17 files changed, 322 insertions, 347 deletions

diff --git a/src/gallium/drivers/svga/svga_context.c b/src/gallium/drivers/svga/svga_context.c
index d499ae6acc..adb7840182 100644
--- a/src/gallium/drivers/svga/svga_context.c
+++ b/src/gallium/drivers/svga/svga_context.c
@@ -164,6 +164,8 @@ struct pipe_context *svga_context_create( struct pipe_screen *screen,
    svga_init_constbuffer_functions(svga);
    svga_init_query_functions(svga);
 
+   svga_init_texture_functions(&svga->pipe);
+
    /* debug */
    svga->debug.no_swtnl = debug_get_bool_option("SVGA_NO_SWTNL", FALSE);
    svga->debug.force_swtnl = debug_get_bool_option("SVGA_FORCE_SWTNL", FALSE);
diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h
index 03302e2a6e..791d30edc0 100644
--- a/src/gallium/drivers/svga/svga_context.h
+++ b/src/gallium/drivers/svga/svga_context.h
@@ -169,6 +169,11 @@ struct svga_sampler_state {
    unsigned view_max_lod;
 };
 
+struct svga_velems_state {
+   unsigned count;
+   struct pipe_vertex_element velem[PIPE_MAX_ATTRIBS];
+};
+
 /* Use to calculate differences between state emitted to hardware and
  * current driver-calculated state.  
  */
@@ -178,13 +183,13 @@ struct svga_state
    const struct svga_depth_stencil_state *depth;
    const struct svga_rasterizer_state *rast;
    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 svga_fragment_shader *fs;
    struct svga_vertex_shader *vs;
 
    struct pipe_vertex_buffer vb[PIPE_MAX_ATTRIBS];
-   struct pipe_vertex_element ve[PIPE_MAX_ATTRIBS];
    struct pipe_buffer *cb[PIPE_SHADER_TYPES];
 
    struct pipe_framebuffer_state framebuffer;
@@ -204,7 +209,6 @@ struct svga_state
 
    unsigned num_samplers;
    unsigned num_textures;
-   unsigned num_vertex_elements;
    unsigned num_vertex_buffers;
    unsigned reduced_prim;
 
diff --git a/src/gallium/drivers/svga/svga_pipe_clear.c b/src/gallium/drivers/svga/svga_pipe_clear.c
index 409b3b41cb..8483a3fad7 100644
--- a/src/gallium/drivers/svga/svga_pipe_clear.c
+++ b/src/gallium/drivers/svga/svga_pipe_clear.c
@@ -54,7 +54,7 @@ try_clear(struct svga_context *svga,
 
    if ((buffers & PIPE_CLEAR_COLOR) && fb->cbufs[0]) {
       flags |= SVGA3D_CLEAR_COLOR;
-      util_pack_color(rgba, PIPE_FORMAT_A8R8G8B8_UNORM, &uc);
+      util_pack_color(rgba, PIPE_FORMAT_B8G8R8A8_UNORM, &uc);
 
       rect.w = fb->cbufs[0]->width;
       rect.h = fb->cbufs[0]->height;
@@ -63,7 +63,7 @@ try_clear(struct svga_context *svga,
    if ((buffers & PIPE_CLEAR_DEPTHSTENCIL) && fb->zsbuf) {
       flags |= SVGA3D_CLEAR_DEPTH;
 
-      if (svga->curr.framebuffer.zsbuf->format == PIPE_FORMAT_Z24S8_UNORM)
+      if (svga->curr.framebuffer.zsbuf->format == PIPE_FORMAT_S8Z24_UNORM)
          flags |= SVGA3D_CLEAR_STENCIL;
 
       rect.w = MAX2(rect.w, fb->zsbuf->width);
diff --git a/src/gallium/drivers/svga/svga_pipe_misc.c b/src/gallium/drivers/svga/svga_pipe_misc.c
index 49b43bebc2..95bf0e6f91 100644
--- a/src/gallium/drivers/svga/svga_pipe_misc.c
+++ b/src/gallium/drivers/svga/svga_pipe_misc.c
@@ -118,10 +118,10 @@ static void svga_set_framebuffer_state(struct pipe_context *pipe,
       case PIPE_FORMAT_Z16_UNORM:
          svga->curr.depthscale = 1.0f / DEPTH_BIAS_SCALE_FACTOR_D16;
          break;
-      case PIPE_FORMAT_S8Z24_UNORM:
-      case PIPE_FORMAT_X8Z24_UNORM:
       case PIPE_FORMAT_Z24S8_UNORM:
       case PIPE_FORMAT_Z24X8_UNORM:
+      case PIPE_FORMAT_S8Z24_UNORM:
+      case PIPE_FORMAT_X8Z24_UNORM:
          svga->curr.depthscale = 1.0f / DEPTH_BIAS_SCALE_FACTOR_D24S8;
          break;
       case PIPE_FORMAT_Z32_UNORM:
diff --git a/src/gallium/drivers/svga/svga_pipe_sampler.c b/src/gallium/drivers/svga/svga_pipe_sampler.c
index 224c4f4c18..1a8ef296ca 100644
--- a/src/gallium/drivers/svga/svga_pipe_sampler.c
+++ b/src/gallium/drivers/svga/svga_pipe_sampler.c
@@ -155,7 +155,7 @@ static void svga_bind_sampler_states(struct pipe_context *pipe,
    /* Check for no-op */
    if (num == svga->curr.num_samplers &&
        !memcmp(svga->curr.sampler, sampler, num * sizeof(void *))) {
-      debug_printf("sampler noop\n");
+      if (0) debug_printf("sampler noop\n");
       return;
    }
 
@@ -201,7 +201,7 @@ static void svga_set_sampler_textures(struct pipe_context *pipe,
       if (!texture[i])
          continue;
 
-      if (texture[i]->format == PIPE_FORMAT_A8R8G8B8_SRGB)
+      if (texture[i]->format == PIPE_FORMAT_B8G8R8A8_SRGB)
          flag_srgb |= 1 << i;
 
       if (texture[i]->target == PIPE_TEXTURE_1D)
diff --git a/src/gallium/drivers/svga/svga_pipe_vertex.c b/src/gallium/drivers/svga/svga_pipe_vertex.c
index 836b8441da..1715a47fc6 100644
--- a/src/gallium/drivers/svga/svga_pipe_vertex.c
+++ b/src/gallium/drivers/svga/svga_pipe_vertex.c
@@ -26,6 +26,7 @@
 #include "util/u_inlines.h"
 #include "pipe/p_defines.h"
 #include "util/u_math.h"
+#include "util/u_memory.h"
 #include "tgsi/tgsi_parse.h"
 
 #include "svga_screen.h"
@@ -64,20 +65,37 @@ static void svga_set_vertex_buffers(struct pipe_context *pipe,
    svga->dirty |= SVGA_NEW_VBUFFER;
 }
 
-static void svga_set_vertex_elements(struct pipe_context *pipe,
-                                     unsigned count,
-                                     const struct pipe_vertex_element *elements)
+
+static void *
+svga_create_vertex_elements_state(struct pipe_context *pipe,
+                                  unsigned count,
+                                  const struct pipe_vertex_element *attribs)
 {
-   struct svga_context *svga = svga_context(pipe);
-   unsigned i;
+   struct svga_velems_state *velems;
+   assert(count <= PIPE_MAX_ATTRIBS);
+   velems = (struct svga_velems_state *) MALLOC(sizeof(struct svga_velems_state));
+   if (velems) {
+      velems->count = count;
+      memcpy(velems->velem, attribs, sizeof(*attribs) * count);
+   }
+   return velems;
+}
 
-   for (i = 0; i < count; i++)
-      svga->curr.ve[i] = elements[i];
+static void svga_bind_vertex_elements_state(struct pipe_context *pipe,
+                                            void *velems)
+{
+   struct svga_context *svga = svga_context(pipe);
+   struct svga_velems_state *svga_velems = (struct svga_velems_state *) velems;
 
-   svga->curr.num_vertex_elements = count;
+   svga->curr.velems = svga_velems;
    svga->dirty |= SVGA_NEW_VELEMENT;
 }
 
+static void svga_delete_vertex_elements_state(struct pipe_context *pipe,
+                                              void *velems)
+{
+   FREE(velems);
+}
 
 void svga_cleanup_vertex_state( struct svga_context *svga )
 {
@@ -91,7 +109,9 @@ void svga_cleanup_vertex_state( struct svga_context *svga )
 void svga_init_vertex_functions( struct svga_context *svga )
 {
    svga->pipe.set_vertex_buffers = svga_set_vertex_buffers;
-   svga->pipe.set_vertex_elements = svga_set_vertex_elements;
+   svga->pipe.create_vertex_elements_state = svga_create_vertex_elements_state;
+   svga->pipe.bind_vertex_elements_state = svga_bind_vertex_elements_state;
+   svga->pipe.delete_vertex_elements_state = svga_delete_vertex_elements_state;
 }
 
 
diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c
index 414ac52e1f..6022c38cfc 100644
--- a/src/gallium/drivers/svga/svga_screen.c
+++ b/src/gallium/drivers/svga/svga_screen.c
@@ -196,23 +196,23 @@ svga_translate_format_cap(enum pipe_format format)
 {
    switch(format) {
    
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_A8R8G8B8;
-   case PIPE_FORMAT_X8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_X8R8G8B8;
 
-   case PIPE_FORMAT_R5G6B5_UNORM:
+   case PIPE_FORMAT_B5G6R5_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_R5G6B5;
-   case PIPE_FORMAT_A1R5G5B5_UNORM:
+   case PIPE_FORMAT_B5G5R5A1_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_A1R5G5B5;
-   case PIPE_FORMAT_A4R4G4B4_UNORM:
+   case PIPE_FORMAT_B4G4R4A4_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_A4R4G4B4;
 
    case PIPE_FORMAT_Z16_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_Z_D16;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_Z_D24S8;
-   case PIPE_FORMAT_Z24X8_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
       return SVGA3D_DEVCAP_SURFACEFMT_Z_D24X8;
 
    case PIPE_FORMAT_A8_UNORM:
@@ -254,8 +254,8 @@ svga_is_format_supported( struct pipe_screen *screen,
       /* Often unsupported/problematic. This means we end up with the same
        * visuals for all virtual hardware implementations.
        */
-      case PIPE_FORMAT_A4R4G4B4_UNORM:
-      case PIPE_FORMAT_A1R5G5B5_UNORM:
+      case PIPE_FORMAT_B4G4R4A4_UNORM:
+      case PIPE_FORMAT_B5G5R5A1_UNORM:
          return FALSE;
          
       /* Simulate ability to render into compressed textures */
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index 12f3531a1d..4a058eda88 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -41,8 +41,6 @@
 #include "svga_debug.h"
 #include "svga_screen_buffer.h"
 
-#include <util/u_string.h>
-
 
 /* XXX: This isn't a real hardware flag, but just a hack for kernel to
  * know about primary surfaces. Find a better way to accomplish this.
@@ -59,21 +57,21 @@ svga_translate_format(enum pipe_format format)
 {
    switch(format) {
    
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
       return SVGA3D_A8R8G8B8;
-   case PIPE_FORMAT_X8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
       return SVGA3D_X8R8G8B8;
 
       /* Required for GL2.1:
        */
-   case PIPE_FORMAT_A8R8G8B8_SRGB:
+   case PIPE_FORMAT_B8G8R8A8_SRGB:
       return SVGA3D_A8R8G8B8;
 
-   case PIPE_FORMAT_R5G6B5_UNORM:
+   case PIPE_FORMAT_B5G6R5_UNORM:
       return SVGA3D_R5G6B5;
-   case PIPE_FORMAT_A1R5G5B5_UNORM:
+   case PIPE_FORMAT_B5G5R5A1_UNORM:
       return SVGA3D_A1R5G5B5;
-   case PIPE_FORMAT_A4R4G4B4_UNORM:
+   case PIPE_FORMAT_B4G4R4A4_UNORM:
       return SVGA3D_A4R4G4B4;
 
       
@@ -83,9 +81,9 @@ svga_translate_format(enum pipe_format format)
     */
    case PIPE_FORMAT_Z16_UNORM:
       return SVGA3D_Z_D16;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
       return SVGA3D_Z_D24S8;
-   case PIPE_FORMAT_Z24X8_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
       return SVGA3D_Z_D24X8;
 
    case PIPE_FORMAT_A8_UNORM:
@@ -111,13 +109,13 @@ SVGA3dSurfaceFormat
 svga_translate_format_render(enum pipe_format format)
 {
    switch(format) { 
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
-   case PIPE_FORMAT_X8R8G8B8_UNORM:
-   case PIPE_FORMAT_A1R5G5B5_UNORM:
-   case PIPE_FORMAT_A4R4G4B4_UNORM:
-   case PIPE_FORMAT_R5G6B5_UNORM:
-   case PIPE_FORMAT_Z24S8_UNORM:
-   case PIPE_FORMAT_Z24X8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+   case PIPE_FORMAT_B5G5R5A1_UNORM:
+   case PIPE_FORMAT_B4G4R4A4_UNORM:
+   case PIPE_FORMAT_B5G6R5_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
    case PIPE_FORMAT_Z32_UNORM:
    case PIPE_FORMAT_Z16_UNORM:
    case PIPE_FORMAT_L8_UNORM:
@@ -315,7 +313,11 @@ svga_texture_create(struct pipe_screen *screen,
       tex->key.cachable = 0;
    }
 
-   if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) {
+   if(templat->tex_usage & PIPE_TEXTURE_USAGE_SHARED) {
+      tex->key.cachable = 0;
+   }
+
+   if(templat->tex_usage & PIPE_TEXTURE_USAGE_SCANOUT) {
       tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT;
       tex->key.cachable = 0;
    }
@@ -355,80 +357,18 @@ error1:
 }
 
 
-static struct pipe_texture *
-svga_texture_blanket(struct pipe_screen * screen,
-                     const struct pipe_texture *base,
-                     const unsigned *stride,
-                     struct pipe_buffer *buffer)
-{
-   struct svga_texture *tex;
-   struct svga_buffer *sbuf = svga_buffer(buffer);
-   struct svga_winsys_screen *sws = svga_winsys_screen(screen);
-   assert(screen);
-
-   /* Only supports one type */
-   if (base->target != PIPE_TEXTURE_2D ||
-       base->last_level != 0 ||
-       base->depth0 != 1) {
-      return NULL;
-   }
-
-   /**
-    * We currently can't do texture blanket on
-    * SVGA3D_BUFFER. Need to blit to a temporary surface?
-    */
-
-   assert(sbuf->handle);
-   if (!sbuf->handle)
-      return NULL;
-
-   if (svga_translate_format(base->format) != sbuf->key.format) {
-      unsigned f1 = svga_translate_format(base->format);
-      unsigned f2 = sbuf->key.format;
-
-      /* It's okay for XRGB and ARGB or depth with/out stencil to get mixed up */
-      if ( !( (f1 == SVGA3D_X8R8G8B8 && f2 == SVGA3D_A8R8G8B8) ||
-              (f1 == SVGA3D_A8R8G8B8 && f2 == SVGA3D_X8R8G8B8) ||
-              (f1 == SVGA3D_Z_D24X8 && f2 == SVGA3D_Z_D24S8) ) ) {
-         debug_printf("%s wrong format %u != %u\n", __FUNCTION__, f1, f2);
-         return NULL;
-      }
-   }
-
-   tex = CALLOC_STRUCT(svga_texture);
-   if (!tex)
-      return NULL;
-
-   tex->base = *base;
-   
-
-   if (sbuf->key.format == 1)
-      tex->base.format = PIPE_FORMAT_X8R8G8B8_UNORM;
-   else if (sbuf->key.format == 2)
-      tex->base.format = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-   pipe_reference_init(&tex->base.reference, 1);
-   tex->base.screen = screen;
-
-   SVGA_DBG(DEBUG_DMA, "blanket sid %p\n", sbuf->handle);
-
-   /* We don't own this storage, so don't try to cache it.
-    */
-   assert(sbuf->key.cachable == 0);
-   tex->key.cachable = 0;
-   sws->surface_reference(sws, &tex->handle, sbuf->handle);
 
-   return &tex->base;
-}
 
 
-struct pipe_texture *
-svga_screen_texture_wrap_surface(struct pipe_screen *screen,
-				 struct pipe_texture *base,
-				 enum SVGA3dSurfaceFormat format,
-				 struct svga_winsys_surface *srf)
+static struct pipe_texture *
+svga_screen_texture_from_handle(struct pipe_screen *screen,
+                                const struct pipe_texture *base,
+                                struct winsys_handle *whandle)
 {
+   struct svga_winsys_screen *sws = svga_winsys_screen(screen);
+   struct svga_winsys_surface *srf;
    struct svga_texture *tex;
+   enum SVGA3dSurfaceFormat format = 0;
    assert(screen);
 
    /* Only supports one type */
@@ -438,6 +378,8 @@ svga_screen_texture_wrap_surface(struct pipe_screen *screen,
       return NULL;
    }
 
+   srf = sws->surface_from_handle(sws, whandle, &format);
+
    if (!srf)
       return NULL;
 
@@ -462,9 +404,9 @@ svga_screen_texture_wrap_surface(struct pipe_screen *screen,
    
 
    if (format == 1)
-      tex->base.format = PIPE_FORMAT_X8R8G8B8_UNORM;
+      tex->base.format = PIPE_FORMAT_B8G8R8X8_UNORM;
    else if (format == 2)
-      tex->base.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+      tex->base.format = PIPE_FORMAT_B8G8R8A8_UNORM;
 
    pipe_reference_init(&tex->base.reference, 1);
    tex->base.screen = screen;
@@ -478,6 +420,22 @@ svga_screen_texture_wrap_surface(struct pipe_screen *screen,
 }
 
 
+static boolean 
+svga_screen_texture_get_handle(struct pipe_screen *screen,
+                               struct pipe_texture *texture,
+                               struct winsys_handle *whandle)
+{
+   struct svga_winsys_screen *sws = svga_winsys_screen(texture->screen);
+   unsigned stride;
+
+   assert(svga_texture(texture)->key.cachable == 0);
+   svga_texture(texture)->key.cachable = 0;
+   stride = util_format_get_nblocksx(texture->format, texture->width0) *
+            util_format_get_blocksize(texture->format);
+   return sws->surface_get_handle(sws, svga_texture(texture)->handle, stride, whandle);
+}
+
+
 static void
 svga_texture_destroy(struct pipe_texture *pt)
 {
@@ -823,15 +781,17 @@ svga_surface_needs_propagation(struct pipe_surface *surf)
    return s->dirty && s->handle != tex->handle;
 }
 
-
+/* XXX: Still implementing this as if it was a screen function, but
+ * can now modify it to queue transfers on the context.
+ */
 static struct pipe_transfer *
-svga_get_tex_transfer(struct pipe_screen *screen,
-                     struct pipe_texture *texture,
-                     unsigned face, unsigned level, unsigned zslice,
-                     enum pipe_transfer_usage usage, unsigned x, unsigned y,
-                     unsigned w, unsigned h)
+svga_get_tex_transfer(struct pipe_context *pipe,
+		      struct pipe_texture *texture,
+		      unsigned face, unsigned level, unsigned zslice,
+		      enum pipe_transfer_usage usage, unsigned x, unsigned y,
+		      unsigned w, unsigned h)
 {
-   struct svga_screen *ss = svga_screen(screen);
+   struct svga_screen *ss = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st;
    unsigned nblocksx = util_format_get_nblocksx(texture->format, w);
@@ -899,11 +859,14 @@ no_hwbuf:
 }
 
 
+/* XXX: Still implementing this as if it was a screen function, but
+ * can now modify it to queue transfers on the context.
+ */
 static void *
-svga_transfer_map( struct pipe_screen *screen,
+svga_transfer_map( struct pipe_context *pipe,
                    struct pipe_transfer *transfer )
 {
-   struct svga_screen *ss = svga_screen(screen);
+   struct svga_screen *ss = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st = svga_transfer(transfer);
 
@@ -917,11 +880,14 @@ svga_transfer_map( struct pipe_screen *screen,
 }
 
 
+/* XXX: Still implementing this as if it was a screen function, but
+ * can now modify it to queue transfers on the context.
+ */
 static void
-svga_transfer_unmap(struct pipe_screen *screen,
+svga_transfer_unmap(struct pipe_context *pipe,
                     struct pipe_transfer *transfer)
 {
-   struct svga_screen *ss = svga_screen(screen);
+   struct svga_screen *ss = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st = svga_transfer(transfer);
    
@@ -931,10 +897,11 @@ svga_transfer_unmap(struct pipe_screen *screen,
 
 
 static void
-svga_tex_transfer_destroy(struct pipe_transfer *transfer)
+svga_tex_transfer_destroy(struct pipe_context *pipe,
+                          struct pipe_transfer *transfer)
 {
    struct svga_texture *tex = svga_texture(transfer->texture);
-   struct svga_screen *ss = svga_screen(transfer->texture->screen);
+   struct svga_screen *ss = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st = svga_transfer(transfer);
 
@@ -951,18 +918,26 @@ svga_tex_transfer_destroy(struct pipe_transfer *transfer)
    FREE(st);
 }
 
+
+void
+svga_init_texture_functions(struct pipe_context *pipe)
+{
+   pipe->get_tex_transfer = svga_get_tex_transfer;
+   pipe->transfer_map = svga_transfer_map;
+   pipe->transfer_unmap = svga_transfer_unmap;
+   pipe->tex_transfer_destroy = svga_tex_transfer_destroy;
+}
+
+
 void
 svga_screen_init_texture_functions(struct pipe_screen *screen)
 {
    screen->texture_create = svga_texture_create;
+   screen->texture_from_handle = svga_screen_texture_from_handle;
+   screen->texture_get_handle = svga_screen_texture_get_handle;
    screen->texture_destroy = svga_texture_destroy;
    screen->get_tex_surface = svga_get_tex_surface;
    screen->tex_surface_destroy = svga_tex_surface_destroy;
-   screen->texture_blanket = svga_texture_blanket;
-   screen->get_tex_transfer = svga_get_tex_transfer;
-   screen->transfer_map = svga_transfer_map;
-   screen->transfer_unmap = svga_transfer_unmap;
-   screen->tex_transfer_destroy = svga_tex_transfer_destroy;
 }
 
 /*********************************************************************** 
@@ -1120,33 +1095,3 @@ svga_destroy_sampler_view_priv(struct svga_sampler_view *v)
    pipe_texture_reference(&v->texture, NULL);
    FREE(v);
 }
-
-boolean
-svga_screen_buffer_from_texture(struct pipe_texture *texture,
-				struct pipe_buffer **buffer,
-				unsigned *stride)
-{
-   struct svga_texture *stex = svga_texture(texture);
-
-   *buffer = svga_screen_buffer_wrap_surface
-      (texture->screen,
-       svga_translate_format(texture->format),
-       stex->handle);
-
-   *stride = util_format_get_stride(texture->format, texture->width0);
-
-   return *buffer != NULL;
-}
-
-
-struct svga_winsys_surface *
-svga_screen_texture_get_winsys_surface(struct pipe_texture *texture)
-{
-   struct svga_winsys_screen *sws = svga_winsys_screen(texture->screen);
-   struct svga_winsys_surface *vsurf = NULL;
-
-   assert(svga_texture(texture)->key.cachable == 0);
-   svga_texture(texture)->key.cachable = 0;
-   sws->surface_reference(sws, &vsurf, svga_texture(texture)->handle);
-   return vsurf;
-}
diff --git a/src/gallium/drivers/svga/svga_screen_texture.h b/src/gallium/drivers/svga/svga_screen_texture.h
index 24c1f78ca5..96d035b12d 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.h
+++ b/src/gallium/drivers/svga/svga_screen_texture.h
@@ -78,7 +78,7 @@ struct svga_texture
 {
    struct pipe_texture base;
 
-   boolean defined[6][PIPE_MAX_TEXTURE_LEVELS];
+   boolean defined[6][SVGA_MAX_TEXTURE_LEVELS];
    
    struct svga_sampler_view *cached_view;
 
@@ -186,6 +186,9 @@ svga_surface_needs_propagation(struct pipe_surface *surf);
 extern void
 svga_screen_init_texture_functions(struct pipe_screen *screen);
 
+void
+svga_init_texture_functions(struct pipe_context *pipe);
+
 enum SVGA3dSurfaceFormat
 svga_translate_format(enum pipe_format format);
 
diff --git a/src/gallium/drivers/svga/svga_state_framebuffer.c b/src/gallium/drivers/svga/svga_state_framebuffer.c
index b4cafb8f21..b710914acd 100644
--- a/src/gallium/drivers/svga/svga_state_framebuffer.c
+++ b/src/gallium/drivers/svga/svga_state_framebuffer.c
@@ -70,7 +70,7 @@ static int emit_framebuffer( struct svga_context *svga,
          return ret;
 
       if (curr->zsbuf &&
-          curr->zsbuf->format == PIPE_FORMAT_Z24S8_UNORM) {
+          curr->zsbuf->format == PIPE_FORMAT_S8Z24_UNORM) {
          ret = SVGA3D_SetRenderTarget(svga->swc, SVGA3D_RT_STENCIL, curr->zsbuf);
          if (ret != PIPE_OK)
             return ret;
@@ -120,174 +120,153 @@ static int emit_viewport( struct svga_context *svga,
    float fb_width = svga->curr.framebuffer.width;
    float fb_height = svga->curr.framebuffer.height;
 
-   memset( &prescale, 0, sizeof(prescale) );
-
-   if (svga->curr.rast->templ.bypass_vs_clip_and_viewport) {
-
-      /* Avoid POSITIONT as it has a non trivial implementation outside the D3D
-       * API. Always generate a vertex shader.
-       */
-      rect.x = 0;
-      rect.y = 0;
-      rect.w = svga->curr.framebuffer.width;
-      rect.h = svga->curr.framebuffer.height;
-
-      prescale.scale[0] = 2.0 / (float)rect.w;
-      prescale.scale[1] = - 2.0 / (float)rect.h;
-      prescale.scale[2] = 1.0;
-      prescale.scale[3] = 1.0;
-      prescale.translate[0] = -1.0f;
-      prescale.translate[1] = 1.0f;
-      prescale.translate[2] = 0;
-      prescale.translate[3] = 0;
-      prescale.enabled = TRUE;
-   } else {
-
-      /* Examine gallium viewport transformation and produce a screen
-       * rectangle and possibly vertex shader pre-transformation to
-       * get the same results.
-       */
-      float fx =        viewport->scale[0] * -1.0 + viewport->translate[0];
-      float fy = flip * viewport->scale[1] * -1.0 + viewport->translate[1];
-      float fw =        viewport->scale[0] * 2; 
-      float fh = flip * viewport->scale[1] * 2; 
-
-      SVGA_DBG(DEBUG_VIEWPORT,
-               "\ninitial %f,%f %fx%f\n",
-               fx,
-               fy,
-               fw,
-               fh);
-
-      prescale.scale[0] = 1.0;
-      prescale.scale[1] = 1.0;
-      prescale.scale[2] = 1.0;
-      prescale.scale[3] = 1.0;
-      prescale.translate[0] = 0;
-      prescale.translate[1] = 0;
-      prescale.translate[2] = 0;
-      prescale.translate[3] = 0;
-      prescale.enabled = TRUE;
-
-
-
-      if (fw < 0) {
-         prescale.scale[0] *= -1.0;
-         prescale.translate[0] += -fw;
-         fw = -fw;
-         fx =        viewport->scale[0] * 1.0 + viewport->translate[0];
-      }
+   float fx =        viewport->scale[0] * -1.0 + viewport->translate[0];
+   float fy = flip * viewport->scale[1] * -1.0 + viewport->translate[1];
+   float fw =        viewport->scale[0] * 2; 
+   float fh = flip * viewport->scale[1] * 2; 
 
-      if (fh < 0) {
-         prescale.scale[1] *= -1.0;
-         prescale.translate[1] += -fh;
-         fh = -fh;
-         fy = flip * viewport->scale[1] * 1.0 + viewport->translate[1];
-      }
+   memset( &prescale, 0, sizeof(prescale) );
 
-      if (fx < 0) {
-         prescale.translate[0] += fx;
-         prescale.scale[0] *= fw / (fw + fx); 
-         fw += fx;
-         fx = 0;
-      }
+   /* Examine gallium viewport transformation and produce a screen
+    * rectangle and possibly vertex shader pre-transformation to
+    * get the same results.
+    */
 
-      if (fy < 0) {
-         prescale.translate[1] += fy;
-         prescale.scale[1] *= fh / (fh + fy); 
-         fh += fy;
-         fy = 0;
-      }
+   SVGA_DBG(DEBUG_VIEWPORT,
+            "\ninitial %f,%f %fx%f\n",
+            fx,
+            fy,
+            fw,
+            fh);
+
+   prescale.scale[0] = 1.0;
+   prescale.scale[1] = 1.0;
+   prescale.scale[2] = 1.0;
+   prescale.scale[3] = 1.0;
+   prescale.translate[0] = 0;
+   prescale.translate[1] = 0;
+   prescale.translate[2] = 0;
+   prescale.translate[3] = 0;
+   prescale.enabled = TRUE;
+
+
+
+   if (fw < 0) {
+      prescale.scale[0] *= -1.0;
+      prescale.translate[0] += -fw;
+      fw = -fw;
+      fx =        viewport->scale[0] * 1.0 + viewport->translate[0];
+   }
 
-      if (fx + fw > fb_width) {
-         prescale.scale[0] *= fw / (fb_width - fx); 
-         prescale.translate[0] -= fx * (fw / (fb_width - fx));
-         prescale.translate[0] += fx;
-         fw = fb_width - fx;
-         
-      }
+   if (fh < 0) {
+      prescale.scale[1] *= -1.0;
+      prescale.translate[1] += -fh;
+      fh = -fh;
+      fy = flip * viewport->scale[1] * 1.0 + viewport->translate[1];
+   }
 
-      if (fy + fh > fb_height) {
-         prescale.scale[1] *= fh / (fb_height - fy);
-         prescale.translate[1] -= fy * (fh / (fb_height - fy));
-         prescale.translate[1] += fy;
-         fh = fb_height - fy;
-      }
+   if (fx < 0) {
+      prescale.translate[0] += fx;
+      prescale.scale[0] *= fw / (fw + fx); 
+      fw += fx;
+      fx = 0;
+   }
 
-      if (fw < 0 || fh < 0) {
-         fw = fh = fx = fy = 0;
-         degenerate = TRUE;
-         goto out;
-      }
+   if (fy < 0) {
+      prescale.translate[1] += fy;
+      prescale.scale[1] *= fh / (fh + fy); 
+      fh += fy;
+      fy = 0;
+   }
 
+   if (fx + fw > fb_width) {
+      prescale.scale[0] *= fw / (fb_width - fx); 
+      prescale.translate[0] -= fx * (fw / (fb_width - fx));
+      prescale.translate[0] += fx;
+      fw = fb_width - fx;
+      
+   }
 
-      /* D3D viewport is integer space.  Convert fx,fy,etc. to
-       * integers.
-       *
-       * TODO: adjust pretranslate correct for any subpixel error
-       * introduced converting to integers.
-       */
-      rect.x = fx;
-      rect.y = fy;
-      rect.w = fw;
-      rect.h = fh;
+   if (fy + fh > fb_height) {
+      prescale.scale[1] *= fh / (fb_height - fy);
+      prescale.translate[1] -= fy * (fh / (fb_height - fy));
+      prescale.translate[1] += fy;
+      fh = fb_height - fy;
+   }
 
-      SVGA_DBG(DEBUG_VIEWPORT,
-               "viewport error %f,%f %fx%f\n",
-               fabs((float)rect.x - fx),
-               fabs((float)rect.y - fy),
-               fabs((float)rect.w - fw),
-               fabs((float)rect.h - fh));
+   if (fw < 0 || fh < 0) {
+      fw = fh = fx = fy = 0;
+      degenerate = TRUE;
+      goto out;
+   }
 
-      SVGA_DBG(DEBUG_VIEWPORT,
-               "viewport %d,%d %dx%d\n",
-               rect.x,
-               rect.y,
-               rect.w,
-               rect.h);
-      
 
-      /* Finally, to get GL rasterization rules, need to tweak the
-       * screen-space coordinates slightly relative to D3D which is
-       * what hardware implements natively.
-       */
-      if (svga->curr.rast->templ.gl_rasterization_rules) {
-         float adjust_x = 0.0;
-         float adjust_y = 0.0;
-
-         switch (svga->curr.reduced_prim) {
-         case PIPE_PRIM_LINES:
-            adjust_x = -0.5;
-            adjust_y = 0;
-            break;
-         case PIPE_PRIM_POINTS:
-         case PIPE_PRIM_TRIANGLES:
-            adjust_x = -0.375;
-            adjust_y = -0.5;
-            break;
-         }
-
-         prescale.translate[0] += adjust_x;
-         prescale.translate[1] += adjust_y;
-         prescale.translate[2] = 0.5; /* D3D clip space */
-         prescale.scale[2]     = 0.5; /* D3D clip space */
+   /* D3D viewport is integer space.  Convert fx,fy,etc. to
+    * integers.
+    *
+    * TODO: adjust pretranslate correct for any subpixel error
+    * introduced converting to integers.
+    */
+   rect.x = fx;
+   rect.y = fy;
+   rect.w = fw;
+   rect.h = fh;
+
+   SVGA_DBG(DEBUG_VIEWPORT,
+            "viewport error %f,%f %fx%f\n",
+            fabs((float)rect.x - fx),
+            fabs((float)rect.y - fy),
+            fabs((float)rect.w - fw),
+            fabs((float)rect.h - fh));
+
+   SVGA_DBG(DEBUG_VIEWPORT,
+            "viewport %d,%d %dx%d\n",
+            rect.x,
+            rect.y,
+            rect.w,
+            rect.h);
+
+
+   /* Finally, to get GL rasterization rules, need to tweak the
+    * screen-space coordinates slightly relative to D3D which is
+    * what hardware implements natively.
+    */
+   if (svga->curr.rast->templ.gl_rasterization_rules) {
+      float adjust_x = 0.0;
+      float adjust_y = 0.0;
+
+      switch (svga->curr.reduced_prim) {
+      case PIPE_PRIM_LINES:
+         adjust_x = -0.5;
+         adjust_y = 0;
+         break;
+      case PIPE_PRIM_POINTS:
+      case PIPE_PRIM_TRIANGLES:
+         adjust_x = -0.375;
+         adjust_y = -0.5;
+         break;
       }
 
+      prescale.translate[0] += adjust_x;
+      prescale.translate[1] += adjust_y;
+      prescale.translate[2] = 0.5; /* D3D clip space */
+      prescale.scale[2]     = 0.5; /* D3D clip space */
+   }
 
-      range_min = viewport->scale[2] * -1.0 + viewport->translate[2];
-      range_max = viewport->scale[2] *  1.0 + viewport->translate[2];
 
-      /* D3D (and by implication SVGA) doesn't like dealing with zmax
-       * less than zmin.  Detect that case, flip the depth range and
-       * invert our z-scale factor to achieve the same effect.
-       */
-      if (range_min > range_max) {
-         float range_tmp;
-         range_tmp = range_min; 
-         range_min = range_max; 
-         range_max = range_tmp;
-         prescale.scale[2]     = -prescale.scale[2];
-      }
+   range_min = viewport->scale[2] * -1.0 + viewport->translate[2];
+   range_max = viewport->scale[2] *  1.0 + viewport->translate[2];
+
+   /* D3D (and by implication SVGA) doesn't like dealing with zmax
+    * less than zmin.  Detect that case, flip the depth range and
+    * invert our z-scale factor to achieve the same effect.
+    */
+   if (range_min > range_max) {
+      float range_tmp;
+      range_tmp = range_min; 
+      range_min = range_max; 
+      range_max = range_tmp;
+      prescale.scale[2]     = -prescale.scale[2];
    }
 
    if (prescale.enabled) {
diff --git a/src/gallium/drivers/svga/svga_state_need_swtnl.c b/src/gallium/drivers/svga/svga_state_need_swtnl.c
index dd13a89d24..dfaab53aef 100644
--- a/src/gallium/drivers/svga/svga_state_need_swtnl.c
+++ b/src/gallium/drivers/svga/svga_state_need_swtnl.c
@@ -43,7 +43,7 @@ svga_translate_vertex_format(enum pipe_format format)
    case PIPE_FORMAT_R32G32_FLOAT:         return SVGA3D_DECLTYPE_FLOAT2;
    case PIPE_FORMAT_R32G32B32_FLOAT:      return SVGA3D_DECLTYPE_FLOAT3;
    case PIPE_FORMAT_R32G32B32A32_FLOAT:   return SVGA3D_DECLTYPE_FLOAT4;
-   case PIPE_FORMAT_B8G8R8A8_UNORM:       return SVGA3D_DECLTYPE_D3DCOLOR;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:       return SVGA3D_DECLTYPE_D3DCOLOR;
    case PIPE_FORMAT_R8G8B8A8_USCALED:     return SVGA3D_DECLTYPE_UBYTE4;
    case PIPE_FORMAT_R16G16_SSCALED:       return SVGA3D_DECLTYPE_SHORT2;
    case PIPE_FORMAT_R16G16B16A16_SSCALED: return SVGA3D_DECLTYPE_SHORT4;
@@ -76,8 +76,13 @@ static int update_need_swvfetch( struct svga_context *svga,
    unsigned i;
    boolean need_swvfetch = FALSE;
 
-   for (i = 0; i < svga->curr.num_vertex_elements; i++) {
-      svga->state.sw.ve_format[i] = svga_translate_vertex_format(svga->curr.ve[i].src_format);
+   if (!svga->curr.velems) {
+      /* No vertex elements bound. */
+      return 0;
+   }
+
+   for (i = 0; i < svga->curr.velems->count; i++) {
+      svga->state.sw.ve_format[i] = svga_translate_vertex_format(svga->curr.velems->velem[i].src_format);
       if (svga->state.sw.ve_format[i] == SVGA3D_DECLTYPE_MAX) {
          need_swvfetch = TRUE;
          break;
@@ -129,8 +134,7 @@ static int update_need_pipeline( struct svga_context *svga,
 
    /* SVGA_NEW_CLIP 
     */
-   if (!svga->curr.rast->templ.bypass_vs_clip_and_viewport &&
-       svga->curr.clip.nr) {
+   if (svga->curr.clip.nr) {
       SVGA_DBG(DEBUG_SWTNL, "%s: userclip\n", __FUNCTION__);
       need_pipeline = TRUE;
    }
diff --git a/src/gallium/drivers/svga/svga_state_rss.c b/src/gallium/drivers/svga/svga_state_rss.c
index 107cc403b4..b7195d246b 100644
--- a/src/gallium/drivers/svga/svga_state_rss.c
+++ b/src/gallium/drivers/svga/svga_state_rss.c
@@ -191,15 +191,24 @@ static int emit_rss( struct svga_context *svga,
       EMIT_RS( svga, svga->curr.stencil_ref.ref_value[0], STENCILREF, fail );
    }
 
-   if (dirty & SVGA_NEW_RAST)
+   if (dirty & (SVGA_NEW_RAST | SVGA_NEW_NEED_PIPELINE))
    {
       const struct svga_rasterizer_state *curr = svga->curr.rast; 
+      unsigned cullmode = curr->cullmode;
 
       /* Shademode: still need to rearrange index list to move
        * flat-shading PV first vertex.
        */
       EMIT_RS( svga, curr->shademode, SHADEMODE, fail );
-      EMIT_RS( svga, curr->cullmode, CULLMODE, fail );
+
+      /* Don't do culling while the software pipeline is active.  It
+       * does it for us, and additionally introduces potentially
+       * back-facing triangles.
+       */
+      if (svga->state.sw.need_pipeline)
+         cullmode = SVGA3D_FACE_NONE;
+
+      EMIT_RS( svga, cullmode, CULLMODE, fail );
       EMIT_RS( svga, curr->scissortestenable, SCISSORTESTENABLE, fail );
       EMIT_RS( svga, curr->multisampleantialias, MULTISAMPLEANTIALIAS, fail );
       EMIT_RS( svga, curr->lastpixel, LASTPIXEL, fail );
diff --git a/src/gallium/drivers/svga/svga_state_vdecl.c b/src/gallium/drivers/svga/svga_state_vdecl.c
index ded903170b..f531e22304 100644
--- a/src/gallium/drivers/svga/svga_state_vdecl.c
+++ b/src/gallium/drivers/svga/svga_state_vdecl.c
@@ -95,17 +95,17 @@ upload_user_buffers( struct svga_context *svga )
 static int emit_hw_vs_vdecl( struct svga_context *svga,
                              unsigned dirty )
 {
-   const struct pipe_vertex_element *ve = svga->curr.ve;
+   const struct pipe_vertex_element *ve = svga->curr.velems->velem;
    SVGA3dVertexDecl decl;
    unsigned i;
 
-   assert(svga->curr.num_vertex_elements >=
+   assert(svga->curr.velems->count >=
           svga->curr.vs->base.info.file_count[TGSI_FILE_INPUT]);
 
    svga_hwtnl_reset_vdecl( svga->hwtnl, 
-                           svga->curr.num_vertex_elements );
+                           svga->curr.velems->count );
 
-   for (i = 0; i < svga->curr.num_vertex_elements; i++) {
+   for (i = 0; i < svga->curr.velems->count; i++) {
       const struct pipe_vertex_buffer *vb = &svga->curr.vb[ve[i].vertex_buffer_index];
       unsigned usage, index;
 
diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c
index d7999fe53d..781f7bf533 100644
--- a/src/gallium/drivers/svga/svga_state_vs.c
+++ b/src/gallium/drivers/svga/svga_state_vs.c
@@ -186,8 +186,8 @@ static int update_zero_stride( struct svga_context *svga,
    svga->curr.zero_stride_vertex_elements = 0;
    svga->curr.num_zero_stride_vertex_elements = 0;
 
-   for (i = 0; i < svga->curr.num_vertex_elements; i++) {
-      const struct pipe_vertex_element *vel = &svga->curr.ve[i];
+   for (i = 0; i < svga->curr.velems->count; i++) {
+      const struct pipe_vertex_element *vel = &svga->curr.velems->velem[i];
       const struct pipe_vertex_buffer *vbuffer = &svga->curr.vb[
          vel->vertex_buffer_index];
       if (vbuffer->stride == 0) {
diff --git a/src/gallium/drivers/svga/svga_swtnl_state.c b/src/gallium/drivers/svga/svga_swtnl_state.c
index 35f36a828f..246d34e649 100644
--- a/src/gallium/drivers/svga/svga_swtnl_state.c
+++ b/src/gallium/drivers/svga/svga_swtnl_state.c
@@ -99,8 +99,8 @@ static int update_swtnl_draw( struct svga_context *svga,
 
    if (dirty & SVGA_NEW_VELEMENT)
       draw_set_vertex_elements(svga->swtnl.draw, 
-                               svga->curr.num_vertex_elements, 
-                               svga->curr.ve );
+                               svga->curr.velems->count, 
+                               svga->curr.velems->velem );
 
    if (dirty & SVGA_NEW_CLIP)
       draw_set_clip_state(svga->swtnl.draw, 
diff --git a/src/gallium/drivers/svga/svga_winsys.h b/src/gallium/drivers/svga/svga_winsys.h
index b4e3af0eaf..d4bb176f9a 100644
--- a/src/gallium/drivers/svga/svga_winsys.h
+++ b/src/gallium/drivers/svga/svga_winsys.h
@@ -51,6 +51,7 @@ struct pipe_context;
 struct pipe_fence_handle;
 struct pipe_texture;
 struct svga_region;
+struct winsys_handle;
 
 
 #define SVGA_BUFFER_USAGE_PINNED  (PIPE_BUFFER_USAGE_CUSTOM << 0)
@@ -187,6 +188,25 @@ struct svga_winsys_screen
                      uint32 numMipLevels);
 
    /**
+    * Creates a surface from a winsys handle.
+    * Used to implement pipe_screen::texture_from_handle.
+    */
+   struct svga_winsys_surface *
+   (*surface_from_handle)(struct svga_winsys_screen *sws,
+                          struct winsys_handle *whandle,
+                          SVGA3dSurfaceFormat *format);
+
+   /**
+    * Get a winsys_handle from a surface.
+    * Used to implement pipe_screen::texture_get_handle.
+    */
+   boolean
+   (*surface_get_handle)(struct svga_winsys_screen *sws,
+                         struct svga_winsys_surface *surface,
+                         unsigned stride,
+                         struct winsys_handle *whandle);
+
+   /**
     * Whether this surface is sitting in a validate list
     */
    boolean
@@ -284,19 +304,6 @@ svga_screen_buffer_wrap_surface(struct pipe_screen *screen,
 				struct svga_winsys_surface *srf);
 
 struct svga_winsys_surface *
-svga_screen_texture_get_winsys_surface(struct pipe_texture *texture);
-struct svga_winsys_surface *
 svga_screen_buffer_get_winsys_surface(struct pipe_buffer *buffer);
 
-boolean
-svga_screen_buffer_from_texture(struct pipe_texture *texture,
-				struct pipe_buffer **buffer,
-				unsigned *stride);
-
-struct pipe_texture *
-svga_screen_texture_wrap_surface(struct pipe_screen *screen,
-				 struct pipe_texture *base,
-				 enum SVGA3dSurfaceFormat format,
-				 struct svga_winsys_surface *srf);
-
 #endif /* SVGA_WINSYS_H_ */
diff --git a/src/gallium/drivers/svga/svgadump/svga_shader_dump.c b/src/gallium/drivers/svga/svgadump/svga_shader_dump.c
index 705ca29e8f..4ee1bf2c35 100644
--- a/src/gallium/drivers/svga/svgadump/svga_shader_dump.c
+++ b/src/gallium/drivers/svga/svgadump/svga_shader_dump.c
@@ -360,7 +360,9 @@ dump_dstreg(struct sh_dstreg dstreg,
    union {
       struct sh_reg reg;
       struct sh_dstreg dstreg;
-   } u = { { 0 } };
+   } u;
+
+   memset(&u, 0, sizeof(u));
 
    assert( (dstreg.modifier & (SVGA3DDSTMOD_SATURATE | SVGA3DDSTMOD_PARTIALPRECISION)) == dstreg.modifier );