINE struct identity_sampler_view *
+identity_sampler_view(struct pipe_sampler_view *_sampler_view)
+{
+   if (!_sampler_view) {
+      return NULL;
+   }
+   return (struct identity_sampler_view *)_sampler_view;
+}
+
 static INLINE struct identity_surface *
 identity_surface(struct pipe_surface *_surface)
 {
@@ -140,6 +157,15 @@ identity_texture_unwrap(struct pipe_texture *_texture)
    return identity_texture(_texture)->texture;
 }
 
+static INLINE struct pipe_sampler_view *
+identity_sampler_view_unwrap(struct pipe_sampler_view *_sampler_view)
+{
+   if (!_sampler_view) {
+      return NULL;
+   }
+   return identity_sampler_view(_sampler_view)->sampler_view;
+}
+
 static INLINE struct pipe_surface *
 identity_surface_unwrap(struct pipe_surface *_surface)
 {
diff --git a/src/gallium/drivers/llvmpipe/Makefile b/src/gallium/drivers/llvmpipe/Makefile
index 89c06ea3ad7..b37c6754c08 100644
--- a/src/gallium/drivers/llvmpipe/Makefile
+++ b/src/gallium/drivers/llvmpipe/Makefile
@@ -42,6 +42,11 @@ C_SOURCES = \
 
 CPP_SOURCES = \
 
+PROGS := lp_test_format	\
+	 lp_test_blend	\
+	 lp_test_conv	\
+	 lp_test_printf
+
 
 include ../../Makefile.template
 
@@ -49,13 +54,7 @@ lp_tile_soa.c: lp_tile_soa.py ../../auxiliary/util/u_format_parse.py ../../auxil
 	python lp_tile_soa.py ../../auxiliary/util/u_format.csv > $@
 
 
-testprogs := lp_test_format	\
-	     lp_test_blend	\
-	     lp_test_conv
-
-LIBS += $(GL_LIB_DEPS) -L. -lllvmpipe -L../../auxiliary/ -lgallium
+LIBS += $(GL_LIB_DEPS) -L../../auxiliary/ -lgallium
 
-#$(testprogs): lp_test_% : lp_test_%.o lp_test_main.o libllvmpipe.a
-#	$(LD) $(filter %.o,$^) -o $@ -Wl,--start-group  $(LIBS) -Wl,--end-group
+$(PROGS): lp_test_main.o
 
-#default: $(testprogs)
diff --git a/src/gallium/drivers/llvmpipe/lp_context.c b/src/gallium/drivers/llvmpipe/lp_context.c
index e0676c80a25..383e4b0614d 100644
--- a/src/gallium/drivers/llvmpipe/lp_context.c
+++ b/src/gallium/drivers/llvmpipe/lp_context.c
@@ -69,11 +69,11 @@ static void llvmpipe_destroy( struct pipe_context *pipe )
    pipe_surface_reference(&llvmpipe->framebuffer.zsbuf, NULL);
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      pipe_texture_reference(&llvmpipe->texture[i], NULL);
+      pipe_sampler_view_reference(&llvmpipe->fragment_sampler_views[i], NULL);
    }
 
    for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
-      pipe_texture_reference(&llvmpipe->vertex_textures[i], NULL);
+      pipe_sampler_view_reference(&llvmpipe->vertex_sampler_views[i], NULL);
    }
 
    for (i = 0; i < Elements(llvmpipe->constants); i++) {
@@ -158,8 +158,10 @@ llvmpipe_create_context( struct pipe_screen *screen, void *priv )
    llvmpipe->pipe.set_framebuffer_state = llvmpipe_set_framebuffer_state;
    llvmpipe->pipe.set_polygon_stipple = llvmpipe_set_polygon_stipple;
    llvmpipe->pipe.set_scissor_state = llvmpipe_set_scissor_state;
-   llvmpipe->pipe.set_fragment_sampler_textures = llvmpipe_set_sampler_textures;
-   llvmpipe->pipe.set_vertex_sampler_textures = llvmpipe_set_vertex_sampler_textures;
+   llvmpipe->pipe.set_fragment_sampler_views = llvmpipe_set_fragment_sampler_views;
+   llvmpipe->pipe.set_vertex_sampler_views = llvmpipe_set_vertex_sampler_views;
+   llvmpipe->pipe.create_sampler_view = llvmpipe_create_sampler_view;
+   llvmpipe->pipe.sampler_view_destroy = llvmpipe_sampler_view_destroy;
    llvmpipe->pipe.set_viewport_state = llvmpipe_set_viewport_state;
 
    llvmpipe->pipe.set_vertex_buffers = llvmpipe_set_vertex_buffers;
diff --git a/src/gallium/drivers/llvmpipe/lp_context.h b/src/gallium/drivers/llvmpipe/lp_context.h
index f391871b0ee..71f991049e5 100644
--- a/src/gallium/drivers/llvmpipe/lp_context.h
+++ b/src/gallium/drivers/llvmpipe/lp_context.h
@@ -69,15 +69,15 @@ struct llvmpipe_context {
    struct pipe_framebuffer_state framebuffer;
    struct pipe_poly_stipple poly_stipple;
    struct pipe_scissor_state scissor;
-   struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
-   struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS];
+   struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_view *vertex_sampler_views[PIPE_MAX_VERTEX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
 
    unsigned num_samplers;
-   unsigned num_textures;
+   unsigned num_fragment_sampler_views;
    unsigned num_vertex_samplers;
-   unsigned num_vertex_textures;
+   unsigned num_vertex_sampler_views;
    unsigned num_vertex_buffers;
 
    unsigned dirty; /**< Mask of LP_NEW_x flags */
diff --git a/src/gallium/drivers/llvmpipe/lp_fence.c b/src/gallium/drivers/llvmpipe/lp_fence.c
index 525c117f316..00dc3eab6b1 100644
--- a/src/gallium/drivers/llvmpipe/lp_fence.c
+++ b/src/gallium/drivers/llvmpipe/lp_fence.c
@@ -29,6 +29,7 @@
 #include "pipe/p_screen.h"
 #include "util/u_memory.h"
 #include "util/u_inlines.h"
+#include "lp_debug.h"
 #include "lp_fence.h"
 
 
@@ -99,6 +100,21 @@ llvmpipe_fence_finish(struct pipe_screen *screen,
 }
 
 
+void
+lp_fence_signal(struct lp_fence *fence)
+{
+   pipe_mutex_lock(fence->mutex);
+
+   fence->count++;
+   assert(fence->count <= fence->rank);
+
+   LP_DBG(DEBUG_RAST, "%s count=%u rank=%u\n", __FUNCTION__,
+          fence->count, fence->rank);
+
+   pipe_condvar_signal(fence->signalled);
+
+   pipe_mutex_unlock(fence->mutex);
+}
 
 
 void
diff --git a/src/gallium/drivers/llvmpipe/lp_fence.h b/src/gallium/drivers/llvmpipe/lp_fence.h
index c90e6de423b..d9270f5784a 100644
--- a/src/gallium/drivers/llvmpipe/lp_fence.h
+++ b/src/gallium/drivers/llvmpipe/lp_fence.h
@@ -54,6 +54,10 @@ lp_fence_create(unsigned rank);
 
 
 void
+lp_fence_signal(struct lp_fence *fence);
+
+
+void
 llvmpipe_init_screen_fence_funcs(struct pipe_screen *screen);
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_flush.c b/src/gallium/drivers/llvmpipe/lp_flush.c
index 636d72a9bb8..782669a1e77 100644
--- a/src/gallium/drivers/llvmpipe/lp_flush.c
+++ b/src/gallium/drivers/llvmpipe/lp_flush.c
@@ -111,7 +111,6 @@ llvmpipe_flush_texture(struct pipe_context *pipe,
                        boolean cpu_access,
                        boolean do_not_flush)
 {
-   struct pipe_fence_handle *last_fence = NULL;
    unsigned referenced;
 
    referenced = pipe->is_texture_referenced(pipe, texture, face, level);
@@ -142,7 +141,7 @@ llvmpipe_flush_texture(struct pipe_context *pipe,
 
          pipe->flush(pipe, flush_flags, &fence);
 
-         if (last_fence) {
+         if (fence) {
             pipe->screen->fence_finish(pipe->screen, fence, 0);
             pipe->screen->fence_reference(pipe->screen, &fence, NULL);
          }
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c
index 5887613120d..927e472ff26 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.c
+++ b/src/gallium/drivers/llvmpipe/lp_jit.c
@@ -91,36 +91,53 @@ lp_jit_init_globals(struct llvmpipe_screen *screen)
 
    /* struct lp_jit_context */
    {
-      LLVMTypeRef elem_types[8];
+      LLVMTypeRef elem_types[LP_JIT_CTX_COUNT];
       LLVMTypeRef context_type;
 
-      elem_types[0] = LLVMPointerType(LLVMFloatType(), 0); /* constants */
-      elem_types[1] = LLVMFloatType();                     /* alpha_ref_value */      elem_types[2] = LLVMFloatType();                     /* scissor_xmin */
-      elem_types[3] = LLVMFloatType();                     /* scissor_ymin */
-      elem_types[4] = LLVMFloatType();                     /* scissor_xmax */
-      elem_types[5] = LLVMFloatType();                     /* scissor_ymax */
-      elem_types[6] = LLVMPointerType(LLVMInt8Type(), 0);  /* blend_color */
-      elem_types[7] = LLVMArrayType(texture_type, PIPE_MAX_SAMPLERS); /* textures */
+      elem_types[LP_JIT_CTX_CONSTANTS] = LLVMPointerType(LLVMFloatType(), 0);
+      elem_types[LP_JIT_CTX_ALPHA_REF] = LLVMFloatType();
+      elem_types[LP_JIT_CTX_STENCIL_REF_FRONT] = LLVMInt32Type();
+      elem_types[LP_JIT_CTX_STENCIL_REF_BACK] = LLVMInt32Type();
+      elem_types[LP_JIT_CTX_SCISSOR_XMIN] = LLVMFloatType();
+      elem_types[LP_JIT_CTX_SCISSOR_YMIN] = LLVMFloatType();
+      elem_types[LP_JIT_CTX_SCISSOR_XMAX] = LLVMFloatType();
+      elem_types[LP_JIT_CTX_SCISSOR_YMAX] = LLVMFloatType();
+      elem_types[LP_JIT_CTX_BLEND_COLOR] = LLVMPointerType(LLVMInt8Type(), 0);
+      elem_types[LP_JIT_CTX_TEXTURES] = LLVMArrayType(texture_type,
+                                                      PIPE_MAX_SAMPLERS);
 
       context_type = LLVMStructType(elem_types, Elements(elem_types), 0);
 
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, constants,
-                             screen->target, context_type, 0);
+                             screen->target, context_type,
+                             LP_JIT_CTX_CONSTANTS);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, alpha_ref_value,
-                             screen->target, context_type, 1);
+                             screen->target, context_type,
+                             LP_JIT_CTX_ALPHA_REF);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, stencil_ref_front,
+                             screen->target, context_type,
+                             LP_JIT_CTX_STENCIL_REF_FRONT);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, stencil_ref_back,
+                             screen->target, context_type,
+                             LP_JIT_CTX_STENCIL_REF_BACK);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, scissor_xmin,
-                             screen->target, context_type, 2);
+                             screen->target, context_type,
+                             LP_JIT_CTX_SCISSOR_XMIN);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, scissor_ymin,
-                             screen->target, context_type, 3);
+                             screen->target, context_type,
+                             LP_JIT_CTX_SCISSOR_YMIN);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, scissor_xmax,
-                             screen->target, context_type, 4);
+                             screen->target, context_type,
+                             LP_JIT_CTX_SCISSOR_XMAX);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, scissor_ymax,
-                             screen->target, context_type, 5);
+                             screen->target, context_type,
+                             LP_JIT_CTX_SCISSOR_YMAX);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, blend_color,
-                             screen->target, context_type, 6);
+                             screen->target, context_type,
+                             LP_JIT_CTX_BLEND_COLOR);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, textures,
                              screen->target, context_type,
-                             LP_JIT_CONTEXT_TEXTURES_INDEX);
+                             LP_JIT_CTX_TEXTURES);
       LP_CHECK_STRUCT_SIZE(struct lp_jit_context,
                            screen->target, context_type);
 
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.h b/src/gallium/drivers/llvmpipe/lp_jit.h
index 13167ae3bf4..4930ff02e6b 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.h
+++ b/src/gallium/drivers/llvmpipe/lp_jit.h
@@ -84,6 +84,8 @@ struct lp_jit_context
 
    float alpha_ref_value;
 
+   uint32_t stencil_ref_front, stencil_ref_back;
+
    /** floats, not ints */
    float scissor_xmin, scissor_ymin, scissor_xmax, scissor_ymax;
 
@@ -94,37 +96,66 @@ struct lp_jit_context
 };
 
 
+/**
+ * These enum values must match the position of the fields in the
+ * lp_jit_context struct above.
+ */
+enum {
+   LP_JIT_CTX_CONSTANTS = 0,
+   LP_JIT_CTX_ALPHA_REF,
+   LP_JIT_CTX_STENCIL_REF_FRONT,
+   LP_JIT_CTX_STENCIL_REF_BACK,
+   LP_JIT_CTX_SCISSOR_XMIN,
+   LP_JIT_CTX_SCISSOR_YMIN,
+   LP_JIT_CTX_SCISSOR_XMAX,
+   LP_JIT_CTX_SCISSOR_YMAX,
+   LP_JIT_CTX_BLEND_COLOR,
+   LP_JIT_CTX_TEXTURES,
+   LP_JIT_CTX_COUNT
+};
+
+
 #define lp_jit_context_constants(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 0, "constants")
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_CONSTANTS, "constants")
 
 #define lp_jit_context_alpha_ref_value(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 1, "alpha_ref_value")
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_ALPHA_REF, "alpha_ref_value")
+
+#define lp_jit_context_stencil_ref_front_value(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_STENCIL_REF_FRONT, "stencil_ref_front")
+
+#define lp_jit_context_stencil_ref_back_value(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_STENCIL_REF_BACK, "stencil_ref_back")
 
 #define lp_jit_context_scissor_xmin_value(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 2, "scissor_xmin")
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_SCISSOR_XMIN, "scissor_xmin")
 
 #define lp_jit_context_scissor_ymin_value(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 3, "scissor_ymin")
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_SCISSOR_YMIN, "scissor_ymin")
 
 #define lp_jit_context_scissor_xmax_value(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 4, "scissor_xmax")
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_SCISSOR_XMAX, "scissor_xmax")
 
 #define lp_jit_context_scissor_ymax_value(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 5, "scissor_ymax")
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_SCISSOR_YMAX, "scissor_ymax")
 
 #define lp_jit_context_blend_color(_builder, _ptr) \
-   lp_build_struct_get(_builder, _ptr, 6, "blend_color")
-
-#define LP_JIT_CONTEXT_TEXTURES_INDEX 7
+   lp_build_struct_get(_builder, _ptr, LP_JIT_CTX_BLEND_COLOR, "blend_color")
 
 #define lp_jit_context_textures(_builder, _ptr) \
-   lp_build_struct_get_ptr(_builder, _ptr, LP_JIT_CONTEXT_TEXTURES_INDEX, "textures")
+   lp_build_struct_get_ptr(_builder, _ptr, LP_JIT_CONTEXT_TEXTURES, "textures")
+
+
+/** Indexes into jit_function[] array */
+#define RAST_WHOLE 0
+#define RAST_EDGE_TEST 1
 
 
 typedef void
 (*lp_jit_frag_func)(const struct lp_jit_context *context,
                     uint32_t x,
                     uint32_t y,
+                    float facing,
                     const void *a0,
                     const void *dadx,
                     const void *dady,
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index 81ea11a16b6..8352f175598 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -189,7 +189,7 @@ lp_rast_clear_zstencil(struct lp_rasterizer_task *task,
 
    LP_DBG(DEBUG_RAST, "%s 0x%x\n", __FUNCTION__, arg.clear_zstencil);
 
-   assert(rast->zsbuf.map);
+   /*assert(rast->zsbuf.map);*/
    if (!rast->zsbuf.map)
       return;
 
@@ -312,15 +312,16 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
          depth = lp_rast_depth_pointer(rast, tile_x + x, tile_y + y);
 
          /* run shader */
-         state->jit_function[0]( &state->jit_context,
-                                 tile_x + x, tile_y + y,
-                                 inputs->a0,
-                                 inputs->dadx,
-                                 inputs->dady,
-                                 color,
-                                 depth,
-                                 INT_MIN, INT_MIN, INT_MIN,
-                                 NULL, NULL, NULL );
+         state->jit_function[RAST_WHOLE]( &state->jit_context,
+                                          tile_x + x, tile_y + y,
+                                          inputs->facing,
+                                          inputs->a0,
+                                          inputs->dadx,
+                                          inputs->dady,
+                                          color,
+                                          depth,
+                                          INT_MIN, INT_MIN, INT_MIN,
+                                          NULL, NULL, NULL );
       }
    }
 }
@@ -375,15 +376,18 @@ void lp_rast_shade_quads( struct lp_rasterizer_task *task,
    assert(lp_check_alignment(inputs->step[2], 16));
 
    /* run shader */
-   state->jit_function[1]( &state->jit_context,
-                        x, y,
-                        inputs->a0,
-                        inputs->dadx,
-                        inputs->dady,
-                        color,
-                        depth,
-                        c1, c2, c3,
-                        inputs->step[0], inputs->step[1], inputs->step[2]);
+   state->jit_function[RAST_EDGE_TEST]( &state->jit_context,
+                                        x, y,
+                                        inputs->facing,
+                                        inputs->a0,
+                                        inputs->dadx,
+                                        inputs->dady,
+                                        color,
+                                        depth,
+                                        c1, c2, c3,
+                                        inputs->step[0],
+                                        inputs->step[1],
+                                        inputs->step[2]);
 }
 
 
@@ -487,18 +491,7 @@ lp_rast_fence(struct lp_rasterizer_task *task,
               const union lp_rast_cmd_arg arg)
 {
    struct lp_fence *fence = arg.fence;
-
-   pipe_mutex_lock( fence->mutex );
-
-   fence->count++;
-   assert(fence->count <= fence->rank);
-
-   LP_DBG(DEBUG_RAST, "%s count=%u rank=%u\n", __FUNCTION__,
-          fence->count, fence->rank);
-
-   pipe_condvar_signal( fence->signalled );
-
-   pipe_mutex_unlock( fence->mutex );
+   lp_fence_signal(fence);
 }
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.h b/src/gallium/drivers/llvmpipe/lp_rast.h
index 303f6e3f7e4..ae838f3fbef 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.h
+++ b/src/gallium/drivers/llvmpipe/lp_rast.h
@@ -82,6 +82,8 @@ struct lp_rast_state {
  * These pointers point into the bin data buffer.
  */
 struct lp_rast_shader_inputs {
+   float facing;     /** Positive for front-facing, negative for back-facing */
+
    float (*a0)[4];
    float (*dadx)[4];
    float (*dady)[4];
diff --git a/src/gallium/drivers/llvmpipe/lp_rast_priv.h b/src/gallium/drivers/llvmpipe/lp_rast_priv.h
index 39bf2c25879..6ee9bcaae3a 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast_priv.h
+++ b/src/gallium/drivers/llvmpipe/lp_rast_priv.h
@@ -195,6 +195,7 @@ lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
    /* run shader */
    state->jit_function[0]( &state->jit_context,
                            x, y,
+                           inputs->facing,
                            inputs->a0,
                            inputs->dadx,
                            inputs->dady,
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c
index 5093f58bb19..f1bbc2092c4 100644
--- a/src/gallium/drivers/llvmpipe/lp_screen.c
+++ b/src/gallium/drivers/llvmpipe/lp_screen.c
@@ -204,8 +204,10 @@ llvmpipe_is_format_supported( struct pipe_screen *_screen,
          return FALSE;
    }
 
-   if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
-      if(!winsys->is_displaytarget_format_supported(winsys, format))
+   if(tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                   PIPE_TEXTURE_USAGE_SCANOUT |
+                   PIPE_TEXTURE_USAGE_SHARED)) {
+      if(!winsys->is_displaytarget_format_supported(winsys, tex_usage, format))
          return FALSE;
    }
 
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.h b/src/gallium/drivers/llvmpipe/lp_screen.h
index d977f98cfaa..af25e043cc9 100644
--- a/src/gallium/drivers/llvmpipe/lp_screen.h
+++ b/src/gallium/drivers/llvmpipe/lp_screen.h
@@ -34,7 +34,7 @@
 #ifndef LP_SCREEN_H
 #define LP_SCREEN_H
 
-#include "os/os_llvm.h"
+#include "gallivm/lp_bld.h"
 #include <llvm-c/ExecutionEngine.h>
 
 #include "pipe/p_screen.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
index 16128c34c80..76a8b87a309 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -402,6 +402,20 @@ lp_setup_set_alpha_ref_value( struct lp_setup_context *setup,
 }
 
 void
+lp_setup_set_stencil_ref_values( struct lp_setup_context *setup,
+                                 const ubyte refs[2] )
+{
+   LP_DBG(DEBUG_SETUP, "%s %d %d\n", __FUNCTION__, refs[0], refs[1]);
+
+   if (setup->fs.current.jit_context.stencil_ref_front != refs[0] ||
+       setup->fs.current.jit_context.stencil_ref_back != refs[1]) {
+      setup->fs.current.jit_context.stencil_ref_front = refs[0];
+      setup->fs.current.jit_context.stencil_ref_back = refs[1];
+      setup->dirty |= LP_SETUP_NEW_FS;
+   }
+}
+
+void
 lp_setup_set_blend_color( struct lp_setup_context *setup,
                           const struct pipe_blend_color *blend_color )
 {
@@ -450,11 +464,12 @@ lp_setup_set_vertex_info( struct lp_setup_context *setup,
 
 
 /**
- * Called during state validation when LP_NEW_TEXTURE is set.
+ * Called during state validation when LP_NEW_SAMPLER_VIEW is set.
  */
 void
-lp_setup_set_sampler_textures( struct lp_setup_context *setup,
-                               unsigned num, struct pipe_texture **texture)
+lp_setup_set_fragment_sampler_views(struct lp_setup_context *setup,
+                                    unsigned num,
+                                    struct pipe_sampler_view **views)
 {
    unsigned i;
 
@@ -463,9 +478,10 @@ lp_setup_set_sampler_textures( struct lp_setup_context *setup,
    assert(num <= PIPE_MAX_SAMPLERS);
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      struct pipe_texture *tex = i < num ? texture[i] : NULL;
+      struct pipe_sampler_view *view = i < num ? views[i] : NULL;
 
-      if(tex) {
+      if(view) {
+         struct pipe_texture *tex = view->texture;
          struct llvmpipe_texture *lp_tex = llvmpipe_texture(tex);
          struct lp_jit_texture *jit_tex;
          jit_tex = &setup->fs.current.jit_context.textures[i];
@@ -473,6 +489,12 @@ lp_setup_set_sampler_textures( struct lp_setup_context *setup,
          jit_tex->height = tex->height0;
          jit_tex->depth = tex->depth0;
          jit_tex->last_level = tex->last_level;
+
+         /* We're referencing the texture's internal data, so save a
+          * reference to it.
+          */
+         pipe_texture_reference(&setup->fs.current_tex[i], tex);
+
          if (!lp_tex->dt) {
             /* regular texture - setup array of mipmap level pointers */
             int j;
@@ -495,12 +517,6 @@ lp_setup_set_sampler_textures( struct lp_setup_context *setup,
             jit_tex->row_stride[0] = lp_tex->stride[0];
             assert(jit_tex->data[0]);
          }
-
-         /* the scene references this texture */
-         {
-            struct lp_scene *scene = lp_setup_get_current_scene(setup);
-            lp_scene_texture_reference(scene, tex);
-         }
       }
    }
 
@@ -635,6 +651,7 @@ lp_setup_update_state( struct lp_setup_context *setup )
           * the new, current state.  So allocate a new lp_rast_state object
           * and append it to the bin's setup data buffer.
           */
+         uint i;
          struct lp_rast_state *stored =
             (struct lp_rast_state *) lp_scene_alloc(scene, sizeof *stored);
          if(stored) {
@@ -648,6 +665,14 @@ lp_setup_update_state( struct lp_setup_context *setup )
 					lp_rast_set_state, 
 					lp_rast_arg_state(setup->fs.stored) );
          }
+
+         /* The scene now references the textures in the rasterization
+          * state record.  Note that now.
+          */
+         for (i = 0; i < Elements(setup->fs.current_tex); i++) {
+            if (setup->fs.current_tex[i])
+               lp_scene_texture_reference(scene, setup->fs.current_tex[i]);
+         }
       }
    }
 
@@ -663,8 +688,14 @@ lp_setup_update_state( struct lp_setup_context *setup )
 void 
 lp_setup_destroy( struct lp_setup_context *setup )
 {
+   uint i;
+
    reset_context( setup );
 
+   for (i = 0; i < Elements(setup->fs.current_tex); i++) {
+      pipe_texture_reference(&setup->fs.current_tex[i], NULL);
+   }
+
    pipe_buffer_reference(&setup->constants.current, NULL);
 
    /* free the scenes in the 'empty' queue */
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.h b/src/gallium/drivers/llvmpipe/lp_setup.h
index be1bf96f12d..dbfc1bf8d4c 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.h
+++ b/src/gallium/drivers/llvmpipe/lp_setup.h
@@ -113,6 +113,10 @@ lp_setup_set_alpha_ref_value( struct lp_setup_context *setup,
                               float alpha_ref_value );
 
 void
+lp_setup_set_stencil_ref_values( struct lp_setup_context *setup,
+                                 const ubyte refs[2] );
+
+void
 lp_setup_set_blend_color( struct lp_setup_context *setup,
                           const struct pipe_blend_color *blend_color );
 
@@ -121,8 +125,9 @@ lp_setup_set_scissor( struct lp_setup_context *setup,
                       const struct pipe_scissor_state *scissor );
 
 void
-lp_setup_set_sampler_textures( struct lp_setup_context *setup,
-                               unsigned num, struct pipe_texture **texture);
+lp_setup_set_fragment_sampler_views(struct lp_setup_context *setup,
+                                    unsigned num,
+                                    struct pipe_sampler_view **views);
 
 unsigned
 lp_setup_is_texture_referenced( const struct lp_setup_context *setup,
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_context.h b/src/gallium/drivers/llvmpipe/lp_setup_context.h
index 464fb369840..ca0dafab627 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_context.h
+++ b/src/gallium/drivers/llvmpipe/lp_setup_context.h
@@ -111,6 +111,7 @@ struct lp_setup_context
 
       const struct lp_rast_state *stored; /**< what's in the scene */
       struct lp_rast_state current;  /**< currently set state */
+      struct pipe_texture *current_tex[PIPE_MAX_SAMPLERS];
    } fs;
 
    /** fragment shader constants */
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_tri.c b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
index ac6264dc73e..ce689d3d568 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_tri.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
@@ -361,6 +361,8 @@ do_triangle_ccw(struct lp_setup_context *setup,
     */
    setup_tri_coefficients( setup, tri, oneoverarea, v1, v2, v3, frontfacing );
 
+   tri->inputs.facing = frontfacing ? 1.0F : -1.0F;
+
    /* half-edge constants, will be interated over the whole render target.
     */
    tri->c1 = tri->dy12 * x1 - tri->dx12 * y1;
diff --git a/src/gallium/drivers/llvmpipe/lp_state.h b/src/gallium/drivers/llvmpipe/lp_state.h
index a5a1a720742..74ebf90d580 100644
--- a/src/gallium/drivers/llvmpipe/lp_state.h
+++ b/src/gallium/drivers/llvmpipe/lp_state.h
@@ -31,7 +31,7 @@
 #ifndef LP_STATE_H
 #define LP_STATE_H
 
-#include "os/os_llvm.h"
+#include "gallivm/lp_bld.h"
 
 #include "pipe/p_state.h"
 #include "tgsi/tgsi_scan.h"
@@ -50,7 +50,7 @@
 #define LP_NEW_DEPTH_STENCIL_ALPHA 0x100
 #define LP_NEW_CONSTANTS     0x200
 #define LP_NEW_SAMPLER       0x400
-#define LP_NEW_TEXTURE       0x800
+#define LP_NEW_SAMPLER_VIEW  0x800
 #define LP_NEW_VERTEX        0x1000
 #define LP_NEW_VS            0x2000
 #define LP_NEW_QUERY         0x4000
@@ -67,6 +67,7 @@ struct lp_fragment_shader;
 struct lp_fragment_shader_variant_key
 {
    struct pipe_depth_state depth;
+   struct pipe_stencil_state stencil[2];
    struct pipe_alpha_state alpha;
    struct pipe_blend_state blend;
    enum pipe_format zsbuf_format;
@@ -192,14 +193,23 @@ void llvmpipe_set_polygon_stipple( struct pipe_context *,
 void llvmpipe_set_scissor_state( struct pipe_context *,
                                  const struct pipe_scissor_state * );
 
-void llvmpipe_set_sampler_textures( struct pipe_context *,
-                                    unsigned num,
-                                    struct pipe_texture ** );
+void llvmpipe_set_fragment_sampler_views(struct pipe_context *,
+                                         unsigned num,
+                                         struct pipe_sampler_view **);
 
 void
-llvmpipe_set_vertex_sampler_textures(struct pipe_context *,
-                                     unsigned num_textures,
-                                     struct pipe_texture **);
+llvmpipe_set_vertex_sampler_views(struct pipe_context *,
+                                  unsigned num,
+                                  struct pipe_sampler_view **);
+
+struct pipe_sampler_view *
+llvmpipe_create_sampler_view(struct pipe_context *pipe,
+                            struct pipe_texture *texture,
+                            const struct pipe_sampler_view *templ);
+
+void
+llvmpipe_sampler_view_destroy(struct pipe_context *pipe,
+                              struct pipe_sampler_view *view);
 
 void llvmpipe_set_viewport_state( struct pipe_context *,
                                   const struct pipe_viewport_state * );
diff --git a/src/gallium/drivers/llvmpipe/lp_state_derived.c b/src/gallium/drivers/llvmpipe/lp_state_derived.c
index bdd906e1a73..777871638f6 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_derived.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_derived.c
@@ -150,7 +150,7 @@ void llvmpipe_update_derived( struct llvmpipe_context *llvmpipe )
     */
    if (llvmpipe->tex_timestamp != lp_screen->timestamp) {
       llvmpipe->tex_timestamp = lp_screen->timestamp;
-      llvmpipe->dirty |= LP_NEW_TEXTURE;
+      llvmpipe->dirty |= LP_NEW_SAMPLER_VIEW;
    }
       
    if (llvmpipe->dirty & (LP_NEW_RASTERIZER |
@@ -164,7 +164,7 @@ void llvmpipe_update_derived( struct llvmpipe_context *llvmpipe )
                           LP_NEW_DEPTH_STENCIL_ALPHA |
                           LP_NEW_RASTERIZER |
                           LP_NEW_SAMPLER |
-                          LP_NEW_TEXTURE))
+                          LP_NEW_SAMPLER_VIEW))
       llvmpipe_update_fs( llvmpipe );
 
    if (llvmpipe->dirty & LP_NEW_BLEND_COLOR)
@@ -174,18 +174,21 @@ void llvmpipe_update_derived( struct llvmpipe_context *llvmpipe )
    if (llvmpipe->dirty & LP_NEW_SCISSOR)
       lp_setup_set_scissor(llvmpipe->setup, &llvmpipe->scissor);
 
-   if (llvmpipe->dirty & LP_NEW_DEPTH_STENCIL_ALPHA)
+   if (llvmpipe->dirty & LP_NEW_DEPTH_STENCIL_ALPHA) {
       lp_setup_set_alpha_ref_value(llvmpipe->setup, 
                                    llvmpipe->depth_stencil->alpha.ref_value);
+      lp_setup_set_stencil_ref_values(llvmpipe->setup,
+                                      llvmpipe->stencil_ref.ref_value);
+   }
 
    if (llvmpipe->dirty & LP_NEW_CONSTANTS)
       lp_setup_set_fs_constants(llvmpipe->setup, 
                                 llvmpipe->constants[PIPE_SHADER_FRAGMENT]);
 
-   if (llvmpipe->dirty & LP_NEW_TEXTURE)
-      lp_setup_set_sampler_textures(llvmpipe->setup, 
-                                    llvmpipe->num_textures,
-                                    llvmpipe->texture);
+   if (llvmpipe->dirty & LP_NEW_SAMPLER_VIEW)
+      lp_setup_set_fragment_sampler_views(llvmpipe->setup, 
+                                          llvmpipe->num_fragment_sampler_views,
+                                          llvmpipe->fragment_sampler_views);
 
    llvmpipe->dirty = 0;
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 9a8de0edfda..7bbf348e0b8 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -138,20 +138,22 @@ generate_pos0(LLVMBuilderRef builder,
 
 
 /**
- * Generate the depth test.
+ * Generate the depth /stencil test code.
  */
 static void
-generate_depth(LLVMBuilderRef builder,
-               const struct lp_fragment_shader_variant_key *key,
-               struct lp_type src_type,
-               struct lp_build_mask_context *mask,
-               LLVMValueRef src,
-               LLVMValueRef dst_ptr)
+generate_depth_stencil(LLVMBuilderRef builder,
+                       const struct lp_fragment_shader_variant_key *key,
+                       struct lp_type src_type,
+                       struct lp_build_mask_context *mask,
+                       LLVMValueRef stencil_refs[2],
+                       LLVMValueRef src,
+                       LLVMValueRef dst_ptr,
+                       LLVMValueRef facing)
 {
    const struct util_format_description *format_desc;
    struct lp_type dst_type;
 
-   if(!key->depth.enabled)
+   if (!key->depth.enabled && !key->stencil[0].enabled && !key->stencil[1].enabled)
       return;
 
    format_desc = util_format_description(key->zsbuf_format);
@@ -178,19 +180,22 @@ generate_depth(LLVMBuilderRef builder,
    assert(dst_type.width == src_type.width);
    assert(dst_type.length == src_type.length);
 
+   /* Convert fragment Z from float to integer */
    lp_build_conv(builder, src_type, dst_type, &src, 1, &src, 1);
 
    dst_ptr = LLVMBuildBitCast(builder,
                               dst_ptr,
                               LLVMPointerType(lp_build_vec_type(dst_type), 0), "");
-
-   lp_build_depth_test(builder,
-                       &key->depth,
-                       dst_type,
-                       format_desc,
-                       mask,
-                       src,
-                       dst_ptr);
+   lp_build_depth_stencil_test(builder,
+                               &key->depth,
+                               key->stencil,
+                               dst_type,
+                               format_desc,
+                               mask,
+                               stencil_refs,
+                               src,
+                               dst_ptr,
+                               facing);
 }
 
 
@@ -252,7 +257,7 @@ generate_tri_edge_mask(LLVMBuilderRef builder,
                                    LLVMConstInt(LLVMInt32Type(), INT_MIN, 0),
                                    "");
 
-      in_out_mask = lp_build_int_const_scalar(i32_type, ~0);
+      in_out_mask = lp_build_const_int_vec(i32_type, ~0);
 
 
       lp_build_flow_scope_declare(flow, &in_out_mask);
@@ -367,7 +372,7 @@ build_int32_vec_const(int value)
    i32_type.norm = FALSE;     /* values are not normalized */
    i32_type.width = 32;       /* 32-bit int values */
    i32_type.length = 4;       /* 4 elements per vector */
-   return lp_build_int_const_scalar(i32_type, value);
+   return lp_build_const_int_vec(i32_type, value);
 }
 
 
@@ -390,6 +395,7 @@ generate_fs(struct llvmpipe_context *lp,
             LLVMValueRef *pmask,
             LLVMValueRef (*color)[4],
             LLVMValueRef depth_ptr,
+            LLVMValueRef facing,
             unsigned do_tri_test,
             LLVMValueRef c0,
             LLVMValueRef c1,
@@ -405,15 +411,19 @@ generate_fs(struct llvmpipe_context *lp,
    LLVMValueRef consts_ptr;
    LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
    LLVMValueRef z = interp->pos[2];
+   LLVMValueRef stencil_refs[2];
    struct lp_build_flow_context *flow;
    struct lp_build_mask_context mask;
-   boolean early_depth_test;
+   boolean early_depth_stencil_test;
    unsigned attrib;
    unsigned chan;
    unsigned cbuf;
 
    assert(i < 4);
 
+   stencil_refs[0] = lp_jit_context_stencil_ref_front_value(builder, context_ptr);
+   stencil_refs[1] = lp_jit_context_stencil_ref_back_value(builder, context_ptr);
+
    elem_type = lp_build_elem_type(type);
    vec_type = lp_build_vec_type(type);
    int_vec_type = lp_build_int_vec_type(type);
@@ -453,16 +463,16 @@ generate_fs(struct llvmpipe_context *lp,
       lp_build_mask_update(&mask, smask);
    }
 
-   early_depth_test =
-      key->depth.enabled &&
+   early_depth_stencil_test =
+      (key->depth.enabled || key->stencil[0].enabled) &&
       !key->alpha.enabled &&
       !shader->info.uses_kill &&
       !shader->info.writes_z;
 
-   if(early_depth_test)
-      generate_depth(builder, key,
-                     type, &mask,
-                     z, depth_ptr);
+   if (early_depth_stencil_test)
+      generate_depth_stencil(builder, key,
+                             type, &mask,
+                             stencil_refs, z, depth_ptr, facing);
 
    lp_build_tgsi_soa(builder, tokens, type, &mask,
                      consts_ptr, interp->pos, interp->inputs,
@@ -506,10 +516,10 @@ generate_fs(struct llvmpipe_context *lp,
       }
    }
 
-   if(!early_depth_test)
-      generate_depth(builder, key,
-                     type, &mask,
-                     z, depth_ptr);
+   if (!early_depth_stencil_test)
+      generate_depth_stencil(builder, key,
+                             type, &mask,
+                             stencil_refs, z, depth_ptr, facing);
 
    lp_build_mask_end(&mask);
 
@@ -585,6 +595,20 @@ generate_blend(const struct pipe_blend_state *blend,
 }
 
 
+/** casting function to avoid compiler warnings */
+static lp_jit_frag_func
+cast_voidptr_to_lp_jit_frag_func(void *p)
+{
+   union {
+      void *v;
+      lp_jit_frag_func f;
+   } tmp;
+   assert(sizeof(tmp.v) == sizeof(tmp.f));
+   tmp.v = p;
+   return tmp.f;
+}
+
+
 /**
  * Generate the runtime callable function for the whole fragment pipeline.
  * Note that the function which we generate operates on a block of 16
@@ -606,7 +630,7 @@ generate_fragment(struct llvmpipe_context *lp,
    LLVMTypeRef fs_int_vec_type;
    LLVMTypeRef blend_vec_type;
    LLVMTypeRef blend_int_vec_type;
-   LLVMTypeRef arg_types[14];
+   LLVMTypeRef arg_types[15];
    LLVMTypeRef func_type;
    LLVMTypeRef int32_vec4_type = lp_build_int32_vec4_type();
    LLVMValueRef context_ptr;
@@ -629,6 +653,7 @@ generate_fragment(struct llvmpipe_context *lp,
    LLVMValueRef blend_mask;
    LLVMValueRef blend_in_color[NUM_CHANNELS];
    LLVMValueRef function;
+   LLVMValueRef facing;
    unsigned num_fs;
    unsigned i;
    unsigned chan;
@@ -668,20 +693,21 @@ generate_fragment(struct llvmpipe_context *lp,
    arg_types[0] = screen->context_ptr_type;            /* context */
    arg_types[1] = LLVMInt32Type();                     /* x */
    arg_types[2] = LLVMInt32Type();                     /* y */
-   arg_types[3] = LLVMPointerType(fs_elem_type, 0);    /* a0 */
-   arg_types[4] = LLVMPointerType(fs_elem_type, 0);    /* dadx */
-   arg_types[5] = LLVMPointerType(fs_elem_type, 0);    /* dady */
-   arg_types[6] = LLVMPointerType(LLVMPointerType(blend_vec_type, 0), 0);  /* color */
-   arg_types[7] = LLVMPointerType(fs_int_vec_type, 0); /* depth */
-   arg_types[8] = LLVMInt32Type();                     /* c0 */
-   arg_types[9] = LLVMInt32Type();                     /* c1 */
-   arg_types[10] = LLVMInt32Type();                    /* c2 */
+   arg_types[3] = LLVMFloatType();                     /* facing */
+   arg_types[4] = LLVMPointerType(fs_elem_type, 0);    /* a0 */
+   arg_types[5] = LLVMPointerType(fs_elem_type, 0);    /* dadx */
+   arg_types[6] = LLVMPointerType(fs_elem_type, 0);    /* dady */
+   arg_types[7] = LLVMPointerType(LLVMPointerType(blend_vec_type, 0), 0);  /* color */
+   arg_types[8] = LLVMPointerType(fs_int_vec_type, 0); /* depth */
+   arg_types[9] = LLVMInt32Type();                     /* c0 */
+   arg_types[10] = LLVMInt32Type();                    /* c1 */
+   arg_types[11] = LLVMInt32Type();                    /* c2 */
    /* Note: the step arrays are built as int32[16] but we interpret
     * them here as int32_vec4[4].
     */
-   arg_types[11] = LLVMPointerType(int32_vec4_type, 0);/* step0 */
-   arg_types[12] = LLVMPointerType(int32_vec4_type, 0);/* step1 */
-   arg_types[13] = LLVMPointerType(int32_vec4_type, 0);/* step2 */
+   arg_types[12] = LLVMPointerType(int32_vec4_type, 0);/* step0 */
+   arg_types[13] = LLVMPointerType(int32_vec4_type, 0);/* step1 */
+   arg_types[14] = LLVMPointerType(int32_vec4_type, 0);/* step2 */
 
    func_type = LLVMFunctionType(LLVMVoidType(), arg_types, Elements(arg_types), 0);
 
@@ -701,17 +727,18 @@ generate_fragment(struct llvmpipe_context *lp,
    context_ptr  = LLVMGetParam(function, 0);
    x            = LLVMGetParam(function, 1);
    y            = LLVMGetParam(function, 2);
-   a0_ptr       = LLVMGetParam(function, 3);
-   dadx_ptr     = LLVMGetParam(function, 4);
-   dady_ptr     = LLVMGetParam(function, 5);
-   color_ptr_ptr = LLVMGetParam(function, 6);
-   depth_ptr    = LLVMGetParam(function, 7);
-   c0           = LLVMGetParam(function, 8);
-   c1           = LLVMGetParam(function, 9);
-   c2           = LLVMGetParam(function, 10);
-   step0_ptr    = LLVMGetParam(function, 11);
-   step1_ptr    = LLVMGetParam(function, 12);
-   step2_ptr    = LLVMGetParam(function, 13);
+   facing       = LLVMGetParam(function, 3);
+   a0_ptr       = LLVMGetParam(function, 4);
+   dadx_ptr     = LLVMGetParam(function, 5);
+   dady_ptr     = LLVMGetParam(function, 6);
+   color_ptr_ptr = LLVMGetParam(function, 7);
+   depth_ptr    = LLVMGetParam(function, 8);
+   c0           = LLVMGetParam(function, 9);
+   c1           = LLVMGetParam(function, 10);
+   c2           = LLVMGetParam(function, 11);
+   step0_ptr    = LLVMGetParam(function, 12);
+   step1_ptr    = LLVMGetParam(function, 13);
+   step2_ptr    = LLVMGetParam(function, 14);
 
    lp_build_name(context_ptr, "context");
    lp_build_name(x, "x");
@@ -770,6 +797,7 @@ generate_fragment(struct llvmpipe_context *lp,
                   &fs_mask[i], /* output */
                   out_color,
                   depth_ptr_i,
+                  facing,
                   do_tri_test,
                   c0, c1, c2,
                   step0_ptr, step1_ptr, step2_ptr);
@@ -845,10 +873,14 @@ generate_fragment(struct llvmpipe_context *lp,
    /*
     * Translate the LLVM IR into machine code.
     */
-   variant->jit_function[do_tri_test] = (lp_jit_frag_func)LLVMGetPointerToGlobal(screen->engine, function);
+   {
+      void *f = LLVMGetPointerToGlobal(screen->engine, function);
+
+      variant->jit_function[do_tri_test] = cast_voidptr_to_lp_jit_frag_func(f);
 
-   if (LP_DEBUG & DEBUG_ASM)
-      lp_disassemble(variant->jit_function[do_tri_test]);
+      if (LP_DEBUG & DEBUG_ASM)
+         lp_disassemble(f);
+   }
 }
 
 
@@ -1054,10 +1086,15 @@ make_variant_key(struct llvmpipe_context *lp,
 
    memset(key, 0, sizeof *key);
 
-   if(lp->framebuffer.zsbuf &&
-      lp->depth_stencil->depth.enabled) {
-      key->zsbuf_format = lp->framebuffer.zsbuf->format;
-      memcpy(&key->depth, &lp->depth_stencil->depth, sizeof key->depth);
+   if (lp->framebuffer.zsbuf) {
+      if (lp->depth_stencil->depth.enabled) {
+         key->zsbuf_format = lp->framebuffer.zsbuf->format;
+         memcpy(&key->depth, &lp->depth_stencil->depth, sizeof key->depth);
+      }
+      if (lp->depth_stencil->stencil[0].enabled) {
+         key->zsbuf_format = lp->framebuffer.zsbuf->format;
+         memcpy(&key->stencil, &lp->depth_stencil->stencil, sizeof key->stencil);
+      }
    }
 
    key->alpha.enabled = lp->depth_stencil->alpha.enabled;
@@ -1094,7 +1131,7 @@ make_variant_key(struct llvmpipe_context *lp,
 
    for(i = 0; i < PIPE_MAX_SAMPLERS; ++i)
       if(shader->info.file_mask[TGSI_FILE_SAMPLER] & (1 << i))
-         lp_sampler_static_state(&key->sampler[i], lp->texture[i], lp->sampler[i]);
+         lp_sampler_static_state(&key->sampler[i], lp->fragment_sampler_views[i]->texture, lp->sampler[i]);
 }
 
 
@@ -1140,6 +1177,7 @@ llvmpipe_update_fs(struct llvmpipe_context *lp)
    opaque = !key.blend.logicop_enable &&
             !key.blend.rt[0].blend_enable &&
             key.blend.rt[0].colormask == 0xf &&
+            !key.stencil[0].enabled &&
             !key.alpha.enabled &&
             !key.depth.enabled &&
             !key.scissor &&
@@ -1147,7 +1185,7 @@ llvmpipe_update_fs(struct llvmpipe_context *lp)
             ? TRUE : FALSE;
 
    lp_setup_set_fs_functions(lp->setup, 
-                             shader->current->jit_function[0],
-                             shader->current->jit_function[1],
+                             shader->current->jit_function[RAST_WHOLE],
+                             shader->current->jit_function[RAST_EDGE_TEST],
                              opaque);
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_state_sampler.c b/src/gallium/drivers/llvmpipe/lp_state_sampler.c
index b30a0757768..2645441b58c 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_sampler.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_sampler.c
@@ -105,8 +105,9 @@ llvmpipe_bind_vertex_sampler_states(struct pipe_context *pipe,
 
 
 void
-llvmpipe_set_sampler_textures(struct pipe_context *pipe,
-                              unsigned num, struct pipe_texture **texture)
+llvmpipe_set_fragment_sampler_views(struct pipe_context *pipe,
+                                    unsigned num,
+                                    struct pipe_sampler_view **views)
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
    uint i;
@@ -114,51 +115,79 @@ llvmpipe_set_sampler_textures(struct pipe_context *pipe,
    assert(num <= PIPE_MAX_SAMPLERS);
 
    /* Check for no-op */
-   if (num == llvmpipe->num_textures &&
-       !memcmp(llvmpipe->texture, texture, num * sizeof(struct pipe_texture *)))
+   if (num == llvmpipe->num_fragment_sampler_views &&
+       !memcmp(llvmpipe->fragment_sampler_views, views, num * sizeof(struct pipe_sampler_view *)))
       return;
 
    draw_flush(llvmpipe->draw);
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      struct pipe_texture *tex = i < num ? texture[i] : NULL;
+      struct pipe_sampler_view *view = i < num ? views[i] : NULL;
 
-      pipe_texture_reference(&llvmpipe->texture[i], tex);
+      pipe_sampler_view_reference(&llvmpipe->fragment_sampler_views[i], view);
    }
 
-   llvmpipe->num_textures = num;
+   llvmpipe->num_fragment_sampler_views = num;
 
-   llvmpipe->dirty |= LP_NEW_TEXTURE;
+   llvmpipe->dirty |= LP_NEW_SAMPLER_VIEW;
 }
 
 
 void
-llvmpipe_set_vertex_sampler_textures(struct pipe_context *pipe,
-                                     unsigned num_textures,
-                                     struct pipe_texture **textures)
+llvmpipe_set_vertex_sampler_views(struct pipe_context *pipe,
+                                  unsigned num,
+                                  struct pipe_sampler_view **views)
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
    uint i;
 
-   assert(num_textures <= PIPE_MAX_VERTEX_SAMPLERS);
+   assert(num <= PIPE_MAX_VERTEX_SAMPLERS);
 
    /* Check for no-op */
-   if (num_textures == llvmpipe->num_vertex_textures &&
-       !memcmp(llvmpipe->vertex_textures, textures, num_textures * sizeof(struct pipe_texture *))) {
+   if (num == llvmpipe->num_vertex_sampler_views &&
+       !memcmp(llvmpipe->vertex_sampler_views, views, num * sizeof(struct pipe_sampler_view *))) {
       return;
    }
 
    draw_flush(llvmpipe->draw);
 
    for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
-      struct pipe_texture *tex = i < num_textures ? textures[i] : NULL;
+      struct pipe_sampler_view *view = i < num ? views[i] : NULL;
 
-      pipe_texture_reference(&llvmpipe->vertex_textures[i], tex);
+      pipe_sampler_view_reference(&llvmpipe->vertex_sampler_views[i], view);
    }
 
-   llvmpipe->num_vertex_textures = num_textures;
+   llvmpipe->num_vertex_sampler_views = num;
 
-   llvmpipe->dirty |= LP_NEW_TEXTURE;
+   llvmpipe->dirty |= LP_NEW_SAMPLER_VIEW;
+}
+
+
+struct pipe_sampler_view *
+llvmpipe_create_sampler_view(struct pipe_context *pipe,
+                            struct pipe_texture *texture,
+                            const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+
+void
+llvmpipe_sampler_view_destroy(struct pipe_context *pipe,
+                              struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
 }
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_test.h b/src/gallium/drivers/llvmpipe/lp_test.h
index 1df98978988..338a04a4878 100644
--- a/src/gallium/drivers/llvmpipe/lp_test.h
+++ b/src/gallium/drivers/llvmpipe/lp_test.h
@@ -41,7 +41,7 @@
 #include <stdio.h>
 #include <float.h>
 
-#include "os/os_llvm.h"
+#include "gallivm/lp_bld.h"
 #include <llvm-c/Analysis.h>
 #include <llvm-c/ExecutionEngine.h>
 #include <llvm-c/Target.h>
diff --git a/src/gallium/drivers/llvmpipe/lp_test_format.c b/src/gallium/drivers/llvmpipe/lp_test_format.c
index 2c4d7fb6e14..fb595893bd0 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_format.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_format.c
@@ -29,7 +29,7 @@
 #include <stdlib.h>
 #include <stdio.h>
 
-#include "os/os_llvm.h"
+#include "gallivm/lp_bld.h"
 #include <llvm-c/Analysis.h>
 #include <llvm-c/ExecutionEngine.h>
 #include <llvm-c/Target.h>
diff --git a/src/gallium/drivers/llvmpipe/lp_test_printf.c b/src/gallium/drivers/llvmpipe/lp_test_printf.c
new file mode 100644
index 00000000000..e5e5925012a
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_test_printf.c
@@ -0,0 +1,162 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMware, Inc.
+ * 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 VMWARE 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.
+ *
+ **************************************************************************/
+
+
+#include <stdlib.h>
+#include <stdio.h>
+
+#include "gallivm/lp_bld.h"
+#include "gallivm/lp_bld_printf.h"
+
+#include <llvm-c/Analysis.h>
+#include <llvm-c/ExecutionEngine.h>
+#include <llvm-c/Target.h>
+#include <llvm-c/Transforms/Scalar.h>
+
+#include "lp_test.h"
+
+
+struct printf_test_case {
+};
+
+void
+write_tsv_header(FILE *fp)
+{
+   fprintf(fp,
+           "result\t"
+           "format\n");
+
+   fflush(fp);
+}
+
+
+
+typedef void (*test_printf_t)(int i);
+
+static LLVMValueRef
+add_printf_test(LLVMModuleRef module)
+{
+   LLVMTypeRef args[1] = { LLVMIntType(32) };
+   LLVMValueRef func = LLVMAddFunction(module, "test_printf", LLVMFunctionType(LLVMVoidType(), args, 1, 0));
+   LLVMBuilderRef builder = LLVMCreateBuilder();
+   LLVMBasicBlockRef block = LLVMAppendBasicBlock(func, "entry");
+
+   LLVMSetFunctionCallConv(func, LLVMCCallConv);
+
+   LLVMPositionBuilderAtEnd(builder, block);
+   lp_build_printf(builder, "hello, world\n");
+   lp_build_printf(builder, "print 5 6: %d %d\n", LLVMConstInt(LLVMInt32Type(), 5, 0),
+				LLVMConstInt(LLVMInt32Type(), 6, 0));
+   LLVMBuildRetVoid(builder);
+   LLVMDisposeBuilder(builder);
+   return func;
+}
+
+
+PIPE_ALIGN_STACK
+static boolean
+test_printf(unsigned verbose, FILE *fp, const struct printf_test_case *testcase)
+{
+   LLVMModuleRef module = NULL;
+   LLVMValueRef test = NULL;
+   LLVMExecutionEngineRef engine = NULL;
+   LLVMModuleProviderRef provider = NULL;
+   LLVMPassManagerRef pass = NULL;
+   char *error = NULL;
+   test_printf_t test_printf;
+   float unpacked[4];
+   unsigned packed;
+   boolean success = TRUE;
+
+   module = LLVMModuleCreateWithName("test");
+
+   test = add_printf_test(module);
+
+   if(LLVMVerifyModule(module, LLVMPrintMessageAction, &error)) {
+      LLVMDumpModule(module);
+      abort();
+   }
+   LLVMDisposeMessage(error);
+
+   provider = LLVMCreateModuleProviderForExistingModule(module);
+   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
+      fprintf(stderr, "%s\n", error);
+      LLVMDisposeMessage(error);
+      abort();
+   }
+
+#if 0
+   pass = LLVMCreatePassManager();
+   LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
+   /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
+    * but there are more on SVN. */
+   LLVMAddConstantPropagationPass(pass);
+   LLVMAddInstructionCombiningPass(pass);
+   LLVMAddPromoteMemoryToRegisterPass(pass);
+   LLVMAddGVNPass(pass);
+   LLVMAddCFGSimplificationPass(pass);
+   LLVMRunPassManager(pass, module);
+#else
+   (void)pass;
+#endif
+
+   test_printf = (test_printf_t)LLVMGetPointerToGlobal(engine, test);
+
+   memset(unpacked, 0, sizeof unpacked);
+   packed = 0;
+
+
+   // LLVMDumpModule(module);
+
+   test_printf(0);
+
+   LLVMFreeMachineCodeForFunction(engine, test);
+
+   LLVMDisposeExecutionEngine(engine);
+   if(pass)
+      LLVMDisposePassManager(pass);
+
+   return success;
+}
+
+
+boolean
+test_all(unsigned verbose, FILE *fp)
+{
+   bool success = TRUE;
+
+   test_printf(verbose, fp, NULL);
+
+   return success;
+}
+
+
+boolean
+test_some(unsigned verbose, FILE *fp, unsigned long n)
+{
+   return test_all(verbose, fp);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample.h b/src/gallium/drivers/llvmpipe/lp_tex_sample.h
index 799df182b6a..1228a831f3b 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_sample.h
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample.h
@@ -29,7 +29,7 @@
 #define LP_TEX_SAMPLE_H
 
 
-#include "os/os_llvm.h"
+#include "gallivm/lp_bld.h"
 
 
 struct lp_sampler_static_state;
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
index 662508af61a..4715cfe4f62 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
@@ -105,7 +105,7 @@ lp_llvm_texture_member(struct lp_sampler_dynamic_state *base,
    /* context[0] */
    indices[0] = LLVMConstInt(LLVMInt32Type(), 0, 0);
    /* context[0].textures */
-   indices[1] = LLVMConstInt(LLVMInt32Type(), LP_JIT_CONTEXT_TEXTURES_INDEX, 0);
+   indices[1] = LLVMConstInt(LLVMInt32Type(), LP_JIT_CTX_TEXTURES, 0);
    /* context[0].textures[unit] */
    indices[2] = LLVMConstInt(LLVMInt32Type(), unit, 0);
    /* context[0].textures[unit].member */
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index 9a85a42897d..8137f29af52 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -103,6 +103,7 @@ llvmpipe_displaytarget_layout(struct llvmpipe_screen *screen,
    unsigned height = align(lpt->base.height0, TILE_SIZE);
 
    lpt->dt = winsys->displaytarget_create(winsys,
+                                          lpt->base.tex_usage,
                                           lpt->base.format,
                                           width, height,
                                           16,
@@ -250,6 +251,51 @@ llvmpipe_texture_unmap(struct pipe_texture *texture,
 }
 
 
+static struct pipe_texture *
+llvmpipe_texture_from_handle(struct pipe_screen *screen,
+                             const struct pipe_texture *template,
+                             struct winsys_handle *whandle)
+{
+   struct sw_winsys *winsys = llvmpipe_screen(screen)->winsys;
+   struct llvmpipe_texture *lpt = CALLOC_STRUCT(llvmpipe_texture);
+   if (!lpt)
+      return NULL;
+
+   lpt->base = *template;
+   pipe_reference_init(&lpt->base.reference, 1);
+   lpt->base.screen = screen;
+
+   lpt->dt = winsys->displaytarget_from_handle(winsys,
+                                               template,
+                                               whandle,
+                                               &lpt->stride[0]);
+   if (!lpt->dt)
+      goto fail;
+
+   return &lpt->base;
+
+ fail:
+   FREE(lpt);
+   return NULL;
+}
+
+
+static boolean
+llvmpipe_texture_get_handle(struct pipe_screen *screen,
+                            struct pipe_texture *pt,
+                            struct winsys_handle *whandle)
+{
+   struct sw_winsys *winsys = llvmpipe_screen(screen)->winsys;
+   struct llvmpipe_texture *lpt = llvmpipe_texture(pt);
+
+   assert(lpt->dt);
+   if (!lpt->dt)
+      return FALSE;
+
+   return winsys->displaytarget_get_handle(winsys, lpt->dt, whandle);
+}
+
+
 static struct pipe_surface *
 llvmpipe_get_tex_surface(struct pipe_screen *screen,
                          struct pipe_texture *pt,
@@ -418,6 +464,8 @@ llvmpipe_init_screen_texture_funcs(struct pipe_screen *screen)
 {
    screen->texture_create = llvmpipe_texture_create;
    screen->texture_destroy = llvmpipe_texture_destroy;
+   screen->texture_from_handle = llvmpipe_texture_from_handle;
+   screen->texture_get_handle = llvmpipe_texture_get_handle;
 
    screen->get_tex_surface = llvmpipe_get_tex_surface;
    screen->tex_surface_destroy = llvmpipe_tex_surface_destroy;
diff --git a/src/gallium/drivers/nv50/nv50_clear.c b/src/gallium/drivers/nv50/nv50_clear.c
index 8afc95c9fc6..5447904e9ca 100644
--- a/src/gallium/drivers/nv50/nv50_clear.c
+++ b/src/gallium/drivers/nv50/nv50_clear.c
@@ -35,7 +35,10 @@ nv50_clear(struct pipe_context *pipe, unsigned buffers,
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct pipe_framebuffer_state *fb = &nv50->framebuffer;
 	unsigned mode = 0, i;
+	const unsigned dirty = nv50->dirty;
 
+	/* don't need NEW_BLEND, NV50TCL_COLOR_MASK doesn't affect CLEAR_BUFFERS */
+	nv50->dirty &= NV50_NEW_FRAMEBUFFER | NV50_NEW_SCISSOR;
 	if (!nv50_state_validate(nv50, 64))
 		return;
 
@@ -64,5 +67,6 @@ nv50_clear(struct pipe_context *pipe, unsigned buffers,
 		BEGIN_RING(chan, tesla, NV50TCL_CLEAR_BUFFERS, 1);
 		OUT_RING  (chan, (i << 6) | 0x3c);
 	}
+	nv50->dirty = dirty;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
index 1743f6fb394..bc7831d9aca 100644
--- a/src/gallium/drivers/nv50/nv50_context.h
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -72,12 +72,23 @@ struct nv50_sampler_stateobj {
 	unsigned tsc[8];
 };
 
+struct nv50_sampler_view {
+	struct pipe_sampler_view pipe;
+	uint32_t tic[8];
+};
+
 struct nv50_vtxelt_stateobj {
 	struct pipe_vertex_element pipe[16];
 	unsigned num_elements;
 	uint32_t hw[16];
 };
 
+static INLINE struct nv50_sampler_view *
+nv50_sampler_view(struct pipe_sampler_view *view)
+{
+	return (struct nv50_sampler_view *)view;
+}
+
 static INLINE unsigned
 get_tile_height(uint32_t tile_mode)
 {
@@ -126,7 +137,7 @@ struct nv50_state {
 	struct nouveau_stateobj *hw[64];
 	uint64_t hw_dirty;
 
-	unsigned miptree_nr[PIPE_SHADER_TYPES];
+	unsigned sampler_view_nr[3];
 	struct nouveau_stateobj *vtxbuf;
 	struct nouveau_stateobj *vtxattr;
 	unsigned vtxelt_nr;
@@ -158,10 +169,10 @@ struct nv50_context {
 	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
 	unsigned vtxbuf_nr;
 	struct nv50_vtxelt_stateobj *vtxelt;
-	struct nv50_sampler_stateobj *sampler[PIPE_SHADER_TYPES][PIPE_MAX_SAMPLERS];
-	unsigned sampler_nr[PIPE_SHADER_TYPES];
-	struct nv50_miptree *miptree[PIPE_SHADER_TYPES][PIPE_MAX_SAMPLERS];
-	unsigned miptree_nr[PIPE_SHADER_TYPES];
+	struct nv50_sampler_stateobj *sampler[3][PIPE_MAX_SAMPLERS];
+	unsigned sampler_nr[3];
+	struct pipe_sampler_view *sampler_views[3][PIPE_MAX_SAMPLERS];
+	unsigned sampler_view_nr[3];
 
 	unsigned vbo_fifo;
 };
@@ -243,6 +254,7 @@ extern void nv50_so_init_sifc(struct nv50_context *nv50,
 			      unsigned offset, unsigned size);
 
 /* nv50_tex.c */
+extern boolean nv50_tex_construct(struct nv50_sampler_view *view);
 extern void nv50_tex_relocs(struct nv50_context *);
 extern struct nouveau_stateobj *nv50_tex_validate(struct nv50_context *);
 
diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
index c857816b31a..d7e06c93272 100644
--- a/src/gallium/drivers/nv50/nv50_program.c
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -3762,7 +3762,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 			p->cfg.in[n].hw = rid = aid;
 			i = p->cfg.in[n].id;
 
-			if (p->info.input_semantic_name[n] ==
+			if (p->info.input_semantic_name[i] ==
 			    TGSI_SEMANTIC_FACE) {
 				load_frontfacing(pc, &pc->attr[i * 4]);
 				continue;
diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c
index adf0d3b3741..1a4606d9e25 100644
--- a/src/gallium/drivers/nv50/nv50_screen.c
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -109,7 +109,7 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 	case PIPE_CAP_TWO_SIDED_STENCIL:
 		return 1;
 	case PIPE_CAP_GLSL:
-		return 0;
+		return 1;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
 		return 1;
 	case PIPE_CAP_POINT_SPRITE:
@@ -190,8 +190,6 @@ nv50_screen_destroy(struct pipe_screen *pscreen)
 		nouveau_bo_ref(NULL, &screen->tic);
 	if (screen->tsc)
 		nouveau_bo_ref(NULL, &screen->tsc);
-	if (screen->static_init)
-		so_ref(NULL, &screen->static_init);
 
 	nouveau_notifier_free(&screen->sync);
 	nouveau_grobj_free(&screen->tesla);
@@ -204,16 +202,65 @@ nv50_screen_destroy(struct pipe_screen *pscreen)
 	FREE(screen);
 }
 
+#define BGN_RELOC(ch, bo, gr, m, n, fl) \
+   OUT_RELOC(ch, bo, (n << 18) | (gr->subc << 13) | m, fl, 0, 0)
+
+void
+nv50_screen_relocs(struct nv50_screen *screen)
+{
+	struct nouveau_channel *chan = screen->base.channel;
+	struct nouveau_grobj *tesla = screen->tesla;
+	unsigned i;
+	const unsigned rl = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_DUMMY;
+
+	MARK_RING (chan, 28, 26);
+
+	/* cause grobj autobind */
+	BEGIN_RING(chan, tesla, 0x0100, 1);
+	OUT_RING  (chan, 0);
+
+	BGN_RELOC (chan, screen->tic, tesla, NV50TCL_TIC_ADDRESS_HIGH, 2, rl);
+	OUT_RELOCh(chan, screen->tic, 0, rl);
+	OUT_RELOCl(chan, screen->tic, 0, rl);
+
+	BGN_RELOC (chan, screen->tsc, tesla, NV50TCL_TSC_ADDRESS_HIGH, 2, rl);
+	OUT_RELOCh(chan, screen->tsc, 0, rl);
+	OUT_RELOCl(chan, screen->tsc, 0, rl);
+
+	BGN_RELOC (chan, screen->constbuf_misc[0],
+		   tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3, rl);
+	OUT_RELOCh(chan, screen->constbuf_misc[0], 0, rl);
+	OUT_RELOCl(chan, screen->constbuf_misc[0], 0, rl);
+	OUT_RELOC (chan, screen->constbuf_misc[0],
+		   (NV50_CB_PMISC << 16) | 0x0200, rl, 0, 0);
+
+	BGN_RELOC (chan, screen->constbuf_misc[0],
+		   tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3, rl);
+	OUT_RELOCh(chan, screen->constbuf_misc[0], 0x200, rl);
+	OUT_RELOCl(chan, screen->constbuf_misc[0], 0x200, rl);
+	OUT_RELOC (chan, screen->constbuf_misc[0],
+		   (NV50_CB_AUX << 16) | 0x0200, rl, 0, 0);
+
+	for (i = 0; i < 3; ++i) {
+		BGN_RELOC (chan, screen->constbuf_parm[i],
+			   tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3, rl);
+		OUT_RELOCh(chan, screen->constbuf_parm[i], 0, rl);
+		OUT_RELOCl(chan, screen->constbuf_parm[i], 0, rl);
+		OUT_RELOC (chan, screen->constbuf_parm[i],
+			   ((NV50_CB_PVP + i) << 16) | 0x0800, rl, 0, 0);
+	}
+}
+
 struct pipe_screen *
 nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv50_screen *screen = CALLOC_STRUCT(nv50_screen);
 	struct nouveau_channel *chan;
 	struct pipe_screen *pscreen;
-	struct nouveau_stateobj *so;
 	unsigned chipset = dev->chipset;
 	unsigned tesla_class = 0;
 	int ret, i;
+	const unsigned rl = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD;
 
 	if (!screen)
 		return NULL;
@@ -296,64 +343,58 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 	}
 
 	/* Static M2MF init */
-	so = so_new(1, 3, 0);
-	so_method(so, screen->m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 3);
-	so_data  (so, screen->sync->handle);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->vram->handle);
-	so_emit(chan, so);
-	so_ref (NULL, &so);
+	BEGIN_RING(chan, screen->m2mf,
+		   NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 3);
+	OUT_RING  (chan, screen->sync->handle);
+	OUT_RING  (chan, chan->vram->handle);
+	OUT_RING  (chan, chan->vram->handle);
 
 	/* Static 2D init */
-	so = so_new(4, 7, 0);
-	so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4);
-	so_data  (so, screen->sync->handle);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->vram->handle);
-	so_data  (so, chan->vram->handle);
-	so_method(so, screen->eng2d, NV50_2D_OPERATION, 1);
-	so_data  (so, NV50_2D_OPERATION_SRCCOPY);
-	so_method(so, screen->eng2d, NV50_2D_CLIP_ENABLE, 1);
-	so_data  (so, 0);
-	so_method(so, screen->eng2d, 0x0888, 1);
-	so_data  (so, 1);
-	so_emit(chan, so);
-	so_ref(NULL, &so);
+	BEGIN_RING(chan, screen->eng2d, NV50_2D_DMA_NOTIFY, 4);
+	OUT_RING  (chan, screen->sync->handle);
+	OUT_RING  (chan, chan->vram->handle);
+	OUT_RING  (chan, chan->vram->handle);
+	OUT_RING  (chan, chan->vram->handle);
+	BEGIN_RING(chan, screen->eng2d, NV50_2D_OPERATION, 1);
+	OUT_RING  (chan, NV50_2D_OPERATION_SRCCOPY);
+	BEGIN_RING(chan, screen->eng2d, NV50_2D_CLIP_ENABLE, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, screen->eng2d, 0x0888, 1);
+	OUT_RING  (chan, 1);
 
 	/* Static tesla init */
-	so = so_new(47, 95, 24);
-
-	so_method(so, screen->tesla, NV50TCL_COND_MODE, 1);
-	so_data  (so, NV50TCL_COND_MODE_ALWAYS);
-	so_method(so, screen->tesla, NV50TCL_DMA_NOTIFY, 1);
-	so_data  (so, screen->sync->handle);
-	so_method(so, screen->tesla, NV50TCL_DMA_ZETA, 11);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_COND_MODE, 1);
+	OUT_RING  (chan, NV50TCL_COND_MODE_ALWAYS);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_DMA_NOTIFY, 1);
+	OUT_RING  (chan, screen->sync->handle);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_DMA_ZETA, 11);
 	for (i = 0; i < 11; i++)
-		so_data(so, chan->vram->handle);
-	so_method(so, screen->tesla, NV50TCL_DMA_COLOR(0),
-				     NV50TCL_DMA_COLOR__SIZE);
+		OUT_RING  (chan, chan->vram->handle);
+	BEGIN_RING(chan, screen->tesla,
+		   NV50TCL_DMA_COLOR(0), NV50TCL_DMA_COLOR__SIZE);
 	for (i = 0; i < NV50TCL_DMA_COLOR__SIZE; i++)
-		so_data(so, chan->vram->handle);
-	so_method(so, screen->tesla, NV50TCL_RT_CONTROL, 1);
-	so_data  (so, 1);
+		OUT_RING  (chan, chan->vram->handle);
+
+	BEGIN_RING(chan, screen->tesla, NV50TCL_RT_CONTROL, 1);
+	OUT_RING  (chan, 1);
 
 	/* activate all 32 lanes (threads) in a warp */
-	so_method(so, screen->tesla, NV50TCL_WARP_HALVES, 1);
-	so_data  (so, 0x2);
-	so_method(so, screen->tesla, 0x1400, 1);
-	so_data  (so, 0xf);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_WARP_HALVES, 1);
+	OUT_RING  (chan, 2);
+	BEGIN_RING(chan, screen->tesla, 0x1400, 1);
+	OUT_RING  (chan, 0xf);
 
 	/* max TIC (bits 4:8) & TSC (ignored) bindings, per program type */
 	for (i = 0; i < 3; ++i) {
-		so_method(so, screen->tesla, NV50TCL_TEX_LIMITS(i), 1);
-		so_data  (so, 0x54);
+		BEGIN_RING(chan, screen->tesla, NV50TCL_TEX_LIMITS(i), 1);
+		OUT_RING  (chan, 0x54);
 	}
 
 	/* origin is top left (set to 1 for bottom left) */
-	so_method(so, screen->tesla, NV50TCL_Y_ORIGIN_BOTTOM, 1);
-	so_data  (so, 0);
-	so_method(so, screen->tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1);
-	so_data  (so, 8);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_Y_ORIGIN_BOTTOM, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1);
+	OUT_RING  (chan, 8);
 
 	/* constant buffers for immediates and VP/FP parameters */
 	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, (32 * 4) * 4,
@@ -362,6 +403,14 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BEGIN_RING(chan, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+	OUT_RELOCh(chan, screen->constbuf_misc[0], 0, rl);
+	OUT_RELOCl(chan, screen->constbuf_misc[0], 0, rl);
+	OUT_RING  (chan, (NV50_CB_PMISC << 16) | 0x0200);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+	OUT_RELOCh(chan, screen->constbuf_misc[0], 0x200, rl);
+	OUT_RELOCl(chan, screen->constbuf_misc[0], 0x200, rl);
+	OUT_RING  (chan, (NV50_CB_AUX << 16) | 0x0200);
 
 	for (i = 0; i < 3; i++) {
 		ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, (256 * 4) * 4,
@@ -370,6 +419,10 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 			nv50_screen_destroy(pscreen);
 			return NULL;
 		}
+		BEGIN_RING(chan, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+		OUT_RELOCh(chan, screen->constbuf_parm[i], 0, rl);
+		OUT_RELOCl(chan, screen->constbuf_parm[i], 0, rl);
+		OUT_RING  (chan, ((NV50_CB_PVP + i) << 16) | 0x0800);
 	}
 
 	if (nouveau_resource_init(&screen->immd_heap[0], 0, 128) ||
@@ -381,118 +434,67 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 		return NULL;
 	}
 
-	/*
-	// map constant buffers:
-	//  B = buffer ID (maybe more than 1 byte)
-	//  N = CB index used in shader instruction
-	//  P = program type (0 = VP, 2 = GP, 3 = FP)
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x000BBNP1);
-	*/
-
-	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_PMISC << 16) | 0x00000200);
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000001 | (NV50_CB_PMISC << 12));
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000021 | (NV50_CB_PMISC << 12));
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000031 | (NV50_CB_PMISC << 12));
-
-	/* bind auxiliary constbuf to immediate data bo */
-	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->constbuf_misc[0], (128 * 4) * 4,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf_misc[0], (128 * 4) * 4,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_AUX << 16) | 0x00000200);
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000201 | (NV50_CB_AUX << 12));
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000221 | (NV50_CB_AUX << 12));
-
-	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->constbuf_parm[PIPE_SHADER_VERTEX], 0,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf_parm[PIPE_SHADER_VERTEX], 0,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_PVP << 16) | 0x00000800);
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000101 | (NV50_CB_PVP << 12));
-
-	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->constbuf_parm[PIPE_SHADER_GEOMETRY], 0,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf_parm[PIPE_SHADER_GEOMETRY], 0,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_PGP << 16) | 0x00000800);
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000121 | (NV50_CB_PGP << 12));
-
-	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->constbuf_parm[PIPE_SHADER_FRAGMENT], 0,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf_parm[PIPE_SHADER_FRAGMENT], 0,
-		  NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_PFP << 16) | 0x00000800);
-	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
-	so_data  (so, 0x00000131 | (NV50_CB_PFP << 12));
-
-	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, PIPE_SHADER_TYPES*32*32,
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 3 * 32 * (8 * 4),
 			     &screen->tic);
 	if (ret) {
 		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BEGIN_RING(chan, screen->tesla, NV50TCL_TIC_ADDRESS_HIGH, 3);
+	OUT_RELOCh(chan, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RELOCl(chan, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RING  (chan, 3 * 32 - 1);
 
-	so_method(so, screen->tesla, NV50TCL_TIC_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, PIPE_SHADER_TYPES * 32 - 1);
-
-	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, PIPE_SHADER_TYPES*32*32,
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 3 * 32 * (8 * 4),
 			     &screen->tsc);
 	if (ret) {
 		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
-
-	so_method(so, screen->tesla, NV50TCL_TSC_ADDRESS_HIGH, 3);
-	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, 0x00000000); /* ignored if TSC_LINKED (0x1234) = 1 */
-
+	BEGIN_RING(chan, screen->tesla, NV50TCL_TSC_ADDRESS_HIGH, 3);
+	OUT_RELOCh(chan, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RELOCl(chan, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RING  (chan, 0); /* ignored if TSC_LINKED (0x1234) == 1 */
+
+	/* map constant buffers:
+	 *  B = buffer ID (maybe more than 1 byte)
+	 *  N = CB index used in shader instruction
+	 *  P = program type (0 = VP, 2 = GP, 3 = FP)
+	 * SET_PROGRAM_CB = 0x000BBNP1
+	 */
+	BEGIN_RING_NI(chan, screen->tesla, NV50TCL_SET_PROGRAM_CB, 8);
+	/* bind immediate buffer */
+	OUT_RING  (chan, 0x001 | (NV50_CB_PMISC << 12));
+	OUT_RING  (chan, 0x021 | (NV50_CB_PMISC << 12));
+	OUT_RING  (chan, 0x031 | (NV50_CB_PMISC << 12));
+	/* bind auxiliary constbuf to immediate data bo */
+	OUT_RING  (chan, 0x201 | (NV50_CB_AUX << 12));
+	OUT_RING  (chan, 0x221 | (NV50_CB_AUX << 12));
+	/* bind parameter buffers */
+	OUT_RING  (chan, 0x101 | (NV50_CB_PVP << 12));
+	OUT_RING  (chan, 0x121 | (NV50_CB_PGP << 12));
+	OUT_RING  (chan, 0x131 | (NV50_CB_PFP << 12));
 
 	/* Vertex array limits - max them out */
 	for (i = 0; i < 16; i++) {
-		so_method(so, screen->tesla, NV50TCL_VERTEX_ARRAY_LIMIT_HIGH(i), 2);
-		so_data  (so, 0x000000ff);
-		so_data  (so, 0xffffffff);
+		BEGIN_RING(chan, screen->tesla,
+			   NV50TCL_VERTEX_ARRAY_LIMIT_HIGH(i), 2);
+		OUT_RING  (chan, 0x000000ff);
+		OUT_RING  (chan, 0xffffffff);
 	}
 
-	so_method(so, screen->tesla, NV50TCL_DEPTH_RANGE_NEAR(0), 2);
-	so_data  (so, fui(0.0));
-	so_data  (so, fui(1.0));
+	BEGIN_RING(chan, screen->tesla, NV50TCL_DEPTH_RANGE_NEAR(0), 2);
+	OUT_RINGf (chan, 0.0f);
+	OUT_RINGf (chan, 1.0f);
 
 	/* no dynamic combination of TIC & TSC entries => only BIND_TIC used */
-	so_method(so, screen->tesla, NV50TCL_LINKED_TSC, 1);
-	so_data  (so, 1);
+	BEGIN_RING(chan, screen->tesla, NV50TCL_LINKED_TSC, 1);
+	OUT_RING  (chan, 1);
 
-	so_method(so, screen->tesla, NV50TCL_EDGEFLAG_ENABLE, 1);
-	so_data  (so, 1); /* default edgeflag to TRUE */
+	BEGIN_RING(chan, screen->tesla, NV50TCL_EDGEFLAG_ENABLE, 1);
+	OUT_RING  (chan, 1); /* default edgeflag to TRUE */
 
-	so_emit(chan, so);
-	so_ref (so, &screen->static_init);
-	so_ref (NULL, &so);
-	nouveau_pushbuf_flush(chan, 0);
+	FIRE_RING (chan);
 
 	screen->force_push = debug_get_bool_option("NV50_ALWAYS_PUSH", FALSE);
 	return pscreen;
diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h
index ec19ea655b1..15bd4eed399 100644
--- a/src/gallium/drivers/nv50/nv50_screen.h
+++ b/src/gallium/drivers/nv50/nv50_screen.h
@@ -27,8 +27,6 @@ struct nv50_screen {
 	struct nouveau_bo *tic;
 	struct nouveau_bo *tsc;
 
-	struct nouveau_stateobj *static_init;
-
 	boolean force_push;
 };
 
@@ -38,4 +36,6 @@ nv50_screen(struct pipe_screen *screen)
 	return (struct nv50_screen *)screen;
 }
 
+extern void nv50_screen_relocs(struct nv50_screen *);
+
 #endif
diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c
index b0e5552eff4..c1628089285 100644
--- a/src/gallium/drivers/nv50/nv50_state.c
+++ b/src/gallium/drivers/nv50/nv50_state.c
@@ -238,6 +238,9 @@ nv50_sampler_state_create(struct pipe_context *pipe,
 	return (void *)sso;
 }
 
+/* type == 0 for VPs, 1 for GPs, 2 for FPs, which is how the
+ * relevant tesla methods are indexed (NV50TCL_BIND_TSC etc.)
+ */
 static INLINE void
 nv50_sampler_state_bind(struct pipe_context *pipe, unsigned type,
 			unsigned nr, void **sampler)
@@ -253,13 +256,13 @@ nv50_sampler_state_bind(struct pipe_context *pipe, unsigned type,
 static void
 nv50_vp_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **s)
 {
-	nv50_sampler_state_bind(pipe, PIPE_SHADER_VERTEX, nr, s);
+	nv50_sampler_state_bind(pipe, 0, nr, s);
 }
 
 static void
 nv50_fp_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **s)
 {
-	nv50_sampler_state_bind(pipe, PIPE_SHADER_FRAGMENT, nr, s);
+	nv50_sampler_state_bind(pipe, 2, nr, s);
 }
 
 static void
@@ -269,35 +272,69 @@ nv50_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
 }
 
 static INLINE void
-nv50_set_sampler_texture(struct pipe_context *pipe, unsigned type,
-			 unsigned nr, struct pipe_texture **pt)
+nv50_set_sampler_views(struct pipe_context *pipe, unsigned p,
+		       unsigned nr,
+		       struct pipe_sampler_view **views)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
 	unsigned i;
 
 	for (i = 0; i < nr; i++)
-		pipe_texture_reference((void *)&nv50->miptree[type][i], pt[i]);
-	for (i = nr; i < nv50->miptree_nr[type]; i++)
-		pipe_texture_reference((void *)&nv50->miptree[type][i], NULL);
+		pipe_sampler_view_reference(&nv50->sampler_views[p][i],
+					    views[i]);
+
+	for (i = nr; i < nv50->sampler_view_nr[p]; i++)
+		pipe_sampler_view_reference(&nv50->sampler_views[p][i], NULL);
 
-	nv50->miptree_nr[type] = nr;
+	nv50->sampler_view_nr[p] = nr;
 	nv50->dirty |= NV50_NEW_TEXTURE;
 }
 
 static void
-nv50_set_vp_sampler_textures(struct pipe_context *pipe,
-			     unsigned nr, struct pipe_texture **pt)
+nv50_set_vp_sampler_views(struct pipe_context *pipe,
+			  unsigned nr,
+			  struct pipe_sampler_view **views)
 {
-	nv50_set_sampler_texture(pipe, PIPE_SHADER_VERTEX, nr, pt);
+	nv50_set_sampler_views(pipe, 0, nr, views);
 }
 
 static void
-nv50_set_fp_sampler_textures(struct pipe_context *pipe,
-			     unsigned nr, struct pipe_texture **pt)
+nv50_set_fp_sampler_views(struct pipe_context *pipe,
+			  unsigned nr,
+			  struct pipe_sampler_view **views)
 {
-	nv50_set_sampler_texture(pipe, PIPE_SHADER_FRAGMENT, nr, pt);
+	nv50_set_sampler_views(pipe, 2, nr, views);
 }
 
+static void
+nv50_sampler_view_destroy(struct pipe_context *pipe,
+			  struct pipe_sampler_view *view)
+{
+	pipe_texture_reference(&view->texture, NULL);
+	FREE(nv50_sampler_view(view));
+}
+
+static struct pipe_sampler_view *
+nv50_create_sampler_view(struct pipe_context *pipe,
+			 struct pipe_texture *texture,
+			 const struct pipe_sampler_view *templ)
+{
+	struct nv50_sampler_view *view = CALLOC_STRUCT(nv50_sampler_view);
+
+	view->pipe = *templ;
+	view->pipe.reference.count = 1;
+	view->pipe.texture = NULL;
+	pipe_texture_reference(&view->pipe.texture, texture);
+	view->pipe.context = pipe;
+
+	if (!nv50_tex_construct(view)) {
+		nv50_sampler_view_destroy(pipe, &view->pipe);
+		return NULL;
+	}
+	return &view->pipe;
+}
+
+
 static void *
 nv50_rasterizer_state_create(struct pipe_context *pipe,
 			     const struct pipe_rasterizer_state *cso)
@@ -765,8 +802,10 @@ nv50_init_state_functions(struct nv50_context *nv50)
 	nv50->pipe.delete_sampler_state = nv50_sampler_state_delete;
 	nv50->pipe.bind_fragment_sampler_states = nv50_fp_sampler_state_bind;
 	nv50->pipe.bind_vertex_sampler_states   = nv50_vp_sampler_state_bind;
-	nv50->pipe.set_fragment_sampler_textures = nv50_set_fp_sampler_textures;
-	nv50->pipe.set_vertex_sampler_textures   = nv50_set_vp_sampler_textures;
+	nv50->pipe.set_fragment_sampler_views = nv50_set_fp_sampler_views;
+	nv50->pipe.set_vertex_sampler_views   = nv50_set_vp_sampler_views;
+	nv50->pipe.create_sampler_view = nv50_create_sampler_view;
+	nv50->pipe.sampler_view_destroy = nv50_sampler_view_destroy;
 
 	nv50->pipe.create_rasterizer_state = nv50_rasterizer_state_create;
 	nv50->pipe.bind_rasterizer_state = nv50_rasterizer_state_bind;
diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c
index 2c8e7ca7982..b7e355283c8 100644
--- a/src/gallium/drivers/nv50/nv50_state_validate.c
+++ b/src/gallium/drivers/nv50/nv50_state_validate.c
@@ -310,15 +310,13 @@ validate_sampler(struct nv50_context *nv50)
 	struct nouveau_stateobj *so;
 	unsigned nr = 0, i;
 
-	for (i = 0; i < PIPE_SHADER_TYPES; ++i)
+	for (i = 0; i < 3; ++i)
 		nr += nv50->sampler_nr[i];
 
-	so = so_new(1 + 5 * PIPE_SHADER_TYPES,
-		    1 + 19 * PIPE_SHADER_TYPES + nr * 8,
-		    PIPE_SHADER_TYPES * 2);
+	so = so_new(1 + 5 * 3, 1 + 19 * 3 + nr * 8, 3 * 2);
 
-	nv50_validate_samplers(nv50, so, PIPE_SHADER_VERTEX);
-	nv50_validate_samplers(nv50, so, PIPE_SHADER_FRAGMENT);
+	nv50_validate_samplers(nv50, so, 0); /* VP */
+	nv50_validate_samplers(nv50, so, 2); /* FP */
 
 	so_method(so, tesla, 0x1334, 1); /* flush TSC */
 	so_data  (so, 0);
@@ -437,7 +435,7 @@ nv50_state_validate(struct nv50_context *nv50, unsigned wait_dwords)
 	so_emit_reloc_markers(chan, nv50->state.hw[3]); /* vp */
 	so_emit_reloc_markers(chan, nv50->state.hw[4]); /* fp */
 	so_emit_reloc_markers(chan, nv50->state.hw[17]); /* vb */
-	so_emit_reloc_markers(chan, nv50->screen->static_init);
+	nv50_screen_relocs(nv50->screen);
 
 	/* No idea.. */
 	BEGIN_RING(chan, tesla, 0x142c, 1);
diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c
index cabd148bc5b..6467c48a32a 100644
--- a/src/gallium/drivers/nv50/nv50_surface.c
+++ b/src/gallium/drivers/nv50/nv50_surface.c
@@ -28,6 +28,7 @@
 #include "util/u_inlines.h"
 
 #include "util/u_tile.h"
+#include "util/u_format.h"
 
 static INLINE int
 nv50_format(enum pipe_format format)
@@ -37,10 +38,35 @@ nv50_format(enum pipe_format format)
 		return NV50_2D_DST_FORMAT_A8R8G8B8_UNORM;
 	case PIPE_FORMAT_B8G8R8X8_UNORM:
 		return NV50_2D_DST_FORMAT_X8R8G8B8_UNORM;
+	case PIPE_FORMAT_B8G8R8A8_SRGB:
+		return NV50_2D_DST_FORMAT_A8R8G8B8_SRGB;
+	case PIPE_FORMAT_B8G8R8X8_SRGB:
+		return NV50_2D_DST_FORMAT_X8R8G8B8_SRGB;
 	case PIPE_FORMAT_B5G6R5_UNORM:
 		return NV50_2D_DST_FORMAT_R5G6B5_UNORM;
+	case PIPE_FORMAT_B5G5R5A1_UNORM:
+		return NV50_2D_DST_FORMAT_A1R5G5B5_UNORM;
 	case PIPE_FORMAT_A8_UNORM:
+	case PIPE_FORMAT_I8_UNORM:
+	case PIPE_FORMAT_L8_UNORM:
 		return NV50_2D_DST_FORMAT_R8_UNORM;
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+		return NV50_2D_DST_FORMAT_R32G32B32A32_FLOAT;
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+		return NV50_2D_DST_FORMAT_R32G32B32X32_FLOAT;
+	case PIPE_FORMAT_Z32_FLOAT:
+		return NV50_2D_DST_FORMAT_R32_FLOAT;
+
+	/* only because we require src format == dst format: */
+	case PIPE_FORMAT_R16G16_SNORM:
+	case PIPE_FORMAT_R16G16_UNORM:
+	case PIPE_FORMAT_S8Z24_UNORM:
+	case PIPE_FORMAT_Z24S8_UNORM:
+		return NV50_2D_DST_FORMAT_A8R8G8B8_UNORM;
+	case PIPE_FORMAT_L8A8_UNORM:
+	case PIPE_FORMAT_B4G4R4A4_UNORM:
+		return NV50_2D_DST_FORMAT_R16_UNORM;
+
 	default:
 		return -1;
 	}
@@ -57,8 +83,11 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
  	int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD);
 
  	format = nv50_format(ps->format);
- 	if (format < 0)
+	if (format < 0) {
+		NOUVEAU_ERR("invalid/unsupported surface format: %s\n",
+			    util_format_name(ps->format));
  		return 1;
+	}
 
  	if (!bo->tile_flags) {
 		MARK_RING (chan, 9, 2); /* flush on lack of space or relocs */
diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c
index 4c48b12cd87..85ab947c006 100644
--- a/src/gallium/drivers/nv50/nv50_tex.c
+++ b/src/gallium/drivers/nv50/nv50_tex.c
@@ -29,26 +29,16 @@
 #include "util/u_format.h"
 
 #define _MIXED(pf, t0, t1, t2, t3, cr, cg, cb, ca, f)		\
-{                                                       	\
-	PIPE_FORMAT_##pf,					\
+[PIPE_FORMAT_##pf] = (						\
 	NV50TIC_0_0_MAPR_##cr | NV50TIC_0_0_TYPER_##t0 |	\
 	NV50TIC_0_0_MAPG_##cg | NV50TIC_0_0_TYPEG_##t1 |	\
 	NV50TIC_0_0_MAPB_##cb | NV50TIC_0_0_TYPEB_##t2 |	\
 	NV50TIC_0_0_MAPA_##ca | NV50TIC_0_0_TYPEA_##t3 |	\
-	NV50TIC_0_0_FMT_##f					\
-}
+	NV50TIC_0_0_FMT_##f)
 
 #define _(pf, t, cr, cg, cb, ca, f) _MIXED(pf, t, t, t, t, cr, cg, cb, ca, f)
 
-struct nv50_texture_format {
-	enum pipe_format pf;
-	uint32_t hw;
-};
-
-#define NV50_TEX_FORMAT_LIST_SIZE \
-	(sizeof(nv50_tex_format_list) / sizeof(struct nv50_texture_format))
-
-static const struct nv50_texture_format nv50_tex_format_list[] =
+static const uint32_t nv50_texture_formats[PIPE_FORMAT_COUNT] =
 {
 	_(B8G8R8A8_UNORM, UNORM, C2, C1, C0, C3,  8_8_8_8),
 	_(B8G8R8A8_SRGB,  UNORM, C2, C1, C0, C3,  8_8_8_8),
@@ -60,10 +50,12 @@ static const struct nv50_texture_format nv50_tex_format_list[] =
 	_(B5G6R5_UNORM, UNORM, C2, C1, C0, ONE, 5_6_5),
 
 	_(L8_UNORM, UNORM, C0, C0, C0, ONE, 8),
+	_(L8_SRGB,  UNORM, C0, C0, C0, ONE, 8),
 	_(A8_UNORM, UNORM, ZERO, ZERO, ZERO, C0, 8),
 	_(I8_UNORM, UNORM, C0, C0, C0, C0, 8),
 
 	_(L8A8_UNORM, UNORM, C0, C0, C0, C1, 8_8),
+	_(L8A8_SRGB,  UNORM, C0, C0, C0, C1, 8_8),
 
 	_(DXT1_RGB, UNORM, C0, C1, C2, ONE, DXT1),
 	_(DXT1_RGBA, UNORM, C0, C1, C2, C3, DXT1),
@@ -81,117 +73,143 @@ static const struct nv50_texture_format nv50_tex_format_list[] =
 	_(R16G16_UNORM, UNORM, C0, C1, ZERO, ONE, 16_16),
 
 	_MIXED(Z32_FLOAT, FLOAT, UINT, UINT, UINT, C0, C0, C0, ONE, 32_DEPTH)
-
 };
 
 #undef _
 #undef _MIXED
 
-static int
-nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so,
-		   struct nv50_miptree *mt, int unit, unsigned p)
+static INLINE uint32_t
+nv50_tic_swizzle(uint32_t tc, unsigned swz)
+{
+	switch (swz) {
+	case PIPE_SWIZZLE_RED:
+		return (tc & NV50TIC_0_0_MAPR_MASK) >> NV50TIC_0_0_MAPR_SHIFT;
+	case PIPE_SWIZZLE_GREEN:
+		return (tc & NV50TIC_0_0_MAPG_MASK) >> NV50TIC_0_0_MAPG_SHIFT;
+	case PIPE_SWIZZLE_BLUE:
+		return (tc & NV50TIC_0_0_MAPB_MASK) >> NV50TIC_0_0_MAPB_SHIFT;
+	case PIPE_SWIZZLE_ALPHA:
+		return (tc & NV50TIC_0_0_MAPA_MASK) >> NV50TIC_0_0_MAPA_SHIFT;
+	case PIPE_SWIZZLE_ONE:
+		return 7;
+	case PIPE_SWIZZLE_ZERO:
+	default:
+		return 0;
+	}
+}
+
+boolean
+nv50_tex_construct(struct nv50_sampler_view *view)
 {
-	unsigned i;
-	uint32_t mode;
 	const struct util_format_description *desc;
+	struct nv50_miptree *mt = nv50_miptree(view->pipe.texture);
+	uint32_t swz[4], *tic = view->tic;
 
-	for (i = 0; i < NV50_TEX_FORMAT_LIST_SIZE; i++)
-		if (nv50_tex_format_list[i].pf == mt->base.base.format)
-			break;
-	if (i == NV50_TEX_FORMAT_LIST_SIZE)
-                return 1;
-
-	if (nv50->sampler[p][unit]->normalized)
-		mode = 0x50001000 | (1 << 31);
-	else {
-		mode = 0x50001000 | (7 << 14);
-		assert(mt->base.base.target == PIPE_TEXTURE_2D);
-	}
+	tic[0] = nv50_texture_formats[view->pipe.format];
 
-	mode |= ((mt->base.bo->tile_mode & 0x0f) << 22) |
-		((mt->base.bo->tile_mode & 0xf0) << 21);
+	swz[0] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_r);
+	swz[1] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_g);
+	swz[2] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_b);
+	swz[3] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_a);
+	view->tic[0] = (tic[0] &  ~NV50TIC_0_0_SWIZZLE_MASK) |
+		(swz[0] << NV50TIC_0_0_MAPR_SHIFT) |
+		(swz[1] << NV50TIC_0_0_MAPG_SHIFT) |
+		(swz[2] << NV50TIC_0_0_MAPB_SHIFT) |
+		(swz[3] << NV50TIC_0_0_MAPA_SHIFT);
 
-	desc = util_format_description(mt->base.base.format);
-	assert(desc);
+	tic[2] = 0x50001000;
+	tic[2] |= ((mt->base.bo->tile_mode & 0x0f) << 22) |
+		  ((mt->base.bo->tile_mode & 0xf0) << 21);
 
+	desc = util_format_description(mt->base.base.format);
 	if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB)
-		mode |= 0x0400;
+		tic[2] |= NV50TIC_0_2_COLORSPACE_SRGB;
 
 	switch (mt->base.base.target) {
 	case PIPE_TEXTURE_1D:
+		tic[2] |= NV50TIC_0_2_TARGET_1D;
 		break;
 	case PIPE_TEXTURE_2D:
-		mode |= (1 << 14);
+		tic[2] |= NV50TIC_0_2_TARGET_2D;
 		break;
 	case PIPE_TEXTURE_3D:
-		mode |= (2 << 14);
+		tic[2] |= NV50TIC_0_2_TARGET_3D;
 		break;
 	case PIPE_TEXTURE_CUBE:
-		mode |= (3 << 14);
+		tic[2] |= NV50TIC_0_2_TARGET_CUBE;
 		break;
 	default:
-		assert(!"unsupported texture target");
-		break;
+		NOUVEAU_ERR("invalid texture target: %d\n",
+			    mt->base.base.target);
+		return FALSE;
 	}
 
-	so_data (so, nv50_tex_format_list[i].hw);
-	so_reloc(so, mt->base.bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
-		 NOUVEAU_BO_RD, 0, 0);
-	so_data (so, mode);
-	so_data (so, 0x00300000);
-	so_data (so, mt->base.base.width0 | (1 << 31));
-	so_data (so, (mt->base.base.last_level << 28) |
-		 (mt->base.base.depth0 << 16) | mt->base.base.height0);
-	so_data (so, 0x03000000);
-	so_data (so, mt->base.base.last_level << 4);
-
-	return 0;
-}
+	tic[3] = 0x00300000;
+
+	tic[4] = (1 << 31) | mt->base.base.width0;
+	tic[5] = (mt->base.base.last_level << 28) |
+		(mt->base.base.depth0 << 16) | mt->base.base.height0;
+
+	tic[6] = 0x03000000;
 
-#ifndef NV50TCL_BIND_TIC
-#define NV50TCL_BIND_TIC(n) (0x1448 + 8 * n)
-#endif
+	tic[7] = (view->pipe.last_level << 4) | view->pipe.first_level;
+
+	return TRUE;
+}
 
-static boolean
+static int
 nv50_validate_textures(struct nv50_context *nv50, struct nouveau_stateobj *so,
 		       unsigned p)
 {
-	static const unsigned p_remap[PIPE_SHADER_TYPES] = { 0, 2, 1 };
-
 	struct nouveau_grobj *eng2d = nv50->screen->eng2d;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	unsigned unit, j, p_hw = p_remap[p];
+	unsigned unit, j;
+
+	const unsigned rll = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW;
+	const unsigned rlh = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH
+		| NOUVEAU_BO_OR;
 
 	nv50_so_init_sifc(nv50, so, nv50->screen->tic, NOUVEAU_BO_VRAM,
-			  p * (32 * 8 * 4), nv50->miptree_nr[p] * 8 * 4);
+			  p * (32 * 8 * 4), nv50->sampler_view_nr[p] * 8 * 4);
 
-	for (unit = 0; unit < nv50->miptree_nr[p]; ++unit) {
-		struct nv50_miptree *mt = nv50->miptree[p][unit];
+	for (unit = 0; unit < nv50->sampler_view_nr[p]; ++unit) {
+		struct nv50_sampler_view *view =
+			nv50_sampler_view(nv50->sampler_views[p][unit]);
 
 		so_method(so, eng2d, NV50_2D_SIFC_DATA | (2 << 29), 8);
-		if (mt) {
-			if (nv50_tex_construct(nv50, so, mt, unit, p))
-				return FALSE;
+		if (view) {
+			uint32_t tic2 = view->tic[2];
+			struct nv50_miptree *mt =
+				nv50_miptree(view->pipe.texture);
+
+			if (nv50->sampler[p][unit]->normalized)
+				tic2 |= NV50TIC_0_2_NORMALIZED_COORDS;
+
+			so_data  (so, view->tic[0]);
+			so_reloc (so, mt->base.bo, 0, rll, 0, 0);
+			so_reloc (so, mt->base.bo, 0, rlh, tic2, tic2);
+			so_datap (so, &view->tic[3], 5);
+
 			/* Set TEX insn $t src binding $unit in program type p
 			 * to TIC, TSC entry (32 * p + unit), mark valid (1).
 			 */
-			so_method(so, tesla, NV50TCL_BIND_TIC(p_hw), 1);
+			so_method(so, tesla, NV50TCL_BIND_TIC(p), 1);
 			so_data  (so, ((32 * p + unit) << 9) | (unit << 1) | 1);
 		} else {
 			for (j = 0; j < 8; ++j)
 				so_data(so, 0);
-			so_method(so, tesla, NV50TCL_BIND_TIC(p_hw), 1);
+			so_method(so, tesla, NV50TCL_BIND_TIC(p), 1);
 			so_data  (so, (unit << 1) | 0);
 		}
 	}
 
-	for (; unit < nv50->state.miptree_nr[p]; unit++) {
+	for (; unit < nv50->state.sampler_view_nr[p]; unit++) {
 		/* Make other bindings invalid. */
-		so_method(so, tesla, NV50TCL_BIND_TIC(p_hw), 1);
+		so_method(so, tesla, NV50TCL_BIND_TIC(p), 1);
 		so_data  (so, (unit << 1) | 0);
 	}
 
-	nv50->state.miptree_nr[p] = nv50->miptree_nr[p];
+	nv50->state.sampler_view_nr[p] = nv50->sampler_view_nr[p];
 	return TRUE;
 }
 
@@ -202,23 +220,25 @@ nv50_tex_relocs(struct nv50_context *nv50)
 	int p, unit;
 
 	p = PIPE_SHADER_FRAGMENT;
-	for (unit = 0; unit < nv50->miptree_nr[p]; unit++) {
-		if (!nv50->miptree[p][unit])
+	for (unit = 0; unit < nv50->sampler_view_nr[p]; unit++) {
+		struct pipe_sampler_view *view = nv50->sampler_views[p][unit];
+		if (!view)
 			continue;
 		nouveau_reloc_emit(chan, nv50->screen->tic,
 				   ((p * 32) + unit) * 32, NULL,
-				   nv50->miptree[p][unit]->base.bo, 0, 0,
+				   nv50_miptree(view->texture)->base.bo, 0, 0,
 				   NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
 				   NOUVEAU_BO_RD, 0, 0);
 	}
 
 	p = PIPE_SHADER_VERTEX;
-	for (unit = 0; unit < nv50->miptree_nr[p]; unit++) {
-		if (!nv50->miptree[p][unit])
+	for (unit = 0; unit < nv50->sampler_view_nr[p]; unit++) {
+		struct pipe_sampler_view *view = nv50->sampler_views[p][unit];
+		if (!view)
 			continue;
 		nouveau_reloc_emit(chan, nv50->screen->tic,
 				   ((p * 32) + unit) * 32, NULL,
-				   nv50->miptree[p][unit]->base.bo, 0, 0,
+				   nv50_miptree(view->texture)->base.bo, 0, 0,
 				   NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
 				   NOUVEAU_BO_RD, 0, 0);
 	}
@@ -229,21 +249,23 @@ nv50_tex_validate(struct nv50_context *nv50)
 {
 	struct nouveau_stateobj *so;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	unsigned p, start, push, nrlc;
-
-	for (nrlc = 0, start = 0, push = 0, p = 0; p < PIPE_SHADER_TYPES; ++p) {
-		start += MAX2(nv50->miptree_nr[p], nv50->state.miptree_nr[p]);
-		push += MAX2(nv50->miptree_nr[p], nv50->state.miptree_nr[p]);
-		nrlc += nv50->miptree_nr[p];
+	unsigned p, m = 0, d = 0, r = 0;
+
+	for (p = 0; p < 3; ++p) {
+		unsigned nr = MAX2(nv50->sampler_view_nr[p],
+				   nv50->state.sampler_view_nr[p]);
+		m += nr;
+		d += nr;
+		r += nv50->sampler_view_nr[p];
 	}
-	start = start * 2 + 4 * PIPE_SHADER_TYPES + 2;
-	push = push * 9 + 19 * PIPE_SHADER_TYPES + 2;
-	nrlc = nrlc * 2 + 2 * PIPE_SHADER_TYPES;
+	m = m * 2 + 3 * 4 + 1;
+	d = d * 9 + 3 * 19 + 1;
+	r = r * 2 + 3 * 2;
 
-	so = so_new(start, push, nrlc);
+	so = so_new(m, d, r);
 
-	if (nv50_validate_textures(nv50, so, PIPE_SHADER_VERTEX) == FALSE ||
-	    nv50_validate_textures(nv50, so, PIPE_SHADER_FRAGMENT) == FALSE) {
+	if (nv50_validate_textures(nv50, so, 0) == FALSE ||
+	    nv50_validate_textures(nv50, so, 2) == FALSE) {
 		so_ref(NULL, &so);
 
 		NOUVEAU_ERR("failed tex validate\n");
diff --git a/src/gallium/drivers/nv50/nv50_texture.h b/src/gallium/drivers/nv50/nv50_texture.h
index b870302019a..3475d3e4326 100644
--- a/src/gallium/drivers/nv50/nv50_texture.h
+++ b/src/gallium/drivers/nv50/nv50_texture.h
@@ -7,7 +7,9 @@
  */
 
 /* Texture image control block */
+#define NV50TIC_0_0_SWIZZLE_MASK                                  0x3ffc0000
 #define NV50TIC_0_0_MAPA_MASK                                     0x38000000
+#define NV50TIC_0_0_MAPA_SHIFT                                            27
 #define NV50TIC_0_0_MAPA_ZERO                                     0x00000000
 #define NV50TIC_0_0_MAPA_C0                                       0x10000000
 #define NV50TIC_0_0_MAPA_C1                                       0x18000000
@@ -15,6 +17,7 @@
 #define NV50TIC_0_0_MAPA_C3                                       0x28000000
 #define NV50TIC_0_0_MAPA_ONE                                      0x38000000
 #define NV50TIC_0_0_MAPB_MASK                                     0x07000000
+#define NV50TIC_0_0_MAPB_SHIFT                                            24
 #define NV50TIC_0_0_MAPB_ZERO                                     0x00000000
 #define NV50TIC_0_0_MAPB_C0                                       0x02000000
 #define NV50TIC_0_0_MAPB_C1                                       0x03000000
@@ -22,6 +25,7 @@
 #define NV50TIC_0_0_MAPB_C3                                       0x05000000
 #define NV50TIC_0_0_MAPB_ONE                                      0x07000000
 #define NV50TIC_0_0_MAPG_MASK                                     0x00e00000
+#define NV50TIC_0_0_MAPG_SHIFT                                            21
 #define NV50TIC_0_0_MAPG_ZERO                                     0x00000000
 #define NV50TIC_0_0_MAPG_C0                                       0x00400000
 #define NV50TIC_0_0_MAPG_C1                                       0x00600000
@@ -29,6 +33,7 @@
 #define NV50TIC_0_0_MAPG_C3                                       0x00a00000
 #define NV50TIC_0_0_MAPG_ONE                                      0x00e00000
 #define NV50TIC_0_0_MAPR_MASK                                     0x001c0000
+#define NV50TIC_0_0_MAPR_SHIFT                                            18
 #define NV50TIC_0_0_MAPR_ZERO                                     0x00000000
 #define NV50TIC_0_0_MAPR_C0                                       0x00080000
 #define NV50TIC_0_0_MAPR_C1                                       0x000c0000
@@ -89,22 +94,39 @@
 #define NV50TIC_0_1_OFFSET_LOW_MASK                               0xffffffff
 #define NV50TIC_0_1_OFFSET_LOW_SHIFT                                       0
 
-#define NV50TIC_0_2_UNKNOWN_MASK                                  0xffffffff
+#define NV50TIC_0_2_COLORSPACE_SRGB                               0x00000400
+#define NV50TIC_0_2_TARGET_1D                                     0x00000000
+#define NV50TIC_0_2_TARGET_2D                                     0x00004000
+#define NV50TIC_0_2_TARGET_3D                                     0x00008000
+#define NV50TIC_0_2_TARGET_CUBE                                   0x0000c000
+#define NV50TIC_0_2_TARGET_1D_ARRAY                               0x00010000
+#define NV50TIC_0_2_TARGET_2D_ARRAY                               0x00014000
+#define NV50TIC_0_2_TARGET_BUFFER                                 0x00018000
+#define NV50TIC_0_2_TARGET_RECT                                   0x0001c000
+/* #define NV50TIC_0_0_TILE_MODE_LINEAR                           0x00040000 */
+#define NV50TIC_0_2_TILE_MODE_Y_MASK                              0x01c00000
+#define NV50TIC_0_2_TILE_MODE_Y_SHIFT                                     22
+#define NV50TIC_0_2_TILE_MODE_Z_MASK                              0x0e000000
+#define NV50TIC_0_2_TILE_MODE_Z_SHIFT                                     25
+#define NV50TIC_0_2_NORMALIZED_COORDS                             0x80000000
 
 #define NV50TIC_0_3_UNKNOWN_MASK                                  0xffffffff
 
 #define NV50TIC_0_4_WIDTH_MASK                                    0x0000ffff
 #define NV50TIC_0_4_WIDTH_SHIFT                                            0
 
-#define NV50TIC_0_5_DEPTH_MASK                                    0xffff0000
+#define NV50TIC_0_5_LAST_LEVEL_MASK                               0xf0000000
+#define NV50TIC_0_5_LAST_LEVEL_SHIFT                                      28
+#define NV50TIC_0_5_DEPTH_MASK                                    0x0fff0000
 #define NV50TIC_0_5_DEPTH_SHIFT                                           16
 #define NV50TIC_0_5_HEIGHT_MASK                                   0x0000ffff
 #define NV50TIC_0_5_HEIGHT_SHIFT                                           0
-
 #define NV50TIC_0_6_UNKNOWN_MASK                                  0xffffffff
 
-#define NV50TIC_0_7_OFFSET_HIGH_MASK                              0xffffffff
-#define NV50TIC_0_7_OFFSET_HIGH_SHIFT                                      0
+#define NV50TIC_0_7_BASE_LEVEL_MASK                               0x0000000f
+#define NV50TIC_0_7_BASE_LEVEL_SHIFT                                       0
+#define NV50TIC_0_7_MAX_LEVEL_MASK                                0x000000f0
+#define NV50TIC_0_7_MAX_LEVEL_SHIFT                                        4
 
 /* Texture sampler control block */
 #define NV50TSC_1_0_WRAPS_MASK                                   0x00000007
diff --git a/src/gallium/drivers/nvfx/nv30_fragtex.c b/src/gallium/drivers/nvfx/nv30_fragtex.c
index 2b56f454921..54e47574c4d 100644
--- a/src/gallium/drivers/nvfx/nv30_fragtex.c
+++ b/src/gallium/drivers/nvfx/nv30_fragtex.c
@@ -91,7 +91,7 @@ struct nouveau_stateobj *
 nv30_fragtex_build(struct nvfx_context *nvfx, int unit)
 {
 	struct nvfx_sampler_state *ps = nvfx->tex_sampler[unit];
-	struct nvfx_miptree *nv30mt = nvfx->tex_miptree[unit];
+	struct nvfx_miptree *nv30mt = (struct nvfx_miptree *)nvfx->fragment_sampler_views[unit]->texture;
 	struct pipe_texture *pt = &nv30mt->base;
 	struct nouveau_bo *bo = nouveau_bo(nv30mt->buffer);
 	struct nv30_texture_format *tf;
diff --git a/src/gallium/drivers/nvfx/nv40_fragtex.c b/src/gallium/drivers/nvfx/nv40_fragtex.c
index 5889b5e40d5..05506e28099 100644
--- a/src/gallium/drivers/nvfx/nv40_fragtex.c
+++ b/src/gallium/drivers/nvfx/nv40_fragtex.c
@@ -109,7 +109,7 @@ struct nouveau_stateobj *
 nv40_fragtex_build(struct nvfx_context *nvfx, int unit)
 {
 	struct nvfx_sampler_state *ps = nvfx->tex_sampler[unit];
-	struct nvfx_miptree *nv40mt = nvfx->tex_miptree[unit];
+	struct nvfx_miptree *nv40mt = (struct nvfx_miptree *)nvfx->fragment_sampler_views[unit]->texture;
 	struct nouveau_bo *bo = nouveau_bo(nv40mt->buffer);
 	struct pipe_texture *pt = &nv40mt->base;
 	struct nv40_texture_format *tf;
diff --git a/src/gallium/drivers/nvfx/nvfx_context.h b/src/gallium/drivers/nvfx/nvfx_context.h
index 5eed8a560e5..ab7225cf6c3 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.h
+++ b/src/gallium/drivers/nvfx/nvfx_context.h
@@ -158,7 +158,7 @@ struct nvfx_context {
 	struct pipe_buffer *idxbuf;
 	unsigned idxbuf_format;
 	struct nvfx_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
-	struct nvfx_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
 	unsigned nr_samplers;
 	unsigned nr_textures;
 	unsigned dirty_samplers;
diff --git a/src/gallium/drivers/nvfx/nvfx_draw.c b/src/gallium/drivers/nvfx/nvfx_draw.c
index 5379b29efd1..68e50a36479 100644
--- a/src/gallium/drivers/nvfx/nvfx_draw.c
+++ b/src/gallium/drivers/nvfx/nvfx_draw.c
@@ -79,6 +79,12 @@ nvfx_render_vertex(struct nvfx_context *nvfx, const struct vertex_header *v)
 					    float_to_ubyte(v->data[idx][1]),
 					    float_to_ubyte(v->data[idx][2]),
 					    float_to_ubyte(v->data[idx][3])));
+		case EMIT_4UB_BGRA:
+			BEGIN_RING(chan, eng3d, NV34TCL_VTX_ATTR_4UB(hw), 1);
+			OUT_RING  (chan, pack_ub4(float_to_ubyte(v->data[idx][2]),
+					    float_to_ubyte(v->data[idx][1]),
+					    float_to_ubyte(v->data[idx][0]),
+					    float_to_ubyte(v->data[idx][3])));
 			break;
 		default:
 			assert(0);
diff --git a/src/gallium/drivers/nvfx/nvfx_fragprog.c b/src/gallium/drivers/nvfx/nvfx_fragprog.c
index 76351430f44..b9c91cec8ce 100644
--- a/src/gallium/drivers/nvfx/nvfx_fragprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_fragprog.c
@@ -803,7 +803,8 @@ nvfx_fragprog_translate(struct nvfx_context *nvfx,
 		fp->fp_control |= fpc->num_regs << NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT;
 
 	/* Terminate final instruction */
-	fp->insn[fpc->inst_offset] |= 0x00000001;
+	if(fp->insn)
+                fp->insn[fpc->inst_offset] |= 0x00000001;
 
 	/* Append NOP + END instruction, may or may not be necessary. */
 	fpc->inst_offset = fp->insn_len;
diff --git a/src/gallium/drivers/nvfx/nvfx_miptree.c b/src/gallium/drivers/nvfx/nvfx_miptree.c
index 0f5ed61aab7..9de25175e78 100644
--- a/src/gallium/drivers/nvfx/nvfx_miptree.c
+++ b/src/gallium/drivers/nvfx/nvfx_miptree.c
@@ -67,6 +67,9 @@ nvfx_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 	struct nvfx_miptree *mt;
 	unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL |
 	                     NOUVEAU_BUFFER_USAGE_TEXTURE;
+	static int no_swizzle = -1;
+	if(no_swizzle < 0)
+		no_swizzle = debug_get_bool_option("NOUVEAU_NO_SWIZZLE", FALSE);
 
 	mt = MALLOC(sizeof(struct nvfx_miptree));
 	if (!mt)
@@ -106,7 +109,7 @@ nvfx_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 		case PIPE_FORMAT_B8G8R8X8_UNORM:
 		case PIPE_FORMAT_R16_SNORM:
 		{
-			if (debug_get_bool_option("NOUVEAU_NO_SWIZZLE", FALSE))
+			if (no_swizzle)
 				mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
 			break;
 		}
diff --git a/src/gallium/drivers/nvfx/nvfx_screen.c b/src/gallium/drivers/nvfx/nvfx_screen.c
index 8138715cc7d..f7f39218944 100644
--- a/src/gallium/drivers/nvfx/nvfx_screen.c
+++ b/src/gallium/drivers/nvfx/nvfx_screen.c
@@ -68,11 +68,9 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, int param)
 	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
 		return !!screen->is_nv4x;
 	case NOUVEAU_CAP_HW_VTXBUF:
-		/* TODO: this is almost surely wrong */
-		return !!screen->is_nv4x;
+		return 0;
 	case NOUVEAU_CAP_HW_IDXBUF:
-		/* TODO: this is also almost surely wrong */
-		return screen->is_nv4x && screen->eng3d->grclass == NV40TCL;
+		return 0;
 	case PIPE_CAP_MAX_COMBINED_SAMPLERS:
 		return 16;
 	case PIPE_CAP_INDEP_BLEND_ENABLE:
@@ -295,6 +293,36 @@ static void nv40_screen_init(struct nvfx_screen *screen, struct nouveau_stateobj
 	so_data  (so, 0x00000001);
 }
 
+static void
+nvfx_screen_init_buffer_functions(struct nvfx_screen* screen)
+{
+	int vram_hack_default = 0;
+	int vram_hack;
+	// TODO: this is a bit of a guess; also add other cards that may need this hack.
+	// It may also depend on the specific card or the AGP/PCIe chipset.
+	if(screen->base.device->chipset == 0x47 /* G70 */
+		|| screen->base.device->chipset == 0x49 /* G71 */
+		|| screen->base.device->chipset == 0x46 /* G72 */
+		)
+		vram_hack_default = 1;
+	vram_hack = debug_get_bool_option("NOUVEAU_VTXIDX_IN_VRAM", vram_hack_default);
+
+#ifdef DEBUG
+	if(!vram_hack)
+	{
+		fprintf(stderr, "Some systems may experience graphics corruption due to randomly misplaced vertices.\n"
+			"If this is happening, export NOUVEAU_VTXIDX_IN_VRAM=1 may reduce or eliminate the problem\n");
+	}
+	else
+	{
+		fprintf(stderr, "A performance reducing hack is being used to help avoid graphics corruption.\n"
+			"You can try export NOUVEAU_VTXIDX_IN_VRAM=0 to disable it.\n");
+	}
+#endif
+
+	screen->vertex_buffer_flags = vram_hack ? NOUVEAU_BO_VRAM : NOUVEAU_BO_GART;
+}
+
 struct pipe_screen *
 nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
@@ -352,6 +380,7 @@ nvfx_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 		return NULL;
 	}
 
+	nvfx_screen_init_buffer_functions(screen);
 	nvfx_screen_init_miptree_functions(pscreen);
 
 	ret = nouveau_grobj_alloc(chan, 0xbeef3097, eng3d_class, &screen->eng3d);
diff --git a/src/gallium/drivers/nvfx/nvfx_screen.h b/src/gallium/drivers/nvfx/nvfx_screen.h
index c0b4b9899dd..baa848c47aa 100644
--- a/src/gallium/drivers/nvfx/nvfx_screen.h
+++ b/src/gallium/drivers/nvfx/nvfx_screen.h
@@ -12,6 +12,7 @@ struct nvfx_screen {
 	struct nvfx_context *cur_ctx;
 
 	unsigned is_nv4x; /* either 0 or ~0 */
+	int vertex_buffer_flags;
 
 	/* HW graphics objects */
 	struct nv04_surface_2d *eng2d;
diff --git a/src/gallium/drivers/nvfx/nvfx_state.c b/src/gallium/drivers/nvfx/nvfx_state.c
index 88a9d01c509..ecaa0dcb16a 100644
--- a/src/gallium/drivers/nvfx/nvfx_state.c
+++ b/src/gallium/drivers/nvfx/nvfx_state.c
@@ -137,21 +137,22 @@ nvfx_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
 }
 
 static void
-nvfx_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
-			 struct pipe_texture **miptree)
+nvfx_set_fragment_sampler_views(struct pipe_context *pipe,
+				unsigned nr,
+				struct pipe_sampler_view **views)
 {
 	struct nvfx_context *nvfx = nvfx_context(pipe);
 	unsigned unit;
 
 	for (unit = 0; unit < nr; unit++) {
-		pipe_texture_reference((struct pipe_texture **)
-				       &nvfx->tex_miptree[unit], miptree[unit]);
+		pipe_sampler_view_reference(&nvfx->fragment_sampler_views[unit],
+                                            views[unit]);
 		nvfx->dirty_samplers |= (1 << unit);
 	}
 
 	for (unit = nr; unit < nvfx->nr_textures; unit++) {
-		pipe_texture_reference((struct pipe_texture **)
-				       &nvfx->tex_miptree[unit], NULL);
+		pipe_sampler_view_reference(&nvfx->fragment_sampler_views[unit],
+                                            NULL);
 		nvfx->dirty_samplers |= (1 << unit);
 	}
 
@@ -159,6 +160,34 @@ nvfx_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
 	nvfx->dirty |= NVFX_NEW_SAMPLER;
 }
 
+
+static struct pipe_sampler_view *
+nvfx_create_sampler_view(struct pipe_context *pipe,
+			 struct pipe_texture *texture,
+			 const struct pipe_sampler_view *templ)
+{
+	struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+	if (view) {
+		*view = *templ;
+		view->reference.count = 1;
+		view->texture = NULL;
+		pipe_texture_reference(&view->texture, texture);
+		view->context = pipe;
+	}
+
+	return view;
+}
+
+
+static void
+nvfx_sampler_view_destroy(struct pipe_context *pipe,
+			  struct pipe_sampler_view *view)
+{
+	pipe_texture_reference(&view->texture, NULL);
+	FREE(view);
+}
+
 static void *
 nvfx_rasterizer_state_create(struct pipe_context *pipe,
 			     const struct pipe_rasterizer_state *cso)
@@ -581,7 +610,9 @@ nvfx_init_state_functions(struct nvfx_context *nvfx)
 	nvfx->pipe.create_sampler_state = nvfx_sampler_state_create;
 	nvfx->pipe.bind_fragment_sampler_states = nvfx_sampler_state_bind;
 	nvfx->pipe.delete_sampler_state = nvfx_sampler_state_delete;
-	nvfx->pipe.set_fragment_sampler_textures = nvfx_set_sampler_texture;
+	nvfx->pipe.set_fragment_sampler_views = nvfx_set_fragment_sampler_views;
+        nvfx->pipe.create_sampler_view = nvfx_create_sampler_view;
+        nvfx->pipe.sampler_view_destroy = nvfx_sampler_view_destroy;
 
 	nvfx->pipe.create_rasterizer_state = nvfx_rasterizer_state_create;
 	nvfx->pipe.bind_rasterizer_state = nvfx_rasterizer_state_bind;
diff --git a/src/gallium/drivers/nvfx/nvfx_state_viewport.c b/src/gallium/drivers/nvfx/nvfx_state_viewport.c
index 82e0e9220b0..ec730e3a9e9 100644
--- a/src/gallium/drivers/nvfx/nvfx_state_viewport.c
+++ b/src/gallium/drivers/nvfx/nvfx_state_viewport.c
@@ -1,15 +1,16 @@
 #include "nvfx_context.h"
 
+/* Having this depend on FB and RAST looks wrong, but it seems
+   necessary to make this work on nv3x
+   TODO: find the right fix
+*/
+
 static boolean
 nvfx_state_viewport_validate(struct nvfx_context *nvfx)
 {
 	struct pipe_viewport_state *vpt = &nvfx->viewport;
 	struct nouveau_stateobj *so;
 
-	if (nvfx->state.hw[NVFX_STATE_VIEWPORT] &&
-	    !(nvfx->dirty & NVFX_NEW_VIEWPORT))
-		return FALSE;
-
 	so = so_new(2, 9, 0);
 	so_method(so, nvfx->screen->eng3d,
 		  NV34TCL_VIEWPORT_TRANSLATE_X, 8);
@@ -45,7 +46,7 @@ nvfx_state_viewport_validate(struct nvfx_context *nvfx)
 struct nvfx_state_entry nvfx_state_viewport = {
 	.validate = nvfx_state_viewport_validate,
 	.dirty = {
-		.pipe = NVFX_NEW_VIEWPORT,
+		.pipe = NVFX_NEW_VIEWPORT | NVFX_NEW_FB | NVFX_NEW_RAST,
 		.hw = NVFX_STATE_VIEWPORT
 	}
 };
diff --git a/src/gallium/drivers/nvfx/nvfx_transfer.c b/src/gallium/drivers/nvfx/nvfx_transfer.c
index 409b354d582..1c250e9fe44 100644
--- a/src/gallium/drivers/nvfx/nvfx_transfer.c
+++ b/src/gallium/drivers/nvfx/nvfx_transfer.c
@@ -33,15 +33,18 @@ nvfx_compatible_transfer_tex(struct pipe_texture *pt, unsigned width, unsigned h
 }
 
 static struct pipe_transfer *
-nvfx_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
+nvfx_transfer_new(struct pipe_context *pipe, struct pipe_texture *pt,
 		  unsigned face, unsigned level, unsigned zslice,
 		  enum pipe_transfer_usage usage,
 		  unsigned x, unsigned y, unsigned w, unsigned h)
 {
-        struct pipe_screen *pscreen = pcontext->screen;
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nvfx_miptree *mt = (struct nvfx_miptree *)pt;
 	struct nvfx_transfer *tx;
 	struct pipe_texture tx_tex_template, *tx_tex;
+	static int no_transfer = -1;
+	if(no_transfer < 0)
+		no_transfer = debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/);
 
 	tx = CALLOC_STRUCT(nvfx_transfer);
 	if (!tx)
@@ -59,8 +62,7 @@ nvfx_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 	tx->base.zslice = zslice;
 
 	/* Direct access to texture */
-	if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC ||
-	     debug_get_bool_option("NOUVEAU_NO_TRANSFER", TRUE/*XXX:FALSE*/)) &&
+	if ((pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC || no_transfer) &&
 	    pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)
 	{
 		tx->direct = true;
@@ -118,13 +120,13 @@ nvfx_transfer_new(struct pipe_context *pcontext, struct pipe_texture *pt,
 }
 
 static void
-nvfx_transfer_del(struct pipe_context *pcontext,
+nvfx_transfer_del(struct pipe_context *pipe,
                   struct pipe_transfer *ptx)
 {
 	struct nvfx_transfer *tx = (struct nvfx_transfer *)ptx;
 
 	if (!tx->direct && (ptx->usage & PIPE_TRANSFER_WRITE)) {
-		struct pipe_screen *pscreen = pcontext->screen;
+		struct pipe_screen *pscreen = pipe->screen;
 		struct nvfx_screen *nvscreen = nvfx_screen(pscreen);
 		struct pipe_surface *dst;
 
@@ -147,9 +149,9 @@ nvfx_transfer_del(struct pipe_context *pcontext,
 }
 
 static void *
-nvfx_transfer_map(struct pipe_context *pcontext, struct pipe_transfer *ptx)
+nvfx_transfer_map(struct pipe_context *pipe, struct pipe_transfer *ptx)
 {
-        struct pipe_screen *pscreen = pcontext->screen;
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nvfx_transfer *tx = (struct nvfx_transfer *)ptx;
 	struct nv04_surface *ns = (struct nv04_surface *)tx->surface;
 	struct nvfx_miptree *mt = (struct nvfx_miptree *)tx->surface->texture;
@@ -163,9 +165,9 @@ nvfx_transfer_map(struct pipe_context *pcontext, struct pipe_transfer *ptx)
 }
 
 static void
-nvfx_transfer_unmap(struct pipe_context *pcontext, struct pipe_transfer *ptx)
+nvfx_transfer_unmap(struct pipe_context *pipe, struct pipe_transfer *ptx)
 {
-	struct pipe_screen *pscreen = pcontext->screen;
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nvfx_transfer *tx = (struct nvfx_transfer *)ptx;
 	struct nvfx_miptree *mt = (struct nvfx_miptree *)tx->surface->texture;
 
diff --git a/src/gallium/drivers/nvfx/nvfx_vbo.c b/src/gallium/drivers/nvfx/nvfx_vbo.c
index 257087f8f63..c26536b0e77 100644
--- a/src/gallium/drivers/nvfx/nvfx_vbo.c
+++ b/src/gallium/drivers/nvfx/nvfx_vbo.c
@@ -495,7 +495,7 @@ nvfx_vbo_validate(struct nvfx_context *nvfx)
 	struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
 	struct pipe_buffer *ib = nvfx->idxbuf;
 	unsigned ib_format = nvfx->idxbuf_format;
-	unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
+	unsigned vb_flags = nvfx->screen->vertex_buffer_flags | NOUVEAU_BO_RD;
 	int hw;
 
 	vtxbuf = so_new(3, 17, 18);
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
index 27b2e309932..221878c1c0c 100644
--- a/src/gallium/drivers/r300/SConscript
+++ b/src/gallium/drivers/r300/SConscript
@@ -24,6 +24,7 @@ r300 = env.ConvenienceLibrary(
         'r300_query.c',
         'r300_render.c',
         'r300_screen.c',
+        'r300_screen_buffer.c',
         'r300_state.c',
         'r300_state_derived.c',
         'r300_state_invariant.c',
diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c
index b7ad6b20206..bcdf950df9c 100644
--- a/src/gallium/drivers/r300/r300_blit.c
+++ b/src/gallium/drivers/r300/r300_blit.c
@@ -113,9 +113,9 @@ static void r300_hw_copy(struct pipe_context* pipe,
     util_blitter_save_fragment_sampler_states(
         r300->blitter, state->sampler_count, (void**)state->sampler_states);
 
-    util_blitter_save_fragment_sampler_textures(
+    util_blitter_save_fragment_sampler_views(
         r300->blitter, state->texture_count,
-        (struct pipe_texture**)state->textures);
+        state->fragment_sampler_views);
 
     /* Do a copy */
     util_blitter_copy(r300->blitter,
@@ -149,6 +149,9 @@ void r300_surface_copy(struct pipe_context* pipe,
             case 4:
                 new_format = PIPE_FORMAT_B8G8R8A8_UNORM;
                 break;
+            case 8:
+                new_format = PIPE_FORMAT_R16G16B16A16_UNORM;
+                break;
             default:
                 debug_printf("r300: surface_copy: Unhandled format: %s. Falling back to software.\n"
                              "r300: surface_copy: Software fallback doesn't work for tiled textures.\n",
diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c
index 92de297ef1d..41719862635 100644
--- a/src/gallium/drivers/r300/r300_chipset.c
+++ b/src/gallium/drivers/r300/r300_chipset.c
@@ -24,6 +24,8 @@
 
 #include "util/u_debug.h"
 
+#include <stdio.h>
+
 /* r300_chipset: A file all to itself for deducing the various properties of
  * Radeons. */
 
@@ -365,8 +367,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             break;
 
         default:
-            debug_printf("r300: Warning: Unknown chipset 0x%x\n",
-                caps->pci_id);
-            break;
+            fprintf(stderr, "r300: Warning: Unknown chipset 0x%x\n",
+                    caps->pci_id);
     }
 }
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index d994a46ccfe..4b470b2c6af 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -33,11 +33,10 @@
 #include "r300_query.h"
 #include "r300_render.h"
 #include "r300_screen.h"
+#include "r300_screen_buffer.h"
 #include "r300_state_invariant.h"
-#include "r300_texture.h"
 #include "r300_transfer.h"
-
-#include "radeon_winsys.h"
+#include "r300_winsys.h"
 
 static void r300_destroy_context(struct pipe_context* context)
 {
@@ -72,29 +71,23 @@ static void r300_destroy_context(struct pipe_context* context)
 }
 
 static unsigned int
-r300_is_texture_referenced(struct pipe_context *pipe,
+r300_is_texture_referenced(struct pipe_context *context,
                            struct pipe_texture *texture,
                            unsigned face, unsigned level)
 {
-    return 0;
+    struct r300_context* r300 = r300_context(context);
+    struct r300_texture* tex = (struct r300_texture*)texture;
+
+    return r300->rws->is_buffer_referenced(r300->rws, tex->buffer);
 }
 
 static unsigned int
-r300_is_buffer_referenced(struct pipe_context *pipe,
+r300_is_buffer_referenced(struct pipe_context *context,
                           struct pipe_buffer *buf)
 {
-    /* This only checks to see whether actual hardware buffers are
-     * referenced. Since we use managed BOs and transfers, it's actually not
-     * possible for pipe_buffers to ever reference the actual hardware, so
-     * buffers are never referenced. 
-     */
-
-    /* XXX: that doesn't make sense given that
-     * r300_is_texture_referenced is implemented on top of this
-     * function and hardware can certainly refer to textures
-     * directly...
-     */
-    return 0;
+    struct r300_context* r300 = r300_context(context);
+
+    return r300_buffer_is_referenced(r300, buf);
 }
 
 static void r300_flush_cb(void *data)
@@ -204,7 +197,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
 
     /* Open up the OQ BO. */
     r300->oqbo = screen->buffer_create(screen, 4096,
-            PIPE_BUFFER_USAGE_VERTEX, 4096);
+            PIPE_BUFFER_USAGE_PIXEL, 4096);
     make_empty_list(&r300->query_list);
 
     r300_init_flush_functions(r300);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index db2f74e0745..0a82484e89d 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -30,6 +30,7 @@
 #include "pipe/p_context.h"
 #include "util/u_inlines.h"
 
+#include "r300_defines.h"
 #include "r300_screen.h"
 
 struct u_upload_mgr;
@@ -125,8 +126,6 @@ struct r300_texture_format_state {
     uint32_t format2; /* R300_TX_FORMAT2: 0x4500 */
 };
 
-#define R300_MAX_TEXTURE_LEVELS 13
-
 struct r300_texture_fb_state {
     /* Colorbuffer. */
     uint32_t colorpitch[R300_MAX_TEXTURE_LEVELS]; /* R300_RB3D_COLORPITCH[0-3]*/
@@ -139,7 +138,7 @@ struct r300_texture_fb_state {
 
 struct r300_textures_state {
     /* Textures. */
-    struct r300_texture *textures[8];
+    struct pipe_sampler_view *fragment_sampler_views[8];
     int texture_count;
     /* Sampler states. */
     struct r300_sampler_state *sampler_states[8];
@@ -185,12 +184,6 @@ struct r300_ztop_state {
     uint32_t z_buffer_top;      /* R300_ZB_ZTOP: 0x4f14 */
 };
 
-#define R300_NEW_FRAGMENT_SHADER 0x00000020
-#define R300_NEW_FRAGMENT_SHADER_CONSTANTS    0x00000040
-#define R300_NEW_VERTEX_SHADER_CONSTANTS    0x10000000
-#define R300_NEW_QUERY           0x40000000
-#define R300_NEW_KITCHEN_SINK    0x7fffffff
-
 /* The next several objects are not pure Radeon state; they inherit from
  * various Gallium classes. */
 
@@ -228,12 +221,6 @@ struct r300_query {
     struct r300_query* next;
 };
 
-enum r300_buffer_tiling {
-    R300_BUFFER_LINEAR = 0,
-    R300_BUFFER_TILED,
-    R300_BUFFER_SQUARETILED
-};
-
 struct r300_texture {
     /* Parent class */
     struct pipe_texture tex;
@@ -422,7 +409,7 @@ static INLINE void CTX_DBG(struct r300_context * ctx, unsigned flags,
     if (CTX_DBG_ON(ctx, flags)) {
         va_list va;
         va_start(va, fmt);
-        debug_vprintf(fmt, va);
+        vfprintf(stderr, fmt, va);
         va_end(va);
     }
 }
@@ -431,4 +418,3 @@ static INLINE void CTX_DBG(struct r300_context * ctx, unsigned flags,
 #define DBG     CTX_DBG
 
 #endif /* R300_CONTEXT_H */
-
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index ad07efbffdb..456b2ec7b92 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -26,8 +26,7 @@
 #include "util/u_math.h"
 
 #include "r300_reg.h"
-
-#include "radeon_winsys.h"
+#include "r300_winsys.h"
 
 /* Yes, I know macros are ugly. However, they are much prettier than the code
  * that they neatly hide away, and don't have the cost of function setup,so
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
index d6177577c8d..016e8d66061 100644
--- a/src/gallium/drivers/r300/r300_debug.c
+++ b/src/gallium/drivers/r300/r300_debug.c
@@ -22,6 +22,7 @@
 
 #include "r300_context.h"
 
+#include <stdio.h>
 
 struct debug_option {
     const char * name;
@@ -69,7 +70,7 @@ void r300_init_debug(struct r300_screen * screen)
             }
 
             if (!opt->name) {
-                debug_printf("Unknown debug option: %s\n", options);
+                fprintf(stderr, "Unknown debug option: %s\n", options);
                 printhint = TRUE;
             }
 
@@ -81,10 +82,13 @@ void r300_init_debug(struct r300_screen * screen)
     }
 
     if (printhint || screen->debug & DBG_HELP) {
-        debug_printf("You can enable debug output by setting the RADEON_DEBUG environment variable\n"
-                     "to a comma-separated list of debug options. Available options are:\n");
+        fprintf(stderr, "You can enable debug output by setting "
+                        "the RADEON_DEBUG environment variable\n"
+                        "to a comma-separated list of debug options. "
+                        "Available options are:\n");
+
         for(opt = debug_options; opt->name; ++opt) {
-            debug_printf("    %s: %s\n", opt->name, opt->description);
+            fprintf(stderr, "    %s: %s\n", opt->name, opt->description);
         }
     }
 }
diff --git a/src/gallium/drivers/r300/r300_defines.h b/src/gallium/drivers/r300/r300_defines.h
new file mode 100644
index 00000000000..2e04876de92
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_defines.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright 2010 Marek Olšák <[email protected]>
+ *
+ * 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
+ * on 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
+ * THE AUTHOR(S) AND/OR THEIR 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 R300_DEFINES_H
+#define R300_DEFINES_H
+
+#include "pipe/p_defines.h"
+
+#define R300_MAX_TEXTURE_LEVELS         13
+#define R300_MAX_DRAW_VBO_SIZE          (1024 * 1024)
+
+#define R300_TEXTURE_USAGE_TRANSFER     PIPE_TEXTURE_USAGE_CUSTOM
+
+/* Non-atom dirty state flags. */
+#define R300_NEW_FRAGMENT_SHADER                0x00000020
+#define R300_NEW_FRAGMENT_SHADER_CONSTANTS      0x00000040
+#define R300_NEW_VERTEX_SHADER_CONSTANTS        0x10000000
+#define R300_NEW_QUERY                          0x40000000
+#define R300_NEW_KITCHEN_SINK                   0x7fffffff
+
+/* Tiling flags. */
+enum r300_buffer_tiling {
+    R300_BUFFER_LINEAR = 0,
+    R300_BUFFER_TILED,
+    R300_BUFFER_SQUARETILED
+};
+
+#endif
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index d8c64dd9001..15bcf8907f3 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -33,7 +33,6 @@
 #include "r300_fs.h"
 #include "r300_screen.h"
 #include "r300_screen_buffer.h"
-#include "r300_state_inlines.h"
 #include "r300_vs.h"
 
 void r300_emit_blend_state(struct r300_context* r300,
@@ -146,10 +145,8 @@ static const float * get_shader_constant(
     struct rc_constant * constant,
     struct r300_constant_buffer * externals)
 {
-    struct r300_viewport_state* viewport =
-        (struct r300_viewport_state*)r300->viewport_state.state;
-    struct r300_textures_state* texstate =
-        (struct r300_textures_state*)r300->textures_state.state;
+    struct r300_viewport_state* viewport = r300->viewport_state.state;
+    struct r300_textures_state* texstate = r300->textures_state.state;
     static float vec[4] = { 0.0, 0.0, 0.0, 1.0 };
     struct pipe_texture *tex;
 
@@ -165,7 +162,7 @@ static const float * get_shader_constant(
                 /* Factor for converting rectangle coords to
                  * normalized coords. Should only show up on non-r500. */
                 case RC_STATE_R300_TEXRECT_FACTOR:
-                    tex = &texstate->textures[constant->u.State[1]]->tex;
+                    tex = texstate->fragment_sampler_views[constant->u.State[1]]->texture;
                     vec[0] = 1.0 / tex->width0;
                     vec[1] = 1.0 / tex->height0;
                     break;
@@ -189,13 +186,13 @@ static const float * get_shader_constant(
                     break;
 
                 default:
-                    debug_printf("r300: Implementation error: "
+                    fprintf(stderr, "r300: Implementation error: "
                         "Unknown RC_CONSTANT type %d\n", constant->u.State[0]);
             }
             break;
 
         default:
-            debug_printf("r300: Implementation error: "
+            fprintf(stderr, "r300: Implementation error: "
                 "Unhandled constant type %d\n", constant->Type);
     }
 
@@ -517,9 +514,9 @@ static void r300_emit_query_finish(struct r300_context *r300,
                     0, RADEON_GEM_DOMAIN_GTT, 0);
             break;
         default:
-            debug_printf("r300: Implementation error: Chipset reports %d"
+            fprintf(stderr, "r300: Implementation error: Chipset reports %d"
                     " pixel pipes!\n", caps->num_frag_pipes);
-            assert(0);
+            abort();
     }
 
     /* And, finally, reset it to normal... */
@@ -749,8 +746,9 @@ void r300_emit_textures_state(struct r300_context *r300,
             OUT_CS_REG(R300_TX_FORMAT2_0 + (i * 4), texstate->format[2]);
 
             OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (i * 4), 1);
-            OUT_CS_TEX_RELOC(allstate->textures[i], texstate->tile_config,
-                         RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0, 0);
+            OUT_CS_TEX_RELOC((struct r300_texture *)allstate->fragment_sampler_views[i]->texture,
+                             texstate->tile_config,
+                             RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0, 0);
         }
     }
     END_CS;
@@ -893,12 +891,6 @@ void r300_emit_vs_state(struct r300_context* r300, unsigned size, void* state)
 
     CS_LOCALS(r300);
 
-    if (!r300screen->caps->has_tcl) {
-        debug_printf("r300: Implementation error: emit_vs_state called,"
-                " but has_tcl is FALSE!\n");
-        return;
-    }
-
     BEGIN_CS(size);
     /* R300_VAP_PVS_CODE_CNTL_0
      * R300_VAP_PVS_CONST_CNTL
@@ -928,28 +920,19 @@ void r300_emit_vs_state(struct r300_context* r300, unsigned size, void* state)
 void r300_emit_vs_constant_buffer(struct r300_context* r300,
                                   struct rc_constant_list* constants)
 {
-    int i;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
+    unsigned i;
     CS_LOCALS(r300);
 
-    if (!r300screen->caps->has_tcl) {
-        debug_printf("r300: Implementation error: emit_vs_constant_buffer called,"
-        " but has_tcl is FALSE!\n");
-        return;
-    }
-
-    if (constants->Count == 0)
-        return;
-
     BEGIN_CS(constants->Count * 4 + 3);
     OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
                (r300screen->caps->is_r500 ?
                R500_PVS_CONST_START : R300_PVS_CONST_START));
     OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->Count * 4);
     for (i = 0; i < constants->Count; i++) {
-        const float * data = get_shader_constant(r300,
-                                                 &constants->Constants[i],
-                                                 &r300->shader_constants[PIPE_SHADER_VERTEX]);
+        const float *data = get_shader_constant(r300,
+                                                &constants->Constants[i],
+                                                &r300->shader_constants[PIPE_SHADER_VERTEX]);
         OUT_CS_32F(data[0]);
         OUT_CS_32F(data[1]);
         OUT_CS_32F(data[2]);
@@ -1043,9 +1026,11 @@ validate:
     }
     /* ...textures... */
     for (i = 0; i < texstate->count; i++) {
-        tex = texstate->textures[i];
-        if (!tex || !texstate->sampler_states[i])
+        if (!(texstate->tx_enable & (1 << i))) {
             continue;
+        }
+
+        tex = (struct r300_texture*)texstate->fragment_sampler_views[i]->texture;
         if (!r300_add_texture(r300->rws, tex,
 			      RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
             r300->context.flush(&r300->context, 0, NULL);
@@ -1092,8 +1077,8 @@ validate:
         r300->context.flush(&r300->context, 0, NULL);
         if (invalid) {
             /* Well, hell. */
-            debug_printf("r300: Stuck in validation loop, gonna quit now.");
-            exit(1);
+            fprintf(stderr, "r300: Stuck in validation loop, gonna quit now.\n");
+            abort();
         }
         invalid = TRUE;
         goto validate;
@@ -1158,7 +1143,9 @@ void r300_emit_dirty_state(struct r300_context* r300)
 
     if (r300->dirty_state & R300_NEW_VERTEX_SHADER_CONSTANTS) {
         struct r300_vertex_shader* vs = r300->vs_state.state;
-        r300_emit_vs_constant_buffer(r300, &vs->code.constants);
+        if (vs->code.constants.Count) {
+            r300_emit_vs_constant_buffer(r300, &vs->code.constants);
+        }
         r300->dirty_state &= ~R300_NEW_VERTEX_SHADER_CONSTANTS;
     }
 
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
index 9e71e61c303..e23fef8c9f7 100644
--- a/src/gallium/drivers/r300/r300_fs.c
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -204,9 +204,8 @@ static void r300_translate_fragment_shader(
     r3xx_compile_fragment_program(&compiler);
     if (compiler.Base.Error) {
         /* XXX failover maybe? */
-        DBG(r300, DBG_FP, "r300: Error compiling fragment program: %s\n",
-            compiler.Base.ErrorMsg);
-        assert(0);
+        fprintf(stderr, "r300 FP: Compiler Error:\n%s",
+                compiler.Base.ErrorMsg);
         abort();
     }
 
diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c
index ca00b043c51..f8b52d593d5 100644
--- a/src/gallium/drivers/r300/r300_query.c
+++ b/src/gallium/drivers/r300/r300_query.c
@@ -30,6 +30,8 @@
 #include "r300_query.h"
 #include "r300_reg.h"
 
+#include <stdio.h>
+
 static struct pipe_query *r300_create_query(struct pipe_context *pipe,
                                             unsigned query_type)
 {
@@ -97,8 +99,10 @@ static void r300_end_query(struct pipe_context* pipe,
 	                   struct pipe_query* query)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct r300_query* q = (struct r300_query*)query;
 
     r300_emit_query_end(r300);
+    q->begin_emitted = false;
     r300->query_current = NULL;
 }
 
@@ -135,8 +139,8 @@ static boolean r300_get_query_result(struct pipe_context* pipe,
         if (*map == ~0U) {
             /* Looks like our results aren't ready yet. */
             if (wait) {
-                debug_printf("r300: Despite waiting, OQ results haven't"
-                        " come in yet.\n");
+                fprintf(stderr, "r300: Despite waiting, OQ results haven't "
+                                "come in yet.\n");
             }
             temp = ~0U;
             break;
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index 47100c83b0b..1fb7eac2b34 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -41,9 +41,6 @@
 #include "r300_render.h"
 #include "r300_state_derived.h"
 
-/* r300_render: Vertex and index buffer primitive emission. */
-#define R300_MAX_VBO_SIZE  (1024 * 1024)
-
 /* XXX The DRM rejects VAP_ALT_NUM_VERTICES.. */
 //#define ENABLE_ALT_NUM_VERTS
 
@@ -141,7 +138,7 @@ static boolean immd_is_good_idea(struct r300_context *r300,
     unsigned vertex_element_count = r300->velems->count;
     unsigned i, vbi;
 
-    if (count > 4) {
+    if (count > 10) {
         return FALSE;
     }
 
@@ -155,8 +152,7 @@ static boolean immd_is_good_idea(struct r300_context *r300,
         if (!checked[vbi]) {
             vbuf = &r300->vertex_buffer[vbi];
 
-            if (r300_buffer_is_referenced(r300,
-					  vbuf->buffer)) {
+            if (r300_buffer_is_referenced(r300, vbuf->buffer)) {
                 /* It's a very bad idea to map it... */
                 return FALSE;
             }
@@ -304,10 +300,9 @@ static void r300_emit_draw_elements(struct r300_context *r300,
 #endif
     CS_LOCALS(r300);
 
-    assert((start * indexSize) % 4 == 0);
     assert(count < (1 << 24));
 
-    maxIndex = MIN3(maxIndex, r300->vertex_buffer_max_index, count - minIndex);
+    maxIndex = MIN2(maxIndex, r300->vertex_buffer_max_index);
 
     DBG(r300, DBG_DRAW, "r300: Indexbuf of %u indices, min %u max %u\n",
         count, minIndex, maxIndex);
@@ -354,6 +349,7 @@ static void r300_emit_draw_elements(struct r300_context *r300,
 
 static void r300_shorten_ubyte_elts(struct r300_context* r300,
                                     struct pipe_buffer** elts,
+                                    unsigned start,
                                     unsigned count)
 {
     struct pipe_screen* screen = r300->context.screen;
@@ -371,6 +367,8 @@ static void r300_shorten_ubyte_elts(struct r300_context* r300,
     in_map = pipe_buffer_map(screen, *elts, PIPE_BUFFER_USAGE_CPU_READ);
     out_map = pipe_buffer_map(screen, new_elts, PIPE_BUFFER_USAGE_CPU_WRITE);
 
+    in_map += start;
+
     for (i = 0; i < count; i++) {
         *out_map = (unsigned short)*in_map;
         in_map++;
@@ -383,6 +381,32 @@ static void r300_shorten_ubyte_elts(struct r300_context* r300,
     *elts = new_elts;
 }
 
+static void r300_align_ushort_elts(struct r300_context *r300,
+                                   struct pipe_buffer **elts,
+                                   unsigned start, unsigned count)
+{
+    struct pipe_screen* screen = r300->context.screen;
+    struct pipe_buffer* new_elts;
+    unsigned short *in_map;
+    unsigned short *out_map;
+
+    new_elts = screen->buffer_create(screen, 32,
+                                     PIPE_BUFFER_USAGE_INDEX |
+                                     PIPE_BUFFER_USAGE_CPU_WRITE |
+                                     PIPE_BUFFER_USAGE_GPU_READ,
+                                     2 * count);
+
+    in_map = pipe_buffer_map(screen, *elts, PIPE_BUFFER_USAGE_CPU_READ);
+    out_map = pipe_buffer_map(screen, new_elts, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+    memcpy(out_map, in_map+start, 2 * count);
+
+    pipe_buffer_unmap(screen, *elts);
+    pipe_buffer_unmap(screen, new_elts);
+
+    *elts = new_elts;
+}
+
 /* This is the fast-path drawing & emission for HW TCL. */
 void r300_draw_range_elements(struct pipe_context* pipe,
                               struct pipe_buffer* indexBuffer,
@@ -408,8 +432,12 @@ void r300_draw_range_elements(struct pipe_context* pipe,
     }
 
     if (indexSize == 1) {
-        r300_shorten_ubyte_elts(r300, &indexBuffer, count);
+        r300_shorten_ubyte_elts(r300, &indexBuffer, start, count);
         indexSize = 2;
+        start = 0;
+    } else if (indexSize == 2 && start % 2 != 0) {
+        r300_align_ushort_elts(r300, &indexBuffer, start, count);
+        start = 0;
     }
 
     r300_update_derived_state(r300);
@@ -541,13 +569,6 @@ void r300_swtcl_draw_arrays(struct pipe_context* pipe,
 
     draw_set_mapped_element_buffer(r300->draw, 0, NULL);
 
-    draw_set_mapped_constant_buffer(r300->draw,
-				    PIPE_SHADER_VERTEX,
-                                    0,
-				    r300->shader_constants[PIPE_SHADER_VERTEX].constants,
-				    r300->shader_constants[PIPE_SHADER_VERTEX].count *
-                (sizeof(float) * 4));
-
     draw_arrays(r300->draw, mode, start, count);
 
     for (i = 0; i < r300->vertex_buffer_count; i++) {
@@ -586,13 +607,6 @@ void r300_swtcl_draw_range_elements(struct pipe_context* pipe,
     draw_set_mapped_element_buffer_range(r300->draw, indexSize,
                                          minIndex, maxIndex, indices);
 
-    draw_set_mapped_constant_buffer(r300->draw,
-				    PIPE_SHADER_VERTEX,
-                                    0,
-            r300->shader_constants[PIPE_SHADER_VERTEX].constants,
-            r300->shader_constants[PIPE_SHADER_VERTEX].count *
-                (sizeof(float) * 4));
-
     draw_arrays(r300->draw, mode, start, count);
 
     for (i = 0; i < r300->vertex_buffer_count; i++) {
@@ -658,9 +672,9 @@ static boolean r300_render_allocate_vertices(struct vbuf_render* render,
         r300render->vbo = pipe_buffer_create(screen,
                                              64,
                                              PIPE_BUFFER_USAGE_VERTEX,
-                                             R300_MAX_VBO_SIZE);
+                                             R300_MAX_DRAW_VBO_SIZE);
         r300render->vbo_offset = 0;
-        r300render->vbo_size = R300_MAX_VBO_SIZE;
+        r300render->vbo_size = R300_MAX_DRAW_VBO_SIZE;
     }
 
     r300render->vertex_size = vertex_size;
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index 3e31688f8e8..e46f836dd2c 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -26,10 +26,8 @@
 
 #include "r300_context.h"
 #include "r300_texture.h"
-
-#include "radeon_winsys.h"
-
 #include "r300_screen_buffer.h"
+#include "r300_winsys.h"
 
 /* Return the identifier behind whom the brave coders responsible for this
  * amalgamation of code, sweat, and duct tape, routinely obscure their names.
@@ -166,7 +164,7 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
         case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER:
             return 0;
         default:
-            debug_printf("r300: Implementation error: Bad param %d\n",
+            fprintf(stderr, "r300: Implementation error: Bad param %d\n",
                 param);
             return 0;
     }
@@ -195,7 +193,7 @@ static float r300_get_paramf(struct pipe_screen* pscreen, int param)
         case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
             return 16.0f;
         default:
-            debug_printf("r300: Implementation error: Bad paramf %d\n",
+            fprintf(stderr, "r300: Implementation error: Bad paramf %d\n",
                 param);
             return 0.0f;
     }
@@ -214,7 +212,7 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
     boolean is_color2101010 = format == PIPE_FORMAT_R10G10B10A2_UNORM;
 
     if (target >= PIPE_MAX_TEXTURE_TYPES) {
-        debug_printf("r300: Implementation error: Received bogus texture "
+        fprintf(stderr, "r300: Implementation error: Received bogus texture "
             "target %d in %s\n", target, __FUNCTION__);
         return FALSE;
     }
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index 1ccc0bfb7a5..1cf8b7452d7 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -28,9 +28,7 @@
 
 #include "r300_chipset.h"
 
-#define R300_TEXTURE_USAGE_TRANSFER PIPE_TEXTURE_USAGE_CUSTOM
-
-struct radeon_winsys;
+#include <stdio.h>
 
 struct r300_screen {
     /* Parent class */
@@ -84,7 +82,7 @@ static INLINE void SCREEN_DBG(struct r300_screen * screen, unsigned flags,
     if (SCREEN_DBG_ON(screen, flags)) {
         va_list va;
         va_start(va, fmt);
-        debug_vprintf(fmt, va);
+        vfprintf(stderr, fmt, va);
         va_end(va);
     }
 }
diff --git a/src/gallium/drivers/r300/r300_screen_buffer.c b/src/gallium/drivers/r300/r300_screen_buffer.c
index b97d0d76a4a..a1cd48ee730 100644
--- a/src/gallium/drivers/r300/r300_screen_buffer.c
+++ b/src/gallium/drivers/r300/r300_screen_buffer.c
@@ -237,7 +237,7 @@ r300_buffer_map_range(struct pipe_screen *screen,
 	}
     }
 just_map:
-    map = rws->buffer_map(rws, rbuf->buf, usage | R300_USAGE_FLAG_DONT_SYNC);
+    map = rws->buffer_map(rws, rbuf->buf, usage);
    
     return map;
 }
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 712e9280e3c..39e05583dd4 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -38,8 +38,7 @@
 #include "r300_state_inlines.h"
 #include "r300_fs.h"
 #include "r300_vs.h"
-
-#include "radeon_winsys.h"
+#include "r300_winsys.h"
 
 /* r300_state: Functions used to intialize state context by translating
  * Gallium state objects into semi-native r300 state objects. */
@@ -528,8 +527,8 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         if (tex) {
             r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                             tex->pitch[0],
-                                            tex->microtile != 0,
-                                            tex->macrotile != 0);
+                                            tex->microtile,
+                                            tex->macrotile);
         }
     }
     if (old_state->zsbuf &&
@@ -540,8 +539,8 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         if (tex) {
             r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                             tex->pitch[0],
-                                            tex->microtile != 0,
-                                            tex->macrotile != 0);
+                                            tex->microtile,
+                                            tex->macrotile);
         }
     }
 
@@ -552,8 +551,8 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
 
         r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                         tex->pitch[level],
-                                        tex->microtile != 0,
-                                        tex->mip_macrotile[level] != 0);
+                                        tex->microtile,
+                                        tex->mip_macrotile[level]);
     }
     if (new_state->zsbuf) {
         tex = (struct r300_texture*)new_state->zsbuf->texture;
@@ -561,8 +560,8 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
 
         r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                                         tex->pitch[level],
-                                        tex->microtile != 0,
-                                        tex->mip_macrotile[level] != 0);
+                                        tex->microtile,
+                                        tex->mip_macrotile[level]);
     }
 }
 
@@ -576,9 +575,8 @@ static void
     unsigned max_width, max_height;
     uint32_t zbuffer_bpp = 0;
 
-
     if (state->nr_cbufs > 4) {
-        debug_printf("r300: Implementation error: Too many MRTs in %s, "
+        fprintf(stderr, "r300: Implementation error: Too many MRTs in %s, "
             "refusing to bind framebuffer state!\n", __FUNCTION__);
         return;
     }
@@ -592,7 +590,7 @@ static void
     }
 
     if (state->width > max_width || state->height > max_height) {
-        debug_printf("r300: Implementation error: Render targets are too "
+        fprintf(stderr, "r300: Implementation error: Render targets are too "
         "big in %s, refusing to bind framebuffer state!\n", __FUNCTION__);
         return;
     }
@@ -933,13 +931,14 @@ static void r300_delete_sampler_state(struct pipe_context* pipe, void* state)
     FREE(state);
 }
 
-static void r300_set_sampler_textures(struct pipe_context* pipe,
-                                      unsigned count,
-                                      struct pipe_texture** texture)
+static void r300_set_fragment_sampler_views(struct pipe_context* pipe,
+                                            unsigned count,
+                                            struct pipe_sampler_view** views)
 {
     struct r300_context* r300 = r300_context(pipe);
     struct r300_textures_state* state =
         (struct r300_textures_state*)r300->textures_state.state;
+    struct r300_texture *texture;
     unsigned i;
     boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
     boolean dirty_tex = FALSE;
@@ -950,13 +949,20 @@ static void r300_set_sampler_textures(struct pipe_context* pipe,
     }
 
     for (i = 0; i < count; i++) {
-        if (state->textures[i] != (struct r300_texture*)texture[i]) {
-            pipe_texture_reference((struct pipe_texture**)&state->textures[i],
-                                   texture[i]);
+        if (state->fragment_sampler_views[i] != views[i]) {
+            pipe_sampler_view_reference(&state->fragment_sampler_views[i],
+                                        views[i]);
+
+            if (!views[i]) {
+                continue;
+            }
+
+            /* A new sampler view (= texture)... */
             dirty_tex = TRUE;
 
             /* R300-specific - set the texrect factor in the fragment shader */
-            if (!is_r500 && state->textures[i]->is_npot) {
+            texture = (struct r300_texture *)views[i]->texture;
+            if (!is_r500 && texture->is_npot) {
                 /* XXX It would be nice to re-emit just 1 constant,
                  * XXX not all of them */
                 r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
@@ -965,9 +971,9 @@ static void r300_set_sampler_textures(struct pipe_context* pipe,
     }
 
     for (i = count; i < 8; i++) {
-        if (state->textures[i]) {
-            pipe_texture_reference((struct pipe_texture**)&state->textures[i],
-                NULL);
+        if (state->fragment_sampler_views[i]) {
+            pipe_sampler_view_reference(&state->fragment_sampler_views[i],
+                                        NULL);
         }
     }
 
@@ -980,6 +986,32 @@ static void r300_set_sampler_textures(struct pipe_context* pipe,
     }
 }
 
+static struct pipe_sampler_view *
+r300_create_sampler_view(struct pipe_context *pipe,
+                         struct pipe_texture *texture,
+                         const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+static void
+r300_sampler_view_destroy(struct pipe_context *pipe,
+                          struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
+}
+
 static void r300_set_scissor_state(struct pipe_context* pipe,
                                    const struct pipe_scissor_state* state)
 {
@@ -1041,117 +1073,71 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
                                     const struct pipe_vertex_buffer* buffers)
 {
     struct r300_context* r300 = r300_context(pipe);
-    int i;
-    unsigned max_index = (1 << 24) - 1;
-    boolean any_user_buffer = false;
+    struct pipe_vertex_buffer *vbo;
+    unsigned i, max_index = (1 << 24) - 1;
+    boolean any_user_buffer = FALSE;
 
     if (count == r300->vertex_buffer_count &&
-	memcmp(r300->vertex_buffer, buffers, count * sizeof(buffers[0])) == 0)
+        memcmp(r300->vertex_buffer, buffers,
+            sizeof(struct pipe_vertex_buffer) * count) == 0) {
         return;
+    }
 
+    /* Check if the stride is aligned to the size of DWORD. */
     for (i = 0; i < count; i++) {
-	pipe_buffer_reference(&r300->vertex_buffer[i].buffer, buffers[i].buffer);
-	if (r300_buffer_is_user_buffer(buffers[i].buffer))
-	    any_user_buffer = true;
-        max_index = MIN2(buffers[i].max_index, max_index);
+        if (buffers[i].buffer) {
+            if (buffers[i].stride % 4 != 0) {
+                // XXX Shouldn't we align the buffer?
+                fprintf(stderr, "r300_set_vertex_buffers: "
+                        "Unaligned buffer stride %i isn't supported.\n",
+                        buffers[i].stride);
+                assert(0);
+                abort();
+            }
+        }
     }
 
-    for ( ; i < r300->vertex_buffer_count; i++)
-	pipe_buffer_reference(&r300->vertex_buffer[i].buffer, NULL);
-
-    memcpy(r300->vertex_buffer, buffers,
-	   sizeof(struct pipe_vertex_buffer) * count);
+    for (i = 0; i < count; i++) {
+        /* Why, yes, I AM casting away constness. How did you know? */
+        vbo = (struct pipe_vertex_buffer*)&buffers[i];
 
-    r300->vertex_buffer_count = count;
-    r300->vertex_buffer_max_index = max_index;
-    r300->any_user_vbs = any_user_buffer;
+        /* Reference our buffer. */
+        pipe_buffer_reference(&r300->vertex_buffer[i].buffer, vbo->buffer);
 
-    if (r300->draw) {
-        draw_flush(r300->draw);
-        draw_set_vertex_buffers(r300->draw, count, buffers);
-    }
-}
-
-static boolean r300_validate_aos(struct r300_context *r300)
-{
-    struct pipe_vertex_buffer *vbuf = r300->vertex_buffer;
-    struct pipe_vertex_element *velem = r300->velems->velem;
-    int i;
-
-    /* Check if formats and strides are aligned to the size of DWORD. */
-    for (i = 0; i < r300->velems->count; i++) {
-        if (vbuf[velem[i].vertex_buffer_index].stride % 4 != 0 ||
-            util_format_get_blocksize(velem[i].src_format) % 4 != 0) {
-            return FALSE;
+        /* Skip NULL buffers */
+        if (!buffers[i].buffer) {
+            continue;
         }
-    }
-    return TRUE;
-}
 
-static void r300_draw_emit_attrib(struct r300_context* r300,
-                                  enum attrib_emit emit,
-                                  enum interp_mode interp,
-                                  int index)
-{
-    struct r300_vertex_shader* vs = r300->vs_state.state;
-    struct tgsi_shader_info* info = &vs->info;
-    int output;
-
-    output = draw_find_shader_output(r300->draw,
-                                     info->output_semantic_name[index],
-                                     info->output_semantic_index[index]);
-    draw_emit_vertex_attr(&r300->vertex_info, emit, interp, output);
-}
+        if (r300_buffer_is_user_buffer(vbo->buffer)) {
+            any_user_buffer = TRUE;
+        }
 
-static void r300_draw_emit_all_attribs(struct r300_context* r300)
-{
-    struct r300_vertex_shader* vs = r300->vs_state.state;
-    struct r300_shader_semantics* vs_outputs = &vs->outputs;
-    int i, gen_count;
-
-    /* Position. */
-    if (vs_outputs->pos != ATTR_UNUSED) {
-        r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
-                              vs_outputs->pos);
-    } else {
-        assert(0);
-    }
+        if (vbo->max_index == ~0) {
+            /* Bogus value from broken state tracker; hax it. */
+            vbo->max_index =
+                (vbo->buffer->size - vbo->buffer_offset) / vbo->stride;
+        }
 
-    /* Point size. */
-    if (vs_outputs->psize != ATTR_UNUSED) {
-        r300_draw_emit_attrib(r300, EMIT_1F_PSIZE, INTERP_POS,
-                              vs_outputs->psize);
+        max_index = MIN2(vbo->max_index, max_index);
     }
 
-    /* Colors. */
-    for (i = 0; i < ATTR_COLOR_COUNT; i++) {
-        if (vs_outputs->color[i] != ATTR_UNUSED) {
-            r300_draw_emit_attrib(r300, EMIT_4F, INTERP_LINEAR,
-                                  vs_outputs->color[i]);
-        }
+    for (; i < r300->vertex_buffer_count; i++) {
+        /* Dereference any old buffers. */
+        pipe_buffer_reference(&r300->vertex_buffer[i].buffer, NULL);
     }
 
-    /* XXX Back-face colors. */
+    memcpy(r300->vertex_buffer, buffers,
+        sizeof(struct pipe_vertex_buffer) * count);
 
-    /* Texture coordinates. */
-    gen_count = 0;
-    for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
-        if (vs_outputs->generic[i] != ATTR_UNUSED) {
-            r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
-                                  vs_outputs->generic[i]);
-            gen_count++;
-        }
-    }
+    r300->vertex_buffer_count = count;
+    r300->vertex_buffer_max_index = max_index;
+    r300->any_user_vbs = any_user_buffer;
 
-    /* Fog coordinates. */
-    if (vs_outputs->fog != ATTR_UNUSED) {
-        r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
-                              vs_outputs->fog);
-        gen_count++;
+    if (r300->draw) {
+        draw_flush(r300->draw);
+        draw_set_vertex_buffers(r300->draw, count, buffers);
     }
-
-    /* XXX magic */
-    assert(gen_count <= 8);
 }
 
 /* Update the PSC tables. */
@@ -1193,70 +1179,13 @@ static void r300_vertex_psc(struct r300_vertex_element_state *velems)
     vstream->count = (i >> 1) + 1;
 }
 
-/* Update the PSC tables for SW TCL, using Draw. */
-static void r300_swtcl_vertex_psc(struct r300_context *r300,
-                                  struct r300_vertex_element_state *velems)
-{
-    struct r300_vertex_stream_state *vstream = &velems->vertex_stream;
-    struct r300_vertex_shader* vs = r300->vs_state.state;
-    struct vertex_info* vinfo = &r300->vertex_info;
-    uint16_t type, swizzle;
-    enum pipe_format format;
-    unsigned i, attrib_count;
-    int* vs_output_tab = vs->stream_loc_notcl;
-
-    /* For each Draw attribute, route it to the fragment shader according
-     * to the vs_output_tab. */
-    attrib_count = vinfo->num_attribs;
-    DBG(r300, DBG_DRAW, "r300: attrib count: %d\n", attrib_count);
-    for (i = 0; i < attrib_count; i++) {
-        DBG(r300, DBG_DRAW, "r300: attrib: offset %d, interp %d, size %d,"
-               " vs_output_tab %d\n", vinfo->attrib[i].src_index,
-               vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit,
-               vs_output_tab[i]);
-    }
-
-    for (i = 0; i < attrib_count; i++) {
-        /* Make sure we have a proper destination for our attribute. */
-        assert(vs_output_tab[i] != -1);
-
-        format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
-
-        /* Obtain the type of data in this attribute. */
-        type = r300_translate_vertex_data_type(format) |
-            vs_output_tab[i] << R300_DST_VEC_LOC_SHIFT;
-
-        /* Obtain the swizzle for this attribute. Note that the default
-         * swizzle in the hardware is not XYZW! */
-        swizzle = r300_translate_vertex_data_swizzle(format);
-
-        /* Add the attribute to the PSC table. */
-        if (i & 1) {
-            vstream->vap_prog_stream_cntl[i >> 1] |= type << 16;
-            vstream->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 16;
-        } else {
-            vstream->vap_prog_stream_cntl[i >> 1] |= type;
-            vstream->vap_prog_stream_cntl_ext[i >> 1] |= swizzle;
-        }
-    }
-
-    /* Set the last vector in the PSC. */
-    if (i) {
-        i -= 1;
-    }
-    vstream->vap_prog_stream_cntl[i >> 1] |=
-        (R300_LAST_VEC << (i & 1 ? 16 : 0));
-
-    vstream->count = (i >> 1) + 1;
-}
-
 static void* r300_create_vertex_elements_state(struct pipe_context* pipe,
                                                unsigned count,
                                                const struct pipe_vertex_element* attribs)
 {
-    struct r300_context *r300 = r300_context(pipe);
     struct r300_screen* r300screen = r300_screen(pipe->screen);
     struct r300_vertex_element_state *velems;
+    unsigned i, size;
 
     assert(count <= PIPE_MAX_ATTRIBS);
     velems = CALLOC_STRUCT(r300_vertex_element_state);
@@ -1265,12 +1194,21 @@ static void* r300_create_vertex_elements_state(struct pipe_context* pipe,
         memcpy(velems->velem, attribs, sizeof(struct pipe_vertex_element) * count);
 
         if (r300screen->caps->has_tcl) {
+            /* Check if the format is aligned to the size of DWORD. */
+            for (i = 0; i < count; i++) {
+                size = util_format_get_blocksize(attribs[i].src_format);
+
+                if (size % 4 != 0) {
+                    /* XXX Shouldn't we align the format? */
+                    fprintf(stderr, "r300_create_vertex_elements_state: "
+                            "Unaligned format %s:%i isn't supported\n",
+                            util_format_name(attribs[i].src_format), size);
+                    assert(0);
+                    abort();
+                }
+            }
+
             r300_vertex_psc(velems);
-        } else {
-            memset(&r300->vertex_info, 0, sizeof(struct vertex_info));
-            r300_draw_emit_all_attribs(r300);
-            draw_compute_vertex_size(&r300->vertex_info);
-            r300_swtcl_vertex_psc(r300, velems);
         }
     }
     return velems;
@@ -1293,12 +1231,6 @@ static void r300_bind_vertex_elements_state(struct pipe_context *pipe,
         draw_set_vertex_elements(r300->draw, velems->count, velems->velem);
     }
 
-    if (!r300_validate_aos(r300)) {
-        /* XXX We should fallback using draw. */
-        assert(0);
-        abort();
-    }
-
     UPDATE_STATE(&velems->vertex_stream, r300->vertex_stream_state);
     r300->vertex_stream_state.size = (1 + velems->vertex_stream.count) * 2;
 }
@@ -1419,7 +1351,7 @@ static void r300_set_constant_buffer(struct pipe_context *pipe,
 
     /* XXX Subtract immediates and RC_STATE_* variables. */
     if (buf->size > (sizeof(float) * 4 * max_size)) {
-        debug_printf("r300: Max size of the constant buffer is "
+        fprintf(stderr, "r300: Max size of the constant buffer is "
                       "%i*4 floats.\n", max_size);
         abort();
     }
@@ -1432,10 +1364,14 @@ static void r300_set_constant_buffer(struct pipe_context *pipe,
         if (r300screen->caps->has_tcl) {
             r300->dirty_state |= R300_NEW_VERTEX_SHADER_CONSTANTS;
             r300->pvs_flush.dirty = TRUE;
+        } else if (r300->draw) {
+            draw_set_mapped_constant_buffer(r300->draw, PIPE_SHADER_VERTEX,
+                0, r300->shader_constants[PIPE_SHADER_VERTEX].constants,
+                buf->size);
         }
-    }
-    else if (shader == PIPE_SHADER_FRAGMENT)
+    } else if (shader == PIPE_SHADER_FRAGMENT) {
         r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+    }
 }
 
 void r300_init_state_functions(struct r300_context* r300)
@@ -1473,7 +1409,9 @@ void r300_init_state_functions(struct r300_context* r300)
     r300->context.bind_vertex_sampler_states = r300_lacks_vertex_textures;
     r300->context.delete_sampler_state = r300_delete_sampler_state;
 
-    r300->context.set_fragment_sampler_textures = r300_set_sampler_textures;
+    r300->context.set_fragment_sampler_views = r300_set_fragment_sampler_views;
+    r300->context.create_sampler_view = r300_create_sampler_view;
+    r300->context.sampler_view_destroy = r300_sampler_view_destroy;
 
     r300->context.set_scissor_state = r300_set_scissor_state;
 
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 6b9f61acd7b..bc5431c802d 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -37,6 +37,131 @@
 /* r300_state_derived: Various bits of state which are dependent upon
  * currently bound CSO data. */
 
+static void r300_draw_emit_attrib(struct r300_context* r300,
+                                  enum attrib_emit emit,
+                                  enum interp_mode interp,
+                                  int index)
+{
+    struct r300_vertex_shader* vs = r300->vs_state.state;
+    struct tgsi_shader_info* info = &vs->info;
+    int output;
+
+    output = draw_find_shader_output(r300->draw,
+                                     info->output_semantic_name[index],
+                                     info->output_semantic_index[index]);
+    draw_emit_vertex_attr(&r300->vertex_info, emit, interp, output);
+}
+
+static void r300_draw_emit_all_attribs(struct r300_context* r300)
+{
+    struct r300_vertex_shader* vs = r300->vs_state.state;
+    struct r300_shader_semantics* vs_outputs = &vs->outputs;
+    int i, gen_count;
+
+    /* Position. */
+    if (vs_outputs->pos != ATTR_UNUSED) {
+        r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
+                              vs_outputs->pos);
+    } else {
+        assert(0);
+    }
+
+    /* Point size. */
+    if (vs_outputs->psize != ATTR_UNUSED) {
+        r300_draw_emit_attrib(r300, EMIT_1F_PSIZE, INTERP_POS,
+                              vs_outputs->psize);
+    }
+
+    /* Colors. */
+    for (i = 0; i < ATTR_COLOR_COUNT; i++) {
+        if (vs_outputs->color[i] != ATTR_UNUSED) {
+            r300_draw_emit_attrib(r300, EMIT_4F, INTERP_LINEAR,
+                                  vs_outputs->color[i]);
+        }
+    }
+
+    /* XXX Back-face colors. */
+
+    /* Texture coordinates. */
+    gen_count = 0;
+    for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
+        if (vs_outputs->generic[i] != ATTR_UNUSED) {
+            r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
+                                  vs_outputs->generic[i]);
+            gen_count++;
+        }
+    }
+
+    /* Fog coordinates. */
+    if (vs_outputs->fog != ATTR_UNUSED) {
+        r300_draw_emit_attrib(r300, EMIT_4F, INTERP_PERSPECTIVE,
+                              vs_outputs->fog);
+        gen_count++;
+    }
+
+    /* XXX magic */
+    assert(gen_count <= 8);
+}
+
+/* Update the PSC tables for SW TCL, using Draw. */
+static void r300_swtcl_vertex_psc(struct r300_context *r300)
+{
+    struct r300_vertex_stream_state *vstream = r300->vertex_stream_state.state;
+    struct r300_vertex_shader* vs = r300->vs_state.state;
+    struct vertex_info* vinfo = &r300->vertex_info;
+    uint16_t type, swizzle;
+    enum pipe_format format;
+    unsigned i, attrib_count;
+    int* vs_output_tab = vs->stream_loc_notcl;
+
+    /* XXX hax */
+    memset(vstream, 0, sizeof(struct r300_vertex_stream_state));
+
+    /* For each Draw attribute, route it to the fragment shader according
+     * to the vs_output_tab. */
+    attrib_count = vinfo->num_attribs;
+    DBG(r300, DBG_DRAW, "r300: attrib count: %d\n", attrib_count);
+    for (i = 0; i < attrib_count; i++) {
+        DBG(r300, DBG_DRAW, "r300: attrib: offset %d, interp %d, size %d,"
+               " vs_output_tab %d\n", vinfo->attrib[i].src_index,
+               vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit,
+               vs_output_tab[i]);
+
+        /* Make sure we have a proper destination for our attribute. */
+        assert(vs_output_tab[i] != -1);
+
+        format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
+
+        /* Obtain the type of data in this attribute. */
+        type = r300_translate_vertex_data_type(format) |
+            vs_output_tab[i] << R300_DST_VEC_LOC_SHIFT;
+
+        /* Obtain the swizzle for this attribute. Note that the default
+         * swizzle in the hardware is not XYZW! */
+        swizzle = r300_translate_vertex_data_swizzle(format);
+
+        /* Add the attribute to the PSC table. */
+        if (i & 1) {
+            vstream->vap_prog_stream_cntl[i >> 1] |= type << 16;
+            vstream->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 16;
+        } else {
+            vstream->vap_prog_stream_cntl[i >> 1] |= type;
+            vstream->vap_prog_stream_cntl_ext[i >> 1] |= swizzle;
+        }
+    }
+
+    /* Set the last vector in the PSC. */
+    if (i) {
+        i -= 1;
+    }
+    vstream->vap_prog_stream_cntl[i >> 1] |=
+        (R300_LAST_VEC << (i & 1 ? 16 : 0));
+
+    vstream->count = (i >> 1) + 1;
+    r300->vertex_stream_state.dirty = TRUE;
+    r300->vertex_stream_state.size = (1 + vstream->count) * 2;
+}
+
 static void r300_rs_col(struct r300_rs_block* rs, int id, int ptr,
                         boolean swizzle_0001)
 {
@@ -332,20 +457,25 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
         (struct r300_textures_state*)r300->textures_state.state;
     struct r300_texture_sampler_state *texstate;
     struct r300_sampler_state *sampler;
+    struct pipe_sampler_view *view;
     struct r300_texture *tex;
     unsigned min_level, max_level, i, size;
     unsigned count = MIN2(state->texture_count, state->sampler_count);
 
     state->tx_enable = 0;
+    state->count = 0;
     size = 2;
 
     for (i = 0; i < count; i++) {
-        if (state->textures[i] && state->sampler_states[i]) {
+        if (state->fragment_sampler_views[i] && state->sampler_states[i]) {
             state->tx_enable |= 1 << i;
 
-            tex = state->textures[i];
+            view = state->fragment_sampler_views[i];
+            tex = (struct r300_texture *)view->texture;
             sampler = state->sampler_states[i];
 
+            assert(view->format == tex->tex.format);
+
             texstate = &state->regs[i];
             memcpy(texstate->format, &tex->state, sizeof(uint32_t)*3);
             texstate->filter[0] = sampler->filter0;
@@ -367,8 +497,10 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
             } else {
                 /* determine min/max levels */
                 /* the MAX_MIP level is the largest (finest) one */
-                max_level = MIN2(sampler->max_lod, tex->tex.last_level);
-                min_level = MIN2(sampler->min_lod, max_level);
+                max_level = MIN3(sampler->max_lod + view->first_level,
+                                 tex->tex.last_level, view->last_level);
+                min_level = MIN2(sampler->min_lod + view->first_level,
+                                 max_level);
                 texstate->format[0] |= R300_TX_NUM_LEVELS(max_level);
                 texstate->filter[0] |= R300_TX_MAX_MIP_LEVEL(min_level);
             }
@@ -393,5 +525,12 @@ void r300_update_derived_state(struct r300_context* r300)
         r300_merge_textures_and_samplers(r300);
     }
 
+    if (r300->draw) {
+        memset(&r300->vertex_info, 0, sizeof(struct vertex_info));
+        r300_draw_emit_all_attribs(r300);
+        draw_compute_vertex_size(&r300->vertex_info);
+        r300_swtcl_vertex_psc(r300);
+    }
+
     r300_update_ztop(r300);
 }
diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h
index 8485d4f8f94..02d09c008c8 100644
--- a/src/gallium/drivers/r300/r300_state_inlines.h
+++ b/src/gallium/drivers/r300/r300_state_inlines.h
@@ -32,6 +32,8 @@
 
 #include "r300_reg.h"
 
+#include <stdio.h>
+
 /* Some maths. These should probably find their way to u_math, if needed. */
 
 static INLINE int pack_float_16_6x(float f) {
@@ -54,7 +56,7 @@ static INLINE uint32_t r300_translate_blend_function(int blend_func)
         case PIPE_BLEND_MAX:
             return R300_COMB_FCN_MAX;
         default:
-            debug_printf("r300: Unknown blend function %d\n", blend_func);
+            fprintf(stderr, "r300: Unknown blend function %d\n", blend_func);
             assert(0);
             break;
     }
@@ -100,13 +102,13 @@ static INLINE uint32_t r300_translate_blend_factor(int blend_fact)
         case PIPE_BLENDFACTOR_SRC1_ALPHA:
         case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
         case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
-            debug_printf("r300: Implementation error: "
+            fprintf(stderr, "r300: Implementation error: "
                 "Bad blend factor %d not supported!\n", blend_fact);
             assert(0);
             break;
 
         default:
-            debug_printf("r300: Unknown blend factor %d\n", blend_fact);
+            fprintf(stderr, "r300: Unknown blend factor %d\n", blend_fact);
             assert(0);
             break;
     }
@@ -135,7 +137,7 @@ static INLINE uint32_t r300_translate_depth_stencil_function(int zs_func)
         case PIPE_FUNC_ALWAYS:
             return R300_ZS_ALWAYS;
         default:
-            debug_printf("r300: Unknown depth/stencil function %d\n",
+            fprintf(stderr, "r300: Unknown depth/stencil function %d\n",
                 zs_func);
             assert(0);
             break;
@@ -163,7 +165,7 @@ static INLINE uint32_t r300_translate_stencil_op(int s_op)
         case PIPE_STENCIL_OP_INVERT:
             return R300_ZS_INVERT;
         default:
-            debug_printf("r300: Unknown stencil op %d", s_op);
+            fprintf(stderr, "r300: Unknown stencil op %d", s_op);
             assert(0);
             break;
     }
@@ -190,7 +192,7 @@ static INLINE uint32_t r300_translate_alpha_function(int alpha_func)
         case PIPE_FUNC_ALWAYS:
             return R300_FG_ALPHA_FUNC_ALWAYS;
         default:
-            debug_printf("r300: Unknown alpha function %d", alpha_func);
+            fprintf(stderr, "r300: Unknown alpha function %d", alpha_func);
             assert(0);
             break;
     }
@@ -209,7 +211,7 @@ r300_translate_polygon_mode_front(unsigned mode) {
             return R300_GA_POLY_MODE_FRONT_PTYPE_POINT;
 
         default:
-            debug_printf("r300: Bad polygon mode %i in %s\n", mode,
+            fprintf(stderr, "r300: Bad polygon mode %i in %s\n", mode,
                 __FUNCTION__);
             return R300_GA_POLY_MODE_FRONT_PTYPE_TRI;
     }
@@ -227,7 +229,7 @@ r300_translate_polygon_mode_back(unsigned mode) {
             return R300_GA_POLY_MODE_BACK_PTYPE_POINT;
 
         default:
-            debug_printf("r300: Bad polygon mode %i in %s\n", mode,
+            fprintf(stderr, "r300: Bad polygon mode %i in %s\n", mode,
                 __FUNCTION__);
             return R300_GA_POLY_MODE_BACK_PTYPE_TRI;
     }
@@ -255,7 +257,7 @@ static INLINE uint32_t r300_translate_wrap(int wrap)
         case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
             return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED;
         default:
-            debug_printf("r300: Unknown texture wrap %d", wrap);
+            fprintf(stderr, "r300: Unknown texture wrap %d", wrap);
             assert(0);
             return 0;
     }
@@ -276,7 +278,7 @@ static INLINE uint32_t r300_translate_tex_filters(int min, int mag, int mip,
             retval |= R300_TX_MIN_FILTER_LINEAR;
             break;
         default:
-            debug_printf("r300: Unknown texture filter %d\n", min);
+            fprintf(stderr, "r300: Unknown texture filter %d\n", min);
             assert(0);
             break;
         }
@@ -288,7 +290,7 @@ static INLINE uint32_t r300_translate_tex_filters(int min, int mag, int mip,
             retval |= R300_TX_MAG_FILTER_LINEAR;
             break;
         default:
-            debug_printf("r300: Unknown texture filter %d\n", mag);
+            fprintf(stderr, "r300: Unknown texture filter %d\n", mag);
             assert(0);
             break;
         }
@@ -304,7 +306,7 @@ static INLINE uint32_t r300_translate_tex_filters(int min, int mag, int mip,
             retval |= R300_TX_MIN_FILTER_MIP_LINEAR;
             break;
         default:
-            debug_printf("r300: Unknown texture filter %d\n", mip);
+            fprintf(stderr, "r300: Unknown texture filter %d\n", mip);
             assert(0);
             break;
     }
@@ -370,7 +372,7 @@ r300_translate_vertex_data_type(enum pipe_format format) {
     desc = util_format_description(format);
 
     if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN) {
-        debug_printf("r300: Bad format %s in %s:%d\n", util_format_name(format),
+        fprintf(stderr, "r300: Bad format %s in %s:%d\n", util_format_name(format),
             __FUNCTION__, __LINE__);
         assert(0);
     }
@@ -391,7 +393,7 @@ r300_translate_vertex_data_type(enum pipe_format format) {
                     result = R300_DATA_TYPE_FLOAT_1 + (components - 1);
                     break;
                 default:
-                    debug_printf("r300: Bad format %s in %s:%d\n",
+                    fprintf(stderr, "r300: Bad format %s in %s:%d\n",
                         util_format_name(format), __FUNCTION__, __LINE__);
                     assert(0);
             }
@@ -412,15 +414,15 @@ r300_translate_vertex_data_type(enum pipe_format format) {
                     }
                     break;
                 default:
-                    debug_printf("r300: Bad format %s in %s:%d\n",
+                    fprintf(stderr, "r300: Bad format %s in %s:%d\n",
                         util_format_name(format), __FUNCTION__, __LINE__);
-                    debug_printf("r300: util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) == %d\n",
+                    fprintf(stderr, "r300: util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) == %d\n",
                         util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0));
                     assert(0);
             }
             break;
         default:
-            debug_printf("r300: Bad format %s in %s:%d\n",
+            fprintf(stderr, "r300: Bad format %s in %s:%d\n",
                 util_format_name(format), __FUNCTION__, __LINE__);
             assert(0);
     }
@@ -442,7 +444,7 @@ r300_translate_vertex_data_swizzle(enum pipe_format format) {
     assert(format);
 
     if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN) {
-        debug_printf("r300: Bad format %s in %s:%d\n",
+        fprintf(stderr, "r300: Bad format %s in %s:%d\n",
             util_format_name(format), __FUNCTION__, __LINE__);
         return 0;
     }
diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c
index 4a2c68269b1..88ae75bb816 100644
--- a/src/gallium/drivers/r300/r300_state_invariant.c
+++ b/src/gallium/drivers/r300/r300_state_invariant.c
@@ -127,7 +127,7 @@ void r300_emit_invariant_state(struct r300_context* r300,
     OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000);
 
     /* XXX */
-    OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa);
+    OUT_CS_REG(R300_SC_CLIP_RULE, 0xFFFF);
 
     END_CS;
 }
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index 7c7656068bb..22ccadfe3dd 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -31,8 +31,7 @@
 #include "r300_texture.h"
 #include "r300_screen.h"
 #include "r300_state_inlines.h"
-
-#include "radeon_winsys.h"
+#include "r300_winsys.h"
 
 #define TILE_WIDTH 0
 #define TILE_HEIGHT 1
@@ -46,6 +45,18 @@ static const unsigned microblock_table[5][3][2] = {
     {{ 2, 1}, {0, 0}, {0, 0}}  /* 128 bits per pixel */
 };
 
+/* Return true for non-compressed and non-YUV formats. */
+static boolean r300_format_is_plain(enum pipe_format format)
+{
+    const struct util_format_description *desc = util_format_description(format);
+
+    if (!format) {
+        return FALSE;
+    }
+
+    return desc->layout == UTIL_FORMAT_LAYOUT_PLAIN;
+}
+
 /* Translate a pipe_format into a useful texture format for sampling.
  *
  * Some special formats are translated directly using R300_EASY_TX_FORMAT,
@@ -148,7 +159,7 @@ static uint32_t r300_translate_texformat(enum pipe_format format)
     }
 
     /* Compressed formats. */
-    if (desc->layout == UTIL_FORMAT_LAYOUT_COMPRESSED) {
+    if (desc->layout == UTIL_FORMAT_LAYOUT_S3TC) {
         switch (format) {
             case PIPE_FORMAT_DXT1_RGB:
             case PIPE_FORMAT_DXT1_RGBA:
@@ -302,7 +313,6 @@ static uint32_t r300_translate_colorformat(enum pipe_format format)
         case PIPE_FORMAT_A8_UNORM:
         case PIPE_FORMAT_I8_UNORM:
         case PIPE_FORMAT_L8_UNORM:
-        case PIPE_FORMAT_L8_SRGB:
         case PIPE_FORMAT_R8_UNORM:
         case PIPE_FORMAT_R8_SNORM:
             return R300_COLOR_FORMAT_I8;
@@ -311,24 +321,19 @@ static uint32_t r300_translate_colorformat(enum pipe_format format)
         case PIPE_FORMAT_B5G6R5_UNORM:
             return R300_COLOR_FORMAT_RGB565;
         case PIPE_FORMAT_B5G5R5A1_UNORM:
+        case PIPE_FORMAT_B5G5R5X1_UNORM:
             return R300_COLOR_FORMAT_ARGB1555;
         case PIPE_FORMAT_B4G4R4A4_UNORM:
             return R300_COLOR_FORMAT_ARGB4444;
 
         /* 32-bit buffers. */
         case PIPE_FORMAT_B8G8R8A8_UNORM:
-        case PIPE_FORMAT_B8G8R8A8_SRGB:
         case PIPE_FORMAT_B8G8R8X8_UNORM:
-        case PIPE_FORMAT_B8G8R8X8_SRGB:
         case PIPE_FORMAT_A8R8G8B8_UNORM:
-        case PIPE_FORMAT_A8R8G8B8_SRGB:
         case PIPE_FORMAT_X8R8G8B8_UNORM:
-        case PIPE_FORMAT_X8R8G8B8_SRGB:
         case PIPE_FORMAT_A8B8G8R8_UNORM:
         case PIPE_FORMAT_R8G8B8A8_SNORM:
-        case PIPE_FORMAT_A8B8G8R8_SRGB:
         case PIPE_FORMAT_X8B8G8R8_UNORM:
-        case PIPE_FORMAT_X8B8G8R8_SRGB:
         case PIPE_FORMAT_R8SG8SB8UX8U_NORM:
             return R300_COLOR_FORMAT_ARGB8888;
         case PIPE_FORMAT_R10G10B10A2_UNORM:
@@ -393,12 +398,7 @@ static uint32_t r300_translate_out_fmt(enum pipe_format format)
     desc = util_format_description(format);
 
     /* Specifies how the shader output is written to the fog unit. */
-    if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) {
-        /* The gamma correction causes precision loss so we need
-         * higher precision to maintain reasonable quality.
-         * It has nothing to do with the colorbuffer format. */
-        modifier |= R300_US_OUT_FMT_C4_10_GAMMA;
-    } else if (desc->channel[0].type == UTIL_FORMAT_TYPE_FLOAT) {
+    if (desc->channel[0].type == UTIL_FORMAT_TYPE_FLOAT) {
         if (desc->channel[0].size == 32) {
             modifier |= R300_US_OUT_FMT_C4_32_FP;
         } else {
@@ -428,46 +428,39 @@ static uint32_t r300_translate_out_fmt(enum pipe_format format)
             return modifier | R300_C2_SEL_A;
         case PIPE_FORMAT_I8_UNORM:
         case PIPE_FORMAT_L8_UNORM:
-        case PIPE_FORMAT_L8_SRGB:
         case PIPE_FORMAT_R8_UNORM:
         case PIPE_FORMAT_R8_SNORM:
             return modifier | R300_C2_SEL_R;
 
-        /* ARGB 32-bit outputs. */
+        /* BGRA outputs. */
         case PIPE_FORMAT_B5G6R5_UNORM:
         case PIPE_FORMAT_B5G5R5A1_UNORM:
+        case PIPE_FORMAT_B5G5R5X1_UNORM:
         case PIPE_FORMAT_B4G4R4A4_UNORM:
         case PIPE_FORMAT_B8G8R8A8_UNORM:
-        case PIPE_FORMAT_B8G8R8A8_SRGB:
         case PIPE_FORMAT_B8G8R8X8_UNORM:
-        case PIPE_FORMAT_B8G8R8X8_SRGB:
             return modifier |
                 R300_C0_SEL_B | R300_C1_SEL_G |
                 R300_C2_SEL_R | R300_C3_SEL_A;
 
-        /* BGRA 32-bit outputs. */
+        /* ARGB outputs. */
         case PIPE_FORMAT_A8R8G8B8_UNORM:
-        case PIPE_FORMAT_A8R8G8B8_SRGB:
         case PIPE_FORMAT_X8R8G8B8_UNORM:
-        case PIPE_FORMAT_X8R8G8B8_SRGB:
             return modifier |
                 R300_C0_SEL_A | R300_C1_SEL_R |
                 R300_C2_SEL_G | R300_C3_SEL_B;
 
-        /* RGBA 32-bit outputs. */
+        /* ABGR outputs. */
         case PIPE_FORMAT_A8B8G8R8_UNORM:
-        case PIPE_FORMAT_R8G8B8A8_SNORM:
-        case PIPE_FORMAT_A8B8G8R8_SRGB:
         case PIPE_FORMAT_X8B8G8R8_UNORM:
-        case PIPE_FORMAT_X8B8G8R8_SRGB:
             return modifier |
                 R300_C0_SEL_A | R300_C1_SEL_B |
                 R300_C2_SEL_G | R300_C3_SEL_R;
 
-        /* ABGR 32-bit outputs. */
+        /* RGBA outputs. */
+        case PIPE_FORMAT_R8G8B8A8_SNORM:
         case PIPE_FORMAT_R8SG8SB8UX8U_NORM:
         case PIPE_FORMAT_R10G10B10A2_UNORM:
-        /* RGBA high precision outputs (same swizzles as ABGR low precision) */
         case PIPE_FORMAT_R16G16B16A16_UNORM:
         case PIPE_FORMAT_R16G16B16A16_SNORM:
         //case PIPE_FORMAT_R16G16B16A16_FLOAT: /* not in pipe_format */
@@ -658,7 +651,7 @@ unsigned r300_texture_get_stride(struct r300_screen* screen,
 
     width = u_minify(tex->tex.width0, level);
 
-    if (!util_format_is_compressed(tex->tex.format)) {
+    if (r300_format_is_plain(tex->tex.format)) {
         tile_width = r300_texture_get_tile_size(tex, TILE_WIDTH,
                                                 tex->mip_macrotile[level]);
         width = align(width, tile_width);
@@ -676,7 +669,7 @@ static unsigned r300_texture_get_nblocksy(struct r300_texture* tex,
 
     height = u_minify(tex->tex.height0, level);
 
-    if (!util_format_is_compressed(tex->tex.format)) {
+    if (r300_format_is_plain(tex->tex.format)) {
         tile_height = r300_texture_get_tile_size(tex, TILE_HEIGHT,
                                                  tex->mip_macrotile[level]);
         height = align(height, tile_height);
@@ -699,7 +692,8 @@ static void r300_setup_miptree(struct r300_screen* screen,
         /* Let's see if this miplevel can be macrotiled. */
         tex->mip_macrotile[i] =
             (tex->macrotile == R300_BUFFER_TILED &&
-             r300_texture_macro_switch(tex, i, rv350_mode, TILE_WIDTH)) ?
+             r300_texture_macro_switch(tex, i, rv350_mode, TILE_WIDTH) &&
+             r300_texture_macro_switch(tex, i, rv350_mode, TILE_HEIGHT)) ?
              R300_BUFFER_TILED : R300_BUFFER_LINEAR;
 
         stride = r300_texture_get_stride(screen, tex, i);
@@ -733,10 +727,11 @@ static void r300_setup_flags(struct r300_texture* tex)
 static void r300_setup_tiling(struct pipe_screen *screen,
                               struct r300_texture *tex)
 {
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)screen->winsys;
     enum pipe_format format = tex->tex.format;
     boolean rv350_mode = r300_screen(screen)->caps->family >= CHIP_FAMILY_RV350;
 
-    if (util_format_is_compressed(format)) {
+    if (!r300_format_is_plain(format)) {
         return;
     }
 
@@ -752,12 +747,12 @@ static void r300_setup_tiling(struct pipe_screen *screen,
             tex->microtile = R300_BUFFER_TILED;
             break;
 
-        /* XXX Square-tiling doesn't work with kernel older than 2.6.34,
-         * XXX need to check the DRM version */
-        /*case 2:
+        case 2:
         case 8:
-            tex->microtile = R300_BUFFER_SQUARETILED;
-            break;*/
+            if (rws->get_value(rws, R300_VID_SQUARE_TILING_SUPPORT)) {
+                tex->microtile = R300_BUFFER_SQUARETILED;
+            }
+            break;
     }
 
     /* Set macrotiling. */
@@ -795,8 +790,8 @@ static struct pipe_texture* r300_texture_create(struct pipe_screen* screen,
 				     tex->size);
     rws->buffer_set_tiling(rws, tex->buffer,
 			   tex->pitch[0],
-			   tex->microtile != R300_BUFFER_LINEAR,
-			   tex->macrotile != R300_BUFFER_LINEAR);
+			   tex->microtile,
+			   tex->macrotile);
 
     if (!tex->buffer) {
         FREE(tex);
diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
index aff4ddd4e23..3b3802ee2b2 100644
--- a/src/gallium/drivers/r300/r300_tgsi_to_rc.c
+++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
@@ -25,12 +25,11 @@
 #include "radeon_compiler.h"
 #include "radeon_program.h"
 
+#include "tgsi/tgsi_info.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_scan.h"
 #include "tgsi/tgsi_util.h"
 
-#include "util/u_debug.h"
-
 static unsigned translate_opcode(unsigned opcode)
 {
     switch(opcode) {
@@ -145,7 +144,7 @@ static unsigned translate_opcode(unsigned opcode)
         case TGSI_OPCODE_KIL: return RC_OPCODE_KIL;
     }
 
-    debug_printf("r300: Unknown TGSI/RC opcode: %i\n", opcode);
+    fprintf(stderr, "r300: Unknown TGSI/RC opcode: %s\n", tgsi_get_opcode_name(opcode));
     return RC_OPCODE_ILLEGAL_OPCODE;
 }
 
@@ -272,9 +271,6 @@ static void transform_instruction(struct tgsi_to_rc * ttr, struct tgsi_full_inst
     struct rc_instruction * dst;
     int i;
 
-    if (src->Instruction.Opcode == TGSI_OPCODE_END)
-        return;
-
     dst = rc_insert_new_instruction(ttr->compiler, ttr->compiler->Program.Instructions.Prev);
     dst->U.I.Opcode = translate_opcode(src->Instruction.Opcode);
     dst->U.I.SaturateMode = translate_saturate(src->Instruction.Saturate);
@@ -333,6 +329,7 @@ static void handle_immediate(struct tgsi_to_rc * ttr,
 void r300_tgsi_to_rc(struct tgsi_to_rc * ttr,
                      const struct tgsi_token * tokens)
 {
+    struct tgsi_full_instruction *inst;
     struct tgsi_parse_context parser;
     unsigned imm_index = 0;
     int i;
@@ -367,7 +364,15 @@ void r300_tgsi_to_rc(struct tgsi_to_rc * ttr,
                 imm_index++;
                 break;
             case TGSI_TOKEN_TYPE_INSTRUCTION:
-                transform_instruction(ttr, &parser.FullToken.FullInstruction);
+                inst = &parser.FullToken.FullInstruction;
+                /* This hack with the RET opcode woudn't work with
+                 * conditionals. */
+                if (inst->Instruction.Opcode == TGSI_OPCODE_END ||
+                    inst->Instruction.Opcode == TGSI_OPCODE_RET) {
+                    break;
+                }
+
+                transform_instruction(ttr, inst);
                 break;
         }
     }
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index bd6b95dccba..d5690caa68e 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -299,7 +299,7 @@ void r300_translate_vertex_shader(struct r300_context* r300,
     r3xx_compile_vertex_program(&compiler);
     if (compiler.Base.Error) {
         /* XXX We should fallback using Draw. */
-        fprintf(stderr, "r300 VP: Compiler error\n");
+        fprintf(stderr, "r300 VP: Compiler error:\n%s", compiler.Base.ErrorMsg);
         abort();
     }
 
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index e5183a8239c..acfa5dbeb9c 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -30,6 +30,8 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
 
+#include "r300_defines.h"
+
 struct r300_winsys_screen;
 
 /* Creates a new r300 screen. */
@@ -47,10 +49,9 @@ enum r300_value_id {
     R300_VID_PCI_ID,
     R300_VID_GB_PIPES,
     R300_VID_Z_PIPES,
+    R300_VID_SQUARE_TILING_SUPPORT
 };
 
-#define R300_USAGE_FLAG_DONT_SYNC (1 << 17)
-
 struct r300_winsys_screen {
     void (*destroy)(struct r300_winsys_screen *ws);
     
@@ -148,8 +149,8 @@ struct r300_winsys_screen {
     void (*buffer_set_tiling)(struct r300_winsys_screen *winsys,
                               struct r300_winsys_buffer *buffer,
                               uint32_t pitch,
-                              boolean microtiled,
-                              boolean macrotiled);
+                              enum r300_buffer_tiling microtiled,
+                              enum r300_buffer_tiling macrotiled);
 
     uint32_t (*get_value)(struct r300_winsys_screen *winsys,
 			  enum r300_value_id vid);
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index de92a0cd2c7..937a573092f 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -104,12 +104,12 @@ softpipe_destroy( struct pipe_context *pipe )
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
       sp_destroy_tex_tile_cache(softpipe->tex_cache[i]);
-      pipe_texture_reference(&softpipe->texture[i], NULL);
+      pipe_sampler_view_reference(&softpipe->sampler_views[i], NULL);
    }
 
    for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
       sp_destroy_tex_tile_cache(softpipe->vertex_tex_cache[i]);
-      pipe_texture_reference(&softpipe->vertex_textures[i], NULL);
+      pipe_sampler_view_reference(&softpipe->vertex_sampler_views[i], NULL);
    }
 
    for (i = 0; i < PIPE_SHADER_TYPES; i++) {
@@ -257,8 +257,10 @@ softpipe_create_context( struct pipe_screen *screen,
    softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state;
    softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple;
    softpipe->pipe.set_scissor_state = softpipe_set_scissor_state;
-   softpipe->pipe.set_fragment_sampler_textures = softpipe_set_sampler_textures;
-   softpipe->pipe.set_vertex_sampler_textures = softpipe_set_vertex_sampler_textures;
+   softpipe->pipe.set_fragment_sampler_views = softpipe_set_sampler_views;
+   softpipe->pipe.set_vertex_sampler_views = softpipe_set_vertex_sampler_views;
+   softpipe->pipe.create_sampler_view = softpipe_create_sampler_view;
+   softpipe->pipe.sampler_view_destroy = softpipe_sampler_view_destroy;
    softpipe->pipe.set_viewport_state = softpipe_set_viewport_state;
 
    softpipe->pipe.set_vertex_buffers = softpipe_set_vertex_buffers;
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index 9a8158e6a22..75e03c8ae6b 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -70,15 +70,15 @@ struct softpipe_context {
    struct pipe_framebuffer_state framebuffer;
    struct pipe_poly_stipple poly_stipple;
    struct pipe_scissor_state scissor;
-   struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
-   struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS];
+   struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_view *vertex_sampler_views[PIPE_MAX_VERTEX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
 
    unsigned num_samplers;
-   unsigned num_textures;
+   unsigned num_sampler_views;
    unsigned num_vertex_samplers;
-   unsigned num_vertex_textures;
+   unsigned num_vertex_sampler_views;
    unsigned num_vertex_buffers;
 
    unsigned dirty; /**< Mask of SP_NEW_x flags */
diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c
index 3d76af4d8cb..508fe8f764d 100644
--- a/src/gallium/drivers/softpipe/sp_flush.c
+++ b/src/gallium/drivers/softpipe/sp_flush.c
@@ -50,10 +50,10 @@ softpipe_flush( struct pipe_context *pipe,
    draw_flush(softpipe->draw);
 
    if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
-      for (i = 0; i < softpipe->num_textures; i++) {
+      for (i = 0; i < softpipe->num_sampler_views; i++) {
          sp_flush_tex_tile_cache(softpipe->tex_cache[i]);
       }
-      for (i = 0; i < softpipe->num_vertex_textures; i++) {
+      for (i = 0; i < softpipe->num_vertex_sampler_views; i++) {
          sp_flush_tex_tile_cache(softpipe->vertex_tex_cache[i]);
       }
    }
diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
index 4815a0d49f1..17cd5b82072 100644
--- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c
+++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
@@ -1,6 +1,7 @@
 /**************************************************************************
  * 
  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2010 VMware, Inc.
  * All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -18,7 +19,7 @@
  * 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
+ * IN NO EVENT SHALL THE AUTHORS 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.
@@ -26,7 +27,7 @@
  **************************************************************************/
 
 /**
- * \brief  Quad depth testing
+ * \brief  Quad depth / stencil testing
  */
 
 #include "pipe/p_defines.h"
@@ -96,7 +97,9 @@ get_depth_stencil_values( struct depth_data *data,
    }
 }
 
-/* If the shader has not been run, interpolate the depth values
+
+/**
+ * If the shader has not been run, interpolate the depth values
  * ourselves.
  */
 static void
@@ -115,6 +118,9 @@ interpolate_quad_depth( struct quad_header *quad )
 }
 
 
+/**
+ * Compute the depth_data::qzzzz[] values from the float fragment Z values.
+ */
 static void
 convert_quad_depth( struct depth_data *data, 
                     const struct quad_header *quad )
@@ -173,6 +179,9 @@ convert_quad_depth( struct depth_data *data,
 
 
 
+/**
+ * Write data->bzzzz[] values and data->stencilVals into the Z/stencil buffer.
+ */
 static void
 write_depth_stencil_values( struct depth_data *data,
                             struct quad_header *quad )
@@ -225,7 +234,6 @@ write_depth_stencil_values( struct depth_data *data,
 
 
 
-
 /** Only 8-bit stencil supported */
 #define STENCIL_MAX 0xff
 
@@ -408,12 +416,11 @@ apply_stencil_op(struct depth_data *data,
 
    
 
-/*
+/**
  * To increase efficiency, we should probably have multiple versions
  * of this function that are specifically for Z16, Z32 and FP Z buffers.
  * Try to effectively do that with codegen...
  */
-
 static boolean
 depth_test_quad(struct quad_stage *qs, 
                 struct depth_data *data,
@@ -523,7 +530,6 @@ depth_stencil_test_quad(struct quad_stage *qs,
    wrtMask = softpipe->depth_stencil->stencil[face].writemask;
    valMask = softpipe->depth_stencil->stencil[face].valuemask;
 
-
    /* do the stencil test first */
    {
       unsigned passMask, failMask;
@@ -563,7 +569,7 @@ depth_stencil_test_quad(struct quad_stage *qs,
 
 
 #define ALPHATEST( FUNC, COMP )                                         \
-   static int                                                          \
+   static int                                                           \
    alpha_test_quads_##FUNC( struct quad_stage *qs,                      \
                            struct quad_header *quads[],                 \
                            unsigned nr )                                \
@@ -629,6 +635,7 @@ alpha_test_quads(struct quad_stage *qs,
    }
 }
 
+
 static unsigned mask_count[16] = 
 {
    0,                           /* 0x0 */
@@ -665,6 +672,9 @@ get_depth_bits(struct quad_stage *qs)
 
 
 
+/**
+ * General depth/stencil test function.  Used when there's no fast-path.
+ */
 static void
 depth_test_quads_fallback(struct quad_stage *qs, 
                           struct quad_header *quads[],
@@ -712,7 +722,6 @@ depth_test_quads_fallback(struct quad_stage *qs,
                write_depth_stencil_values(&data, quads[i]);
          }
 
-
          quads[pass++] = quads[i];
       }
 
@@ -730,169 +739,36 @@ depth_test_quads_fallback(struct quad_stage *qs,
 
 
 /**
- * Special-case Z testing for 16-bit Zbuffer, PIPE_FUNC_LESS and
- * Z buffer writes enabled.
- *
- * NOTE: there's no guarantee that the quads are sequentially side by
- * side.  The fragment shader may have culled some quads, etc.  Sliver
- * triangles may generate non-sequential quads.
- */
-static void
-depth_interp_z16_less_write(struct quad_stage *qs, 
-                            struct quad_header *quads[],
-                            unsigned nr)
-{
-   unsigned i, pass = 0;
-   const unsigned ix = quads[0]->input.x0;
-   const unsigned iy = quads[0]->input.y0;
-   const float fx = (float) ix;
-   const float fy = (float) iy;
-   const float dzdx = quads[0]->posCoef->dadx[2];
-   const float dzdy = quads[0]->posCoef->dady[2];
-   const float z0 = quads[0]->posCoef->a0[2] + dzdx * fx + dzdy * fy;
-   struct softpipe_cached_tile *tile;
-   ushort (*depth16)[TILE_SIZE];
-   ushort init_idepth[4], idepth[4], depth_step;
-   const float scale = 65535.0;
-
-   /* compute scaled depth of the four pixels in first quad */
-   init_idepth[0] = (ushort)((z0) * scale);
-   init_idepth[1] = (ushort)((z0 + dzdx) * scale);
-   init_idepth[2] = (ushort)((z0 + dzdy) * scale);
-   init_idepth[3] = (ushort)((z0 + dzdx + dzdy) * scale);
-
-   depth_step = (ushort)(dzdx * scale);
-
-   tile = sp_get_cached_tile(qs->softpipe->zsbuf_cache, ix, iy);
-
-   for (i = 0; i < nr; i++) {
-      const unsigned outmask = quads[i]->inout.mask;
-      const int dx = quads[i]->input.x0 - ix;
-      unsigned mask = 0;
-
-      /* compute depth for this quad */
-      idepth[0] = init_idepth[0] + dx * depth_step;
-      idepth[1] = init_idepth[1] + dx * depth_step;
-      idepth[2] = init_idepth[2] + dx * depth_step;
-      idepth[3] = init_idepth[3] + dx * depth_step;
-
-      depth16 = (ushort (*)[TILE_SIZE])
-         &tile->data.depth16[iy % TILE_SIZE][(ix + dx)% TILE_SIZE];
-
-      if ((outmask & 1) && idepth[0] < depth16[0][0]) {
-         depth16[0][0] = idepth[0];
-         mask |= (1 << 0);
-      }
-
-      if ((outmask & 2) && idepth[1] < depth16[0][1]) {
-         depth16[0][1] = idepth[1];
-         mask |= (1 << 1);
-      }
-
-      if ((outmask & 4) && idepth[2] < depth16[1][0]) {
-         depth16[1][0] = idepth[2];
-         mask |= (1 << 2);
-      }
-
-      if ((outmask & 8) && idepth[3] < depth16[1][1]) {
-         depth16[1][1] = idepth[3];
-         mask |= (1 << 3);
-      }
-
-      quads[i]->inout.mask = mask;
-      if (quads[i]->inout.mask)
-         quads[pass++] = quads[i];
-   }
-
-   if (pass)
-      qs->next->run(qs->next, quads, pass);
-
-}
-
-
-/**
- * Special-case Z testing for 16-bit Zbuffer, PIPE_FUNC_LEQUAL and
- * Z buffer writes enabled.
- *
- * NOTE: there's no guarantee that the quads are sequentially side by
- * side.  The fragment shader may have culled some quads, etc.  Sliver
- * triangles may generate non-sequential quads.
+ * Special-case Z testing for 16-bit Zbuffer and Z buffer writes enabled.
  */
-static void
-depth_interp_z16_lequal_write(struct quad_stage *qs, 
-                            struct quad_header *quads[],
-                            unsigned nr)
-{
-   unsigned i, pass = 0;
-   const unsigned ix = quads[0]->input.x0;
-   const unsigned iy = quads[0]->input.y0;
-   const float fx = (float) ix;
-   const float fy = (float) iy;
-   const float dzdx = quads[0]->posCoef->dadx[2];
-   const float dzdy = quads[0]->posCoef->dady[2];
-   const float z0 = quads[0]->posCoef->a0[2] + dzdx * fx + dzdy * fy;
-   struct softpipe_cached_tile *tile;
-   ushort (*depth16)[TILE_SIZE];
-   ushort init_idepth[4], idepth[4], depth_step;
-   const float scale = 65535.0;
-
-   /* compute scaled depth of the four pixels in first quad */
-   init_idepth[0] = (ushort)((z0) * scale);
-   init_idepth[1] = (ushort)((z0 + dzdx) * scale);
-   init_idepth[2] = (ushort)((z0 + dzdy) * scale);
-   init_idepth[3] = (ushort)((z0 + dzdx + dzdy) * scale);
-
-   depth_step = (ushort)(dzdx * scale);
 
-   tile = sp_get_cached_tile(qs->softpipe->zsbuf_cache, ix, iy);
-
-   for (i = 0; i < nr; i++) {
-      const unsigned outmask = quads[i]->inout.mask;
-      const int dx = quads[i]->input.x0 - ix;
-      unsigned mask = 0;
-      
-      /* compute depth for this quad */
-      idepth[0] = init_idepth[0] + dx * depth_step;
-      idepth[1] = init_idepth[1] + dx * depth_step;
-      idepth[2] = init_idepth[2] + dx * depth_step;
-      idepth[3] = init_idepth[3] + dx * depth_step;
-
-      depth16 = (ushort (*)[TILE_SIZE])
-         &tile->data.depth16[iy % TILE_SIZE][(ix + dx)% TILE_SIZE];
-
-      if ((outmask & 1) && idepth[0] <= depth16[0][0]) {
-         depth16[0][0] = idepth[0];
-         mask |= (1 << 0);
-      }
-
-      if ((outmask & 2) && idepth[1] <= depth16[0][1]) {
-         depth16[0][1] = idepth[1];
-         mask |= (1 << 1);
-      }
-
-      if ((outmask & 4) && idepth[2] <= depth16[1][0]) {
-         depth16[1][0] = idepth[2];
-         mask |= (1 << 2);
-      }
-
-      if ((outmask & 8) && idepth[3] <= depth16[1][1]) {
-         depth16[1][1] = idepth[3];
-         mask |= (1 << 3);
-      }
+#define NAME depth_interp_z16_less_write
+#define OPERATOR <
+#include "sp_quad_depth_test_tmp.h"
 
-      depth16 = (ushort (*)[TILE_SIZE]) &depth16[0][2];
+#define NAME depth_interp_z16_equal_write
+#define OPERATOR ==
+#include "sp_quad_depth_test_tmp.h"
 
-      quads[i]->inout.mask = mask;
-      if (quads[i]->inout.mask)
-         quads[pass++] = quads[i];
-   }
+#define NAME depth_interp_z16_lequal_write
+#define OPERATOR <=
+#include "sp_quad_depth_test_tmp.h"
 
-   if (pass)
-      qs->next->run(qs->next, quads, pass);
+#define NAME depth_interp_z16_greater_write
+#define OPERATOR >
+#include "sp_quad_depth_test_tmp.h"
 
-}
+#define NAME depth_interp_z16_notequal_write
+#define OPERATOR !=
+#include "sp_quad_depth_test_tmp.h"
 
+#define NAME depth_interp_z16_gequal_write
+#define OPERATOR >=
+#include "sp_quad_depth_test_tmp.h"
 
+#define NAME depth_interp_z16_always_write
+#define ALWAYS 1
+#include "sp_quad_depth_test_tmp.h"
 
 
 
@@ -926,6 +802,10 @@ choose_depth_test(struct quad_stage *qs,
 
    boolean occlusion = qs->softpipe->active_query_count;
 
+   /* default */
+   qs->run = depth_test_quads_fallback;
+
+   /* look for special cases */
    if (!alpha &&
        !depth &&
        !stencil) {
@@ -938,57 +818,62 @@ choose_depth_test(struct quad_stage *qs,
             !occlusion &&
             !stencil) 
    {
-      switch (depthfunc) {
-      case PIPE_FUNC_LESS:
-         switch (qs->softpipe->framebuffer.zsbuf->format) {
-         case PIPE_FORMAT_Z16_UNORM:
+      if (qs->softpipe->framebuffer.zsbuf->format == PIPE_FORMAT_Z16_UNORM) {
+         switch (depthfunc) {
+         case PIPE_FUNC_NEVER:
+            qs->run = depth_test_quads_fallback;
+            break;
+         case PIPE_FUNC_LESS:
             qs->run = depth_interp_z16_less_write;
             break;
-         default:
-            qs->run = depth_test_quads_fallback;
+         case PIPE_FUNC_EQUAL:
+            qs->run = depth_interp_z16_equal_write;
             break;
-         }
-         break;
-      case PIPE_FUNC_LEQUAL:
-         switch (qs->softpipe->framebuffer.zsbuf->format) {
-         case PIPE_FORMAT_Z16_UNORM:
+         case PIPE_FUNC_LEQUAL:
             qs->run = depth_interp_z16_lequal_write;
             break;
+         case PIPE_FUNC_GREATER:
+            qs->run = depth_interp_z16_greater_write;
+            break;
+         case PIPE_FUNC_NOTEQUAL:
+            qs->run = depth_interp_z16_notequal_write;
+            break;
+         case PIPE_FUNC_GEQUAL:
+            qs->run = depth_interp_z16_gequal_write;
+            break;
+         case PIPE_FUNC_ALWAYS:
+            qs->run = depth_interp_z16_always_write;
+            break;
          default:
             qs->run = depth_test_quads_fallback;
             break;
          }
-         break;
-      default:
-         qs->run = depth_test_quads_fallback;
       }
    }
-   else {
-      qs->run = depth_test_quads_fallback;
-   }
-
 
+   /* next quad/fragment stage */
    qs->run( qs, quads, nr );
 }
 
 
 
-
-
-static void depth_test_begin(struct quad_stage *qs)
+static void
+depth_test_begin(struct quad_stage *qs)
 {
    qs->run = choose_depth_test;
    qs->next->begin(qs->next);
 }
 
 
-static void depth_test_destroy(struct quad_stage *qs)
+static void
+depth_test_destroy(struct quad_stage *qs)
 {
    FREE( qs );
 }
 
 
-struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe )
+struct quad_stage *
+sp_quad_depth_test_stage(struct softpipe_context *softpipe)
 {
    struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test_tmp.h b/src/gallium/drivers/softpipe/sp_quad_depth_test_tmp.h
new file mode 100644
index 00000000000..25af415c256
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_depth_test_tmp.h
@@ -0,0 +1,147 @@
+/**************************************************************************
+ * 
+ * Copyright 2010 VMware, Inc.  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 THE AUTHORS 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.
+ * 
+ **************************************************************************/
+
+
+/*
+ * Template for generating Z test functions
+ * Only PIPE_FORMAT_Z16_UNORM supported at this time.
+ */
+
+
+#ifndef NAME
+#error "NAME is not defined!"
+#endif
+
+#if !defined(OPERATOR) && !defined(ALWAYS)
+#error "neither OPERATOR nor ALWAYS is defined!"
+#endif
+
+
+/*
+ * NOTE: there's no guarantee that the quads are sequentially side by
+ * side.  The fragment shader may have culled some quads, etc.  Sliver
+ * triangles may generate non-sequential quads.
+ */
+static void
+NAME(struct quad_stage *qs, 
+     struct quad_header *quads[],
+     unsigned nr)
+{
+   unsigned i, pass = 0;
+   const unsigned ix = quads[0]->input.x0;
+   const unsigned iy = quads[0]->input.y0;
+   const float fx = (float) ix;
+   const float fy = (float) iy;
+   const float dzdx = quads[0]->posCoef->dadx[2];
+   const float dzdy = quads[0]->posCoef->dady[2];
+   const float z0 = quads[0]->posCoef->a0[2] + dzdx * fx + dzdy * fy;
+   struct softpipe_cached_tile *tile;
+   ushort (*depth16)[TILE_SIZE];
+   ushort init_idepth[4], idepth[4], depth_step;
+   const float scale = 65535.0;
+
+   /* compute scaled depth of the four pixels in first quad */
+   init_idepth[0] = (ushort)((z0) * scale);
+   init_idepth[1] = (ushort)((z0 + dzdx) * scale);
+   init_idepth[2] = (ushort)((z0 + dzdy) * scale);
+   init_idepth[3] = (ushort)((z0 + dzdx + dzdy) * scale);
+
+   depth_step = (ushort)(dzdx * scale);
+
+   tile = sp_get_cached_tile(qs->softpipe->zsbuf_cache, ix, iy);
+
+   for (i = 0; i < nr; i++) {
+      const unsigned outmask = quads[i]->inout.mask;
+      const int dx = quads[i]->input.x0 - ix;
+      unsigned mask = 0;
+      
+      /* compute depth for this quad */
+      idepth[0] = init_idepth[0] + dx * depth_step;
+      idepth[1] = init_idepth[1] + dx * depth_step;
+      idepth[2] = init_idepth[2] + dx * depth_step;
+      idepth[3] = init_idepth[3] + dx * depth_step;
+
+      depth16 = (ushort (*)[TILE_SIZE])
+         &tile->data.depth16[iy % TILE_SIZE][(ix + dx)% TILE_SIZE];
+
+#ifdef ALWAYS
+      if (outmask & 1) {
+         depth16[0][0] = idepth[0];
+         mask |= (1 << 0);
+      }
+
+      if (outmask & 2) {
+         depth16[0][1] = idepth[1];
+         mask |= (1 << 1);
+      }
+
+      if (outmask & 4) {
+         depth16[1][0] = idepth[2];
+         mask |= (1 << 2);
+      }
+
+      if (outmask & 8) {
+         depth16[1][1] = idepth[3];
+         mask |= (1 << 3);
+      }
+#else
+      /* Note: OPERATOR appears here: */
+      if ((outmask & 1) && (idepth[0] OPERATOR depth16[0][0])) {
+         depth16[0][0] = idepth[0];
+         mask |= (1 << 0);
+      }
+
+      if ((outmask & 2) && (idepth[1] OPERATOR depth16[0][1])) {
+         depth16[0][1] = idepth[1];
+         mask |= (1 << 1);
+      }
+
+      if ((outmask & 4) && (idepth[2] OPERATOR depth16[1][0])) {
+         depth16[1][0] = idepth[2];
+         mask |= (1 << 2);
+      }
+
+      if ((outmask & 8) && (idepth[3] OPERATOR depth16[1][1])) {
+         depth16[1][1] = idepth[3];
+         mask |= (1 << 3);
+      }
+#endif
+
+      depth16 = (ushort (*)[TILE_SIZE]) &depth16[0][2];
+
+      quads[i]->inout.mask = mask;
+      if (quads[i]->inout.mask)
+         quads[pass++] = quads[i];
+   }
+
+   if (pass)
+      qs->next->run(qs->next, quads, pass);
+}
+
+
+#undef NAME
+#undef OPERATOR
+#undef ALWAYS
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index d62bfa3d633..757dc861284 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -173,8 +173,10 @@ softpipe_is_format_supported( struct pipe_screen *screen,
       break;
    }
 
-   if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
-      if(!winsys->is_displaytarget_format_supported(winsys, format))
+   if(tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                   PIPE_TEXTURE_USAGE_SCANOUT |
+                   PIPE_TEXTURE_USAGE_SHARED)) {
+      if(!winsys->is_displaytarget_format_supported(winsys, tex_usage, format))
          return FALSE;
    }
 
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index 6b01c0f4d72..ade96b0fd4c 100644
--- a/src/gallium/drivers/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -177,14 +177,23 @@ void softpipe_set_polygon_stipple( struct pipe_context *,
 void softpipe_set_scissor_state( struct pipe_context *,
                                  const struct pipe_scissor_state * );
 
-void softpipe_set_sampler_textures( struct pipe_context *,
-                                    unsigned num,
-                                    struct pipe_texture ** );
+void softpipe_set_sampler_views( struct pipe_context *,
+                                 unsigned num,
+                                 struct pipe_sampler_view ** );
 
 void
-softpipe_set_vertex_sampler_textures(struct pipe_context *,
-                                     unsigned num_textures,
-                                     struct pipe_texture **);
+softpipe_set_vertex_sampler_views(struct pipe_context *,
+                                  unsigned num,
+                                  struct pipe_sampler_view **);
+
+struct pipe_sampler_view *
+softpipe_create_sampler_view(struct pipe_context *pipe,
+                             struct pipe_texture *texture,
+                             const struct pipe_sampler_view *templ);
+
+void
+softpipe_sampler_view_destroy(struct pipe_context *pipe,
+                              struct pipe_sampler_view *view);
 
 void softpipe_set_viewport_state( struct pipe_context *,
                                   const struct pipe_viewport_state * );
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index 2b089c28316..d6f3229bed7 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -181,9 +181,8 @@ softpipe_set_constant_buffer(struct pipe_context *pipe,
    /* note: reference counting */
    pipe_buffer_reference(&softpipe->constants[shader][index], constants);
 
-   if(shader == PIPE_SHADER_VERTEX) {
-      draw_set_mapped_constant_buffer(softpipe->draw, PIPE_SHADER_VERTEX, index,
-                                      data, size);
+   if (shader == PIPE_SHADER_VERTEX || shader == PIPE_SHADER_GEOMETRY) {
+      draw_set_mapped_constant_buffer(softpipe->draw, shader, index, data, size);
    }
 
    softpipe->mapped_constants[shader][index] = data;
diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c
index ceb4e338f1a..d501952bba9 100644
--- a/src/gallium/drivers/softpipe/sp_state_sampler.c
+++ b/src/gallium/drivers/softpipe/sp_state_sampler.c
@@ -121,9 +121,38 @@ softpipe_bind_vertex_sampler_states(struct pipe_context *pipe,
 }
 
 
+struct pipe_sampler_view *
+softpipe_create_sampler_view(struct pipe_context *pipe,
+                             struct pipe_texture *texture,
+                             const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+
 void
-softpipe_set_sampler_textures(struct pipe_context *pipe,
-                              unsigned num, struct pipe_texture **texture)
+softpipe_sampler_view_destroy(struct pipe_context *pipe,
+                              struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
+}
+
+
+void
+softpipe_set_sampler_views(struct pipe_context *pipe,
+                           unsigned num,
+                           struct pipe_sampler_view **views)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
    uint i;
@@ -131,51 +160,51 @@ softpipe_set_sampler_textures(struct pipe_context *pipe,
    assert(num <= PIPE_MAX_SAMPLERS);
 
    /* Check for no-op */
-   if (num == softpipe->num_textures &&
-       !memcmp(softpipe->texture, texture, num * sizeof(struct pipe_texture *)))
+   if (num == softpipe->num_sampler_views &&
+       !memcmp(softpipe->sampler_views, views, num * sizeof(struct pipe_sampler_view *)))
       return;
 
    draw_flush(softpipe->draw);
 
    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      struct pipe_texture *tex = i < num ? texture[i] : NULL;
+      struct pipe_sampler_view *view = i < num ? views[i] : NULL;
 
-      pipe_texture_reference(&softpipe->texture[i], tex);
-      sp_tex_tile_cache_set_texture(softpipe->tex_cache[i], tex);
+      pipe_sampler_view_reference(&softpipe->sampler_views[i], view);
+      sp_tex_tile_cache_set_sampler_view(softpipe->tex_cache[i], view);
    }
 
-   softpipe->num_textures = num;
+   softpipe->num_sampler_views = num;
 
    softpipe->dirty |= SP_NEW_TEXTURE;
 }
 
 
 void
-softpipe_set_vertex_sampler_textures(struct pipe_context *pipe,
-                                     unsigned num_textures,
-                                     struct pipe_texture **textures)
+softpipe_set_vertex_sampler_views(struct pipe_context *pipe,
+                                  unsigned num,
+                                  struct pipe_sampler_view **views)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
    uint i;
 
-   assert(num_textures <= PIPE_MAX_VERTEX_SAMPLERS);
+   assert(num <= PIPE_MAX_VERTEX_SAMPLERS);
 
    /* Check for no-op */
-   if (num_textures == softpipe->num_vertex_textures &&
-       !memcmp(softpipe->vertex_textures, textures, num_textures * sizeof(struct pipe_texture *))) {
+   if (num == softpipe->num_vertex_sampler_views &&
+       !memcmp(softpipe->vertex_sampler_views, views, num * sizeof(struct pipe_sampler_view *))) {
       return;
    }
 
    draw_flush(softpipe->draw);
 
    for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
-      struct pipe_texture *tex = i < num_textures ? textures[i] : NULL;
+      struct pipe_sampler_view *view = i < num ? views[i] : NULL;
 
-      pipe_texture_reference(&softpipe->vertex_textures[i], tex);
-      sp_tex_tile_cache_set_texture(softpipe->vertex_tex_cache[i], tex);
+      pipe_sampler_view_reference(&softpipe->vertex_sampler_views[i], view);
+      sp_tex_tile_cache_set_sampler_view(softpipe->vertex_tex_cache[i], view);
    }
 
-   softpipe->num_vertex_textures = num_textures;
+   softpipe->num_vertex_sampler_views = num;
 
    softpipe->dirty |= SP_NEW_TEXTURE;
 }
@@ -245,29 +274,41 @@ softpipe_reset_sampler_varients(struct softpipe_context *softpipe)
     */
    for (i = 0; i <= softpipe->vs->max_sampler; i++) {
       if (softpipe->vertex_samplers[i]) {
+         struct pipe_texture *texture = NULL;
+
+         if (softpipe->vertex_sampler_views[i]) {
+            texture = softpipe->vertex_sampler_views[i]->texture;
+         }
+
          softpipe->tgsi.vert_samplers_list[i] = 
             get_sampler_varient( i,
-                                sp_sampler(softpipe->vertex_samplers[i]),
-                                softpipe->vertex_textures[i],
+                                 sp_sampler(softpipe->vertex_samplers[i]),
+                                 texture,
                                  TGSI_PROCESSOR_VERTEX );
 
          sp_sampler_varient_bind_texture( softpipe->tgsi.vert_samplers_list[i], 
-                                         softpipe->vertex_tex_cache[i],
-                                         softpipe->vertex_textures[i] );
+                                          softpipe->vertex_tex_cache[i],
+                                          texture );
       }
    }
 
    for (i = 0; i <= softpipe->fs->info.file_max[TGSI_FILE_SAMPLER]; i++) {
       if (softpipe->sampler[i]) {
+         struct pipe_texture *texture = NULL;
+
+         if (softpipe->sampler_views[i]) {
+            texture = softpipe->sampler_views[i]->texture;
+         }
+
          softpipe->tgsi.frag_samplers_list[i] =
             get_sampler_varient( i,
                                  sp_sampler(softpipe->sampler[i]),
-                                 softpipe->texture[i],
+                                 texture,
                                  TGSI_PROCESSOR_FRAGMENT );
 
          sp_sampler_varient_bind_texture( softpipe->tgsi.frag_samplers_list[i], 
                                           softpipe->tex_cache[i],
-                                          softpipe->texture[i] );
+                                          texture );
       }
    }
 }
diff --git a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
index e3a5e37ce44..6594514c38f 100644
--- a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
@@ -116,12 +116,13 @@ sp_tex_tile_cache_validate_texture(struct softpipe_tex_tile_cache *tc)
 }
 
 /**
- * Specify the texture to cache.
+ * Specify the sampler view to cache.
  */
 void
-sp_tex_tile_cache_set_texture(struct softpipe_tex_tile_cache *tc,
-                          struct pipe_texture *texture)
+sp_tex_tile_cache_set_sampler_view(struct softpipe_tex_tile_cache *tc,
+                                   struct pipe_sampler_view *view)
 {
+   struct pipe_texture *texture = view ? view->texture : NULL;
    uint i;
 
    assert(!tc->transfer);
@@ -139,6 +140,14 @@ sp_tex_tile_cache_set_texture(struct softpipe_tex_tile_cache *tc,
          tc->tex_trans = NULL;
       }
 
+      if (view) {
+         tc->swizzle_r = view->swizzle_r;
+         tc->swizzle_g = view->swizzle_g;
+         tc->swizzle_b = view->swizzle_b;
+         tc->swizzle_a = view->swizzle_a;
+         tc->format = view->format;
+      }
+
       /* mark as entries as invalid/empty */
       /* XXX we should try to avoid this when the teximage hasn't changed */
       for (i = 0; i < NUM_ENTRIES; i++) {
@@ -251,12 +260,18 @@ sp_find_cached_tile_tex(struct softpipe_tex_tile_cache *tc,
       }
 
       /* get tile from the transfer (view into texture) */
-      pipe_get_tile_rgba(tc->pipe,
-                         tc->tex_trans,
-                         addr.bits.x * TILE_SIZE, 
-                         addr.bits.y * TILE_SIZE,
-                         TILE_SIZE, TILE_SIZE,
-                         (float *) tile->data.color);
+      pipe_get_tile_swizzle(tc->pipe,
+			    tc->tex_trans,
+                            addr.bits.x * TILE_SIZE, 
+                            addr.bits.y * TILE_SIZE,
+                            TILE_SIZE,
+                            TILE_SIZE,
+                            tc->swizzle_r,
+                            tc->swizzle_g,
+                            tc->swizzle_b,
+                            tc->swizzle_a,
+                            tc->format,
+                            (float *) tile->data.color);
       tile->addr = addr;
    }
 
diff --git a/src/gallium/drivers/softpipe/sp_tex_tile_cache.h b/src/gallium/drivers/softpipe/sp_tex_tile_cache.h
index b1163972587..12ae7ba12d6 100644
--- a/src/gallium/drivers/softpipe/sp_tex_tile_cache.h
+++ b/src/gallium/drivers/softpipe/sp_tex_tile_cache.h
@@ -83,6 +83,12 @@ struct softpipe_tex_tile_cache
    void *tex_trans_map;
    int tex_face, tex_level, tex_z;
 
+   unsigned swizzle_r;
+   unsigned swizzle_g;
+   unsigned swizzle_b;
+   unsigned swizzle_a;
+   unsigned format;
+
    struct softpipe_tex_cached_tile *last_tile;  /**< most recently retrieved tile */
 };
 
@@ -101,8 +107,8 @@ extern void
 sp_tex_tile_cache_unmap_transfers(struct softpipe_tex_tile_cache *tc);
 
 extern void
-sp_tex_tile_cache_set_texture(struct softpipe_tex_tile_cache *tc,
-                          struct pipe_texture *texture);
+sp_tex_tile_cache_set_sampler_view(struct softpipe_tex_tile_cache *tc,
+                                   struct pipe_sampler_view *view);
 
 void
 sp_tex_tile_cache_validate_texture(struct softpipe_tex_tile_cache *tc);
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 2aff6118f41..f4983b7c8e8 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -91,6 +91,7 @@ softpipe_displaytarget_layout(struct pipe_screen *screen,
    /* Round up the surface size to a multiple of the tile size?
     */
    spt->dt = winsys->displaytarget_create(winsys,
+                                          spt->base.tex_usage,
                                           spt->base.format,
                                           spt->base.width0, 
                                           spt->base.height0,
@@ -139,8 +140,6 @@ softpipe_texture_create(struct pipe_screen *screen,
 }
 
 
-
-
 static void
 softpipe_texture_destroy(struct pipe_texture *pt)
 {
@@ -161,6 +160,55 @@ softpipe_texture_destroy(struct pipe_texture *pt)
 }
 
 
+static struct pipe_texture *
+softpipe_texture_from_handle(struct pipe_screen *screen,
+                             const struct pipe_texture *template,
+                             struct winsys_handle *whandle)
+{
+   struct sw_winsys *winsys = softpipe_screen(screen)->winsys;
+   struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture);
+   if (!spt)
+      return NULL;
+
+   spt->base = *template;
+   pipe_reference_init(&spt->base.reference, 1);
+   spt->base.screen = screen;
+
+   spt->pot = (util_is_power_of_two(template->width0) &&
+               util_is_power_of_two(template->height0) &&
+               util_is_power_of_two(template->depth0));
+
+   spt->dt = winsys->displaytarget_from_handle(winsys,
+                                               template,
+                                               whandle,
+                                               &spt->stride[0]);
+   if (!spt->dt)
+      goto fail;
+
+   return &spt->base;
+
+ fail:
+   FREE(spt);
+   return NULL;
+}
+
+
+static boolean
+softpipe_texture_get_handle(struct pipe_screen *screen,
+                            struct pipe_texture *pt,
+                            struct winsys_handle *whandle)
+{
+   struct sw_winsys *winsys = softpipe_screen(screen)->winsys;
+   struct softpipe_texture *spt = softpipe_texture(pt);
+
+   assert(spt->dt);
+   if (!spt->dt)
+      return FALSE;
+
+   return winsys->displaytarget_get_handle(winsys, spt->dt, whandle);
+}
+
+
 /**
  * Get a pipe_surface "view" into a texture.
  */
@@ -461,6 +509,8 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen)
 {
    screen->texture_create = softpipe_texture_create;
    screen->texture_destroy = softpipe_texture_destroy;
+   screen->texture_from_handle = softpipe_texture_from_handle;
+   screen->texture_get_handle = softpipe_texture_get_handle;
 
    screen->get_tex_surface = softpipe_get_tex_surface;
    screen->tex_surface_destroy = softpipe_tex_surface_destroy;
diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h
index 791d30edc0e..1f66437dfe1 100644
--- a/src/gallium/drivers/svga/svga_context.h
+++ b/src/gallium/drivers/svga/svga_context.h
@@ -185,7 +185,7 @@ struct svga_state
    const struct svga_sampler_state *sampler[PIPE_MAX_SAMPLERS];
    const struct svga_velems_state *velems;
 
-   struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; /* or texture ID's? */
+   struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS]; /* or texture ID's? */
    struct svga_fragment_shader *fs;
    struct svga_vertex_shader *vs;
 
@@ -208,7 +208,7 @@ struct svga_state
    struct pipe_viewport_state viewport;
 
    unsigned num_samplers;
-   unsigned num_textures;
+   unsigned num_sampler_views;
    unsigned num_vertex_buffers;
    unsigned reduced_prim;
 
diff --git a/src/gallium/drivers/svga/svga_pipe_sampler.c b/src/gallium/drivers/svga/svga_pipe_sampler.c
index 1a8ef296cac..82d525ca33f 100644
--- a/src/gallium/drivers/svga/svga_pipe_sampler.c
+++ b/src/gallium/drivers/svga/svga_pipe_sampler.c
@@ -176,9 +176,36 @@ static void svga_delete_sampler_state(struct pipe_context *pipe,
 }
 
 
-static void svga_set_sampler_textures(struct pipe_context *pipe,
-                                      unsigned num,
-                                      struct pipe_texture **texture)
+static struct pipe_sampler_view *
+svga_create_sampler_view(struct pipe_context *pipe,
+                         struct pipe_texture *texture,
+                         const struct pipe_sampler_view *templ)
+{
+   struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
+
+   if (view) {
+      *view = *templ;
+      view->reference.count = 1;
+      view->texture = NULL;
+      pipe_texture_reference(&view->texture, texture);
+      view->context = pipe;
+   }
+
+   return view;
+}
+
+
+static void
+svga_sampler_view_destroy(struct pipe_context *pipe,
+                          struct pipe_sampler_view *view)
+{
+   pipe_texture_reference(&view->texture, NULL);
+   FREE(view);
+}
+
+static void svga_set_sampler_views(struct pipe_context *pipe,
+                                   unsigned num,
+                                   struct pipe_sampler_view **views)
 {
    struct svga_context *svga = svga_context(pipe);
    unsigned flag_1d = 0;
@@ -188,31 +215,31 @@ static void svga_set_sampler_textures(struct pipe_context *pipe,
    assert(num <= PIPE_MAX_SAMPLERS);
 
    /* Check for no-op */
-   if (num == svga->curr.num_textures &&
-       !memcmp(svga->curr.texture, texture, num * sizeof(struct pipe_texture *))) {
+   if (num == svga->curr.num_sampler_views &&
+       !memcmp(svga->curr.sampler_views, views, num * sizeof(struct pipe_sampler_view *))) {
       if (0) debug_printf("texture noop\n");
       return;
    }
 
    for (i = 0; i < num; i++) {
-      pipe_texture_reference(&svga->curr.texture[i],
-                             texture[i]);
+      pipe_sampler_view_reference(&svga->curr.sampler_views[i],
+                                  views[i]);
 
-      if (!texture[i])
+      if (!views[i])
          continue;
 
-      if (texture[i]->format == PIPE_FORMAT_B8G8R8A8_SRGB)
+      if (views[i]->texture->format == PIPE_FORMAT_B8G8R8A8_SRGB)
          flag_srgb |= 1 << i;
 
-      if (texture[i]->target == PIPE_TEXTURE_1D)
+      if (views[i]->texture->target == PIPE_TEXTURE_1D)
          flag_1d |= 1 << i;
    }
 
-   for (i = num; i < svga->curr.num_textures; i++)
-      pipe_texture_reference(&svga->curr.texture[i],
-                             NULL);
+   for (i = num; i < svga->curr.num_sampler_views; i++)
+      pipe_sampler_view_reference(&svga->curr.sampler_views[i],
+                                  NULL);
 
-   svga->curr.num_textures = num;
+   svga->curr.num_sampler_views = num;
    svga->dirty |= SVGA_NEW_TEXTURE_BINDING;
 
    if (flag_srgb != svga->curr.tex_flags.flag_srgb ||
@@ -231,7 +258,9 @@ void svga_init_sampler_functions( struct svga_context *svga )
    svga->pipe.create_sampler_state = svga_create_sampler_state;
    svga->pipe.bind_fragment_sampler_states = svga_bind_sampler_states;
    svga->pipe.delete_sampler_state = svga_delete_sampler_state;
-   svga->pipe.set_fragment_sampler_textures = svga_set_sampler_textures;
+   svga->pipe.set_fragment_sampler_views = svga_set_sampler_views;
+   svga->pipe.create_sampler_view = svga_create_sampler_view;
+   svga->pipe.sampler_view_destroy = svga_sampler_view_destroy;
 }
 
 
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index 4a058eda885..70d3a8a46ae 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -330,7 +330,7 @@ svga_texture_create(struct pipe_screen *screen,
     */
 #if 0
    if((templat->tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) &&
-      !util_format_is_compressed(templat->format))
+      !util_format_is_s3tc(templat->format))
       tex->key.flags |= SVGA3D_SURFACE_HINT_RENDERTARGET;
 #endif
    
@@ -969,7 +969,7 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt,
       if (min_lod == 0 && max_lod >= pt->last_level)
          view = FALSE;
 
-      if (util_format_is_compressed(pt->format) && view) {
+      if (util_format_is_s3tc(pt->format) && view) {
          format = svga_translate_format_render(pt->format);
       }
 
diff --git a/src/gallium/drivers/svga/svga_state_constants.c b/src/gallium/drivers/svga/svga_state_constants.c
index bb92f818eae..493f78a9908 100644
--- a/src/gallium/drivers/svga/svga_state_constants.c
+++ b/src/gallium/drivers/svga/svga_state_constants.c
@@ -137,7 +137,7 @@ static int emit_fs_consts( struct svga_context *svga,
 
       for (i = 0; i < key->num_textures; i++) {
          if (key->tex[i].unnormalized) {
-            struct pipe_texture *tex = svga->curr.texture[i];
+            struct pipe_texture *tex = svga->curr.sampler_views[i]->texture;
             float data[4];
 
             data[0] = 1.0 / (float)tex->width0;
diff --git a/src/gallium/drivers/svga/svga_state_fs.c b/src/gallium/drivers/svga/svga_state_fs.c
index 2973444d0ab..1310fd9825f 100644
--- a/src/gallium/drivers/svga/svga_state_fs.c
+++ b/src/gallium/drivers/svga/svga_state_fs.c
@@ -158,10 +158,11 @@ static int make_fs_key( const struct svga_context *svga,
     *
     * SVGA_NEW_TEXTURE_BINDING | SVGA_NEW_SAMPLER
     */
-   for (i = 0; i < svga->curr.num_textures; i++) {
-      if (svga->curr.texture[i]) {
+   for (i = 0; i < svga->curr.num_sampler_views; i++) {
+      if (svga->curr.sampler_views[i]) {
          assert(svga->curr.sampler[i]);
-         key->tex[i].texture_target = svga->curr.texture[i]->target;
+         assert(svga->curr.sampler_views[i]->texture);
+         key->tex[i].texture_target = svga->curr.sampler_views[i]->texture->target;
          if (!svga->curr.sampler[i]->normalized_coords) {
             key->tex[i].width_height_idx = idx++;
             key->tex[i].unnormalized = TRUE;
@@ -169,7 +170,7 @@ static int make_fs_key( const struct svga_context *svga,
          }
       }
    }
-   key->num_textures = svga->curr.num_textures;
+   key->num_textures = svga->curr.num_sampler_views;
 
    idx = 0;
    for (i = 0; i < svga->curr.num_samplers; ++i) {
diff --git a/src/gallium/drivers/svga/svga_state_tss.c b/src/gallium/drivers/svga/svga_state_tss.c
index 17b47859781..c08ec7c2e8c 100644
--- a/src/gallium/drivers/svga/svga_state_tss.c
+++ b/src/gallium/drivers/svga/svga_state_tss.c
@@ -37,14 +37,14 @@
 void svga_cleanup_tss_binding(struct svga_context *svga)
 {
    int i;
-   unsigned count = MAX2( svga->curr.num_textures,
+   unsigned count = MAX2( svga->curr.num_sampler_views,
                           svga->state.hw_draw.num_views );
 
    for (i = 0; i < count; i++) {
       struct svga_hw_view_state *view = &svga->state.hw_draw.views[i];
 
       svga_sampler_view_reference(&view->v, NULL);
-      pipe_texture_reference( &svga->curr.texture[i], NULL );
+      pipe_sampler_view_reference( &svga->curr.sampler_views[i], NULL );
       pipe_texture_reference( &view->texture, NULL );
 
       view->dirty = 1;
@@ -57,7 +57,7 @@ update_tss_binding(struct svga_context *svga,
                    unsigned dirty )
 {
    unsigned i;
-   unsigned count = MAX2( svga->curr.num_textures,
+   unsigned count = MAX2( svga->curr.num_sampler_views,
                           svga->state.hw_draw.num_views );
    unsigned min_lod;
    unsigned max_lod;
@@ -77,30 +77,32 @@ update_tss_binding(struct svga_context *svga,
    for (i = 0; i < count; i++) {
       const struct svga_sampler_state *s = svga->curr.sampler[i];
       struct svga_hw_view_state *view = &svga->state.hw_draw.views[i];
+      struct pipe_texture *texture = NULL;
 
       /* get min max lod */
-      if (svga->curr.texture[i]) {
+      if (svga->curr.sampler_views[i]) {
          min_lod = MAX2(s->view_min_lod, 0);
-         max_lod = MIN2(s->view_max_lod, svga->curr.texture[i]->last_level);
+         max_lod = MIN2(s->view_max_lod, svga->curr.sampler_views[i]->texture->last_level);
+         texture = svga->curr.sampler_views[i]->texture;
       } else {
          min_lod = 0;
          max_lod = 0;
       }
 
-      if (view->texture != svga->curr.texture[i] ||
+      if (view->texture != texture ||
           view->min_lod != min_lod ||
           view->max_lod != max_lod) {
 
          svga_sampler_view_reference(&view->v, NULL);
-         pipe_texture_reference( &view->texture, svga->curr.texture[i] );
+         pipe_texture_reference( &view->texture, texture );
 
          view->dirty = TRUE;
          view->min_lod = min_lod;
          view->max_lod = max_lod;
 
-         if (svga->curr.texture[i])
+         if (texture)
             view->v = svga_get_tex_sampler_view(&svga->pipe, 
-                                                svga->curr.texture[i], 
+                                                texture, 
                                                 min_lod,
                                                 max_lod);
       }
@@ -115,7 +117,7 @@ update_tss_binding(struct svga_context *svga,
       }
    }
 
-   svga->state.hw_draw.num_views = svga->curr.num_textures;
+   svga->state.hw_draw.num_views = svga->curr.num_sampler_views;
 
    if (queue.bind_count) {
       SVGA3dTextureState *ts;
diff --git a/src/gallium/drivers/sw/Makefile b/src/gallium/drivers/sw/Makefile
new file mode 100644
index 00000000000..2713a62ee9f
--- /dev/null
+++ b/src/gallium/drivers/sw/Makefile
@@ -0,0 +1,10 @@
+# Meta-driver which combines whichever software rasterizers have been
+# built into a single convenience library.
+
+TOP = ../../../..
+include $(TOP)/configs/current
+
+C_SOURCES = \
+	sw.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/sw/SConscript b/src/gallium/drivers/sw/SConscript
new file mode 100644
index 00000000000..6fbbdf3cc46
--- /dev/null
+++ b/src/gallium/drivers/sw/SConscript
@@ -0,0 +1,42 @@
+#######################################################################
+# SConscript for swrast convenience library
+#
+# This is a meta-driver which consists of any and all of the software
+# rasterizers into a single driver.  A software rasterizer is defined
+# as any driver which takes an sw_winsys pointer as the only argument
+# to create_screen.
+#
+# XXX: unfortunately users of this driver still need to link in any
+# extra libraries needed for the particular driver (eg llvm for
+# llvmpipe).  Not sure how to get around this.
+
+Import('*')
+
+if not set(('softpipe', 'llvmpipe', 'cell')).intersection(env['drivers']):
+    print 'warning: no supported pipe driver: skipping build of sw meta-driver'
+    Return()
+
+env = env.Clone()
+
+if 'softpipe' in env['drivers']:
+    env.Append(CPPDEFINES = 'GALLIUM_SOFTPIPE')
+    env.Prepend(LIBS = [softpipe])
+
+if 'llvmpipe' in env['drivers']:
+    env.Tool('llvm')
+    if 'LLVM_VERSION' in env:
+        env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
+        env.Tool('udis86')
+        env.Prepend(LIBS = [llvmpipe])
+    
+if 'cell' in env['drivers']:
+    env.Append(CPPDEFINES = 'GALLIUM_CELL')
+    env.Prepend(LIBS = [cell])
+
+sw = env.ConvenienceLibrary(
+	target = 'sw',
+	source = [
+		'sw.c',
+		]
+    )
+    Export('sw')
diff --git a/src/gallium/drivers/sw/sw.c b/src/gallium/drivers/sw/sw.c
new file mode 100644
index 00000000000..9f156df45f5
--- /dev/null
+++ b/src/gallium/drivers/sw/sw.c
@@ -0,0 +1,59 @@
+#include "pipe/p_compiler.h"
+#include "util/u_debug.h"
+#include "target-helpers/wrap_screen.h"
+#include "sw_public.h"
+#include "state_tracker/sw_winsys.h"
+
+
+/* Helper function to choose and instantiate one of the software rasterizers:
+ * cell, llvmpipe, softpipe.
+ */
+
+#ifdef GALLIUM_SOFTPIPE
+#include "softpipe/sp_public.h"
+#endif
+
+#ifdef GALLIUM_LLVMPIPE
+#include "llvmpipe/lp_public.h"
+#endif
+
+#ifdef GALLIUM_CELL
+#include "cell/ppu/cell_public.h"
+#endif
+
+struct pipe_screen *
+swrast_create_screen(struct sw_winsys *winsys)
+{
+   const char *default_driver;
+   const char *driver;
+   struct pipe_screen *screen = NULL;
+
+#if defined(GALLIUM_CELL)
+   default_driver = "cell";
+#elif defined(GALLIUM_LLVMPIPE)
+   default_driver = "llvmpipe";
+#elif defined(GALLIUM_SOFTPIPE)
+   default_driver = "softpipe";
+#else
+   default_driver = "";
+#endif
+
+   driver = debug_get_option("GALLIUM_DRIVER", default_driver);
+
+#if defined(GALLIUM_CELL)
+   if (screen == NULL && strcmp(driver, "cell") == 0)
+      screen = cell_create_screen( winsys );
+#endif
+
+#if defined(GALLIUM_LLVMPIPE)
+   if (screen == NULL && strcmp(driver, "llvmpipe") == 0)
+      screen = llvmpipe_create_screen( winsys );
+#endif
+
+#if defined(GALLIUM_SOFTPIPE)
+   if (screen == NULL)
+      screen = softpipe_create_screen( winsys );
+#endif
+
+   return screen;
+}
diff --git a/src/gallium/drivers/sw/sw_public.h b/src/gallium/drivers/sw/sw_public.h
new file mode 100644
index 00000000000..7085c5c85a0
--- /dev/null
+++ b/src/gallium/drivers/sw/sw_public.h
@@ -0,0 +1,13 @@
+#ifndef SW_PUBLIC_H
+#define SW_PUBLIC_H
+
+/* A convenience library, primarily to isolate the logic required to
+ * figure out which if any software rasterizers have been built and
+ * select between them.
+ */
+struct sw_winsys;
+
+struct pipe_screen *
+swrast_create_screen(struct sw_winsys *winsys);
+
+#endif
diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c
index b7e6bbac68e..5c24bd1f7df 100644
--- a/src/gallium/drivers/trace/tr_context.c
+++ b/src/gallium/drivers/trace/tr_context.c
@@ -25,6 +25,7 @@
  *
  **************************************************************************/
 
+#include "util/u_inlines.h"
 #include "util/u_memory.h"
 #include "util/u_simple_list.h"
 #include "util/u_format.h"
@@ -114,7 +115,7 @@ trace_context_draw_block(struct trace_context *tr_ctx, int flag)
                    (void *) tr_ctx->draw_rule.fs, (void *) tr_ctx->curr.fs,
                    (void *) tr_ctx->draw_rule.vs, (void *) tr_ctx->curr.vs,
                    (void *) tr_ctx->draw_rule.surf, 0,
-                   (void *) tr_ctx->draw_rule.tex, 0);
+                   (void *) tr_ctx->draw_rule.sampler_view, 0);
       if (tr_ctx->draw_rule.fs &&
           tr_ctx->draw_rule.fs == tr_ctx->curr.fs)
          block = TRUE;
@@ -128,12 +129,12 @@ trace_context_draw_block(struct trace_context *tr_ctx, int flag)
          for (k = 0; k < tr_ctx->curr.nr_cbufs; k++)
             if (tr_ctx->draw_rule.surf == tr_ctx->curr.cbufs[k])
                block = TRUE;
-      if (tr_ctx->draw_rule.tex) {
-         for (k = 0; k < tr_ctx->curr.num_texs; k++)
-            if (tr_ctx->draw_rule.tex == tr_ctx->curr.tex[k])
+      if (tr_ctx->draw_rule.sampler_view) {
+         for (k = 0; k < tr_ctx->curr.num_sampler_views; k++)
+            if (tr_ctx->draw_rule.sampler_view == tr_ctx->curr.sampler_views[k])
                block = TRUE;
-         for (k = 0; k < tr_ctx->curr.num_vert_texs; k++) {
-            if (tr_ctx->draw_rule.tex == tr_ctx->curr.vert_tex[k]) {
+         for (k = 0; k < tr_ctx->curr.num_vert_sampler_views; k++) {
+            if (tr_ctx->draw_rule.sampler_view == tr_ctx->curr.vert_sampler_views[k]) {
                block = TRUE;
             }
          }
@@ -1015,63 +1016,119 @@ trace_context_set_viewport_state(struct pipe_context *_pipe,
 }
 
 
+static struct pipe_sampler_view *
+trace_create_sampler_view(struct pipe_context *_pipe,
+                          struct pipe_texture *_texture,
+                          const struct pipe_sampler_view *templ)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_texture *tr_tex = trace_texture(_texture);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   struct pipe_texture *texture = tr_tex->texture;
+   struct trace_sampler_view *result = CALLOC_STRUCT(trace_sampler_view);
+
+   trace_dump_call_begin("pipe_context", "create_sampler_view");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, texture);
+   trace_dump_arg(ptr, templ);
+
+   result->sampler_view = pipe->create_sampler_view(pipe, texture, templ);
+
+   result->base = *templ;
+   result->base.reference.count = 1;
+   result->base.texture = NULL;
+   pipe_texture_reference(&result->base.texture, _texture);
+   result->base.context = _pipe;
+
+   trace_dump_ret(ptr, result);
+
+   trace_dump_call_end();
+
+   return &result->base;
+}
+
+
+static void
+trace_sampler_view_destroy(struct pipe_context *_pipe,
+                           struct pipe_sampler_view *_view)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_sampler_view *tr_view = trace_sampler_view(_view);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   struct pipe_sampler_view *view = tr_view->sampler_view;
+
+   trace_dump_call_begin("pipe_context", "sampler_view_destroy");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, view);
+
+   pipe->sampler_view_destroy(pipe, view);
+
+   trace_dump_call_end();
+
+   pipe_texture_reference(&_view->texture, NULL);
+   FREE(_view);
+}
+
+
 static INLINE void
-trace_context_set_fragment_sampler_textures(struct pipe_context *_pipe,
-                                            unsigned num_textures,
-                                            struct pipe_texture **textures)
+trace_context_set_fragment_sampler_views(struct pipe_context *_pipe,
+                                         unsigned num,
+                                         struct pipe_sampler_view **views)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
-   struct trace_texture *tr_tex;
+   struct trace_sampler_view *tr_view;
    struct pipe_context *pipe = tr_ctx->pipe;
-   struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_view *unwrapped_views[PIPE_MAX_SAMPLERS];
    unsigned i;
 
-   tr_ctx->curr.num_texs = num_textures;
-   for(i = 0; i < num_textures; ++i) {
-      tr_tex = trace_texture(textures[i]);
-      tr_ctx->curr.tex[i] = tr_tex;
-      unwrapped_textures[i] = tr_tex ? tr_tex->texture : NULL;
+   tr_ctx->curr.num_sampler_views = num;
+   for(i = 0; i < num; ++i) {
+      tr_view = trace_sampler_view(views[i]);
+      tr_ctx->curr.sampler_views[i] = tr_view;
+      unwrapped_views[i] = tr_view ? tr_view->sampler_view : NULL;
    }
-   textures = unwrapped_textures;
+   views = unwrapped_views;
 
-   trace_dump_call_begin("pipe_context", "set_fragment_sampler_textures");
+   trace_dump_call_begin("pipe_context", "set_fragment_sampler_views");
 
    trace_dump_arg(ptr, pipe);
-   trace_dump_arg(uint, num_textures);
-   trace_dump_arg_array(ptr, textures, num_textures);
+   trace_dump_arg(uint, num);
+   trace_dump_arg_array(ptr, views, num);
 
-   pipe->set_fragment_sampler_textures(pipe, num_textures, textures);
+   pipe->set_fragment_sampler_views(pipe, num, views);
 
    trace_dump_call_end();
 }
 
 
 static INLINE void
-trace_context_set_vertex_sampler_textures(struct pipe_context *_pipe,
-                                          unsigned num_textures,
-                                          struct pipe_texture **textures)
+trace_context_set_vertex_sampler_views(struct pipe_context *_pipe,
+                                       unsigned num,
+                                       struct pipe_sampler_view **views)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
-   struct trace_texture *tr_tex;
+   struct trace_sampler_view *tr_view;
    struct pipe_context *pipe = tr_ctx->pipe;
-   struct pipe_texture *unwrapped_textures[PIPE_MAX_VERTEX_SAMPLERS];
+   struct pipe_sampler_view *unwrapped_views[PIPE_MAX_VERTEX_SAMPLERS];
    unsigned i;
 
-   tr_ctx->curr.num_vert_texs = num_textures;
-   for(i = 0; i < num_textures; ++i) {
-      tr_tex = trace_texture(textures[i]);
-      tr_ctx->curr.vert_tex[i] = tr_tex;
-      unwrapped_textures[i] = tr_tex ? tr_tex->texture : NULL;
+   tr_ctx->curr.num_vert_sampler_views = num;
+   for(i = 0; i < num; ++i) {
+      tr_view = trace_sampler_view(views[i]);
+      tr_ctx->curr.vert_sampler_views[i] = tr_view;
+      unwrapped_views[i] = tr_view ? tr_view->sampler_view : NULL;
    }
-   textures = unwrapped_textures;
+   views = unwrapped_views;
 
-   trace_dump_call_begin("pipe_context", "set_vertex_sampler_textures");
+   trace_dump_call_begin("pipe_context", "set_vertex_sampler_views");
 
    trace_dump_arg(ptr, pipe);
-   trace_dump_arg(uint, num_textures);
-   trace_dump_arg_array(ptr, textures, num_textures);
+   trace_dump_arg(uint, num);
+   trace_dump_arg_array(ptr, views, num);
 
-   pipe->set_vertex_sampler_textures(pipe, num_textures, textures);
+   pipe->set_vertex_sampler_views(pipe, num, views);
 
    trace_dump_call_end();
 }
@@ -1487,8 +1544,10 @@ trace_context_create(struct trace_screen *tr_scr,
    tr_ctx->base.set_polygon_stipple = trace_context_set_polygon_stipple;
    tr_ctx->base.set_scissor_state = trace_context_set_scissor_state;
    tr_ctx->base.set_viewport_state = trace_context_set_viewport_state;
-   tr_ctx->base.set_fragment_sampler_textures = trace_context_set_fragment_sampler_textures;
-   tr_ctx->base.set_vertex_sampler_textures = trace_context_set_vertex_sampler_textures;
+   tr_ctx->base.set_fragment_sampler_views = trace_context_set_fragment_sampler_views;
+   tr_ctx->base.set_vertex_sampler_views = trace_context_set_vertex_sampler_views;
+   tr_ctx->base.create_sampler_view = trace_create_sampler_view;
+   tr_ctx->base.sampler_view_destroy = trace_sampler_view_destroy;
    tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers;
    if (pipe->surface_copy)
       tr_ctx->base.surface_copy = trace_context_surface_copy;
diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h
index 14284232485..feec9b6bbf3 100644
--- a/src/gallium/drivers/trace/tr_context.h
+++ b/src/gallium/drivers/trace/tr_context.h
@@ -53,11 +53,11 @@ struct trace_context
       struct trace_shader *fs;
       struct trace_shader *vs;
 
-      struct trace_texture *tex[PIPE_MAX_SAMPLERS];
-      unsigned num_texs;
+      struct trace_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
+      unsigned num_sampler_views;
 
-      struct trace_texture *vert_tex[PIPE_MAX_VERTEX_SAMPLERS];
-      unsigned num_vert_texs;
+      struct trace_sampler_view *vert_sampler_views[PIPE_MAX_VERTEX_SAMPLERS];
+      unsigned num_vert_sampler_views;
 
       unsigned nr_cbufs;
       struct trace_texture *cbufs[PIPE_MAX_COLOR_BUFS];
@@ -68,7 +68,7 @@ struct trace_context
       struct trace_shader *fs;
       struct trace_shader *vs;
 
-      struct trace_texture *tex;
+      struct trace_sampler_view *sampler_view;
       struct trace_texture *surf;
 
       int blocker;
diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c
index f4f17566fd3..53ab8c686d8 100644
--- a/src/gallium/drivers/trace/tr_rbug.c
+++ b/src/gallium/drivers/trace/tr_rbug.c
@@ -313,12 +313,12 @@ trace_rbug_context_info(struct trace_rbug *tr_rbug, struct rbug_header *header,
    for (i = 0; i < tr_ctx->curr.nr_cbufs; i++)
       cbufs[i] = VOID2U64(tr_ctx->curr.cbufs[i]);
 
-   for (i = 0; i < tr_ctx->curr.num_texs; i++)
-      texs[i] = VOID2U64(tr_ctx->curr.tex[i]);
+   for (i = 0; i < tr_ctx->curr.num_sampler_views; i++)
+      texs[i] = VOID2U64(tr_ctx->curr.sampler_views[i]);
 
    rbug_send_context_info_reply(tr_rbug->con, serial,
                                 VOID2U64(tr_ctx->curr.vs), VOID2U64(tr_ctx->curr.fs),
-                                texs, tr_ctx->curr.num_texs,
+                                texs, tr_ctx->curr.num_sampler_views,
                                 cbufs, tr_ctx->curr.nr_cbufs,
                                 VOID2U64(tr_ctx->curr.zsbuf),
                                 tr_ctx->draw_blocker, tr_ctx->draw_blocked, NULL);
@@ -444,7 +444,7 @@ trace_rbug_context_draw_rule(struct trace_rbug *tr_rbug, struct rbug_header *hea
    pipe_mutex_lock(tr_ctx->draw_mutex);
    tr_ctx->draw_rule.vs = U642VOID(rule->vertex);
    tr_ctx->draw_rule.fs = U642VOID(rule->fragment);
-   tr_ctx->draw_rule.tex = U642VOID(rule->texture);
+   tr_ctx->draw_rule.sampler_view = U642VOID(rule->texture);
    tr_ctx->draw_rule.surf = U642VOID(rule->surface);
    tr_ctx->draw_rule.blocker = rule->block;
    tr_ctx->draw_blocker |= RBUG_BLOCK_RULE;
diff --git a/src/gallium/drivers/trace/tr_texture.h b/src/gallium/drivers/trace/tr_texture.h
index 4dc95308a79..66250465e44 100644
--- a/src/gallium/drivers/trace/tr_texture.h
+++ b/src/gallium/drivers/trace/tr_texture.h
@@ -56,6 +56,14 @@ struct trace_surface
 };
 
 
+struct trace_sampler_view
+{
+   struct pipe_sampler_view base;
+
+   struct pipe_sampler_view *sampler_view;
+};
+
+
 struct trace_transfer
 {
    struct pipe_transfer base;
@@ -89,6 +97,15 @@ trace_surface(struct pipe_surface *surface)
 }
 
 
+static INLINE struct trace_sampler_view *
+trace_sampler_view(struct pipe_sampler_view *sampler_view)
+{
+   if (!sampler_view)
+      return NULL;
+   return (struct trace_sampler_view *)sampler_view;
+}
+
+
 static INLINE struct trace_transfer *
 trace_transfer(struct pipe_transfer *transfer)
 {