svga: Performance fixes
[mesa.git] / src / gallium / drivers / svga / svga_surface.c
index 3e4bed76c052295dd00d20d25e23f4015c2cf67b..68edf1c21db55af445f50dc2964f02da2b65fd25 100644 (file)
 #include "pipe/p_defines.h"
 #include "util/u_inlines.h"
 #include "os/os_thread.h"
-#include "util/u_format.h"
+#include "util/u_bitmask.h"
+#include "util/format/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
+#include "svga_format.h"
 #include "svga_screen.h"
 #include "svga_context.h"
+#include "svga_sampler_view.h"
 #include "svga_resource_texture.h"
 #include "svga_surface.h"
 #include "svga_debug.h"
 
+static void svga_mark_surface_dirty(struct pipe_surface *surf);
+
+void
+svga_texture_copy_region(struct svga_context *svga,
+                         struct svga_winsys_surface *src_handle,
+                         unsigned srcSubResource,
+                         unsigned src_x, unsigned src_y, unsigned src_z,
+                         struct svga_winsys_surface *dst_handle,
+                         unsigned dstSubResource,
+                         unsigned dst_x, unsigned dst_y, unsigned dst_z,
+                         unsigned width, unsigned height, unsigned depth)
+{
+   SVGA3dCopyBox box;
+
+   assert(svga_have_vgpu10(svga));
+
+   box.x = dst_x;
+   box.y = dst_y;
+   box.z = dst_z;
+   box.w = width;
+   box.h = height;
+   box.d = depth;
+   box.srcx = src_x;
+   box.srcy = src_y;
+   box.srcz = src_z;
+
+   SVGA_RETRY(svga, SVGA3D_vgpu10_PredCopyRegion
+              (svga->swc, dst_handle, dstSubResource,
+               src_handle, srcSubResource, &box));
+}
+
 
 void
 svga_texture_copy_handle(struct svga_context *svga,
                          struct svga_winsys_surface *src_handle,
                          unsigned src_x, unsigned src_y, unsigned src_z,
-                         unsigned src_level, unsigned src_face,
+                         unsigned src_level, unsigned src_layer,
                          struct svga_winsys_surface *dst_handle,
                          unsigned dst_x, unsigned dst_y, unsigned dst_z,
-                         unsigned dst_level, unsigned dst_face,
+                         unsigned dst_level, unsigned dst_layer,
                          unsigned width, unsigned height, unsigned depth)
 {
    struct svga_surface dst, src;
-   enum pipe_error ret;
    SVGA3dCopyBox box, *boxes;
 
    assert(svga);
 
    src.handle = src_handle;
    src.real_level = src_level;
-   src.real_face = src_face;
+   src.real_layer = src_layer;
    src.real_zslice = 0;
 
    dst.handle = dst_handle;
    dst.real_level = dst_level;
-   dst.real_face = dst_face;
+   dst.real_layer = dst_layer;
    dst.real_zslice = 0;
 
    box.x = dst_x;
@@ -82,129 +115,207 @@ svga_texture_copy_handle(struct svga_context *svga,
             dst_handle, dst_level, dst_x, dst_y, dst_z);
 */
 
-   ret = SVGA3D_BeginSurfaceCopy(svga->swc,
-                                 &src.base,
-                                 &dst.base,
-                                 &boxes, 1);
-   if(ret != PIPE_OK) {
-      svga_context_flush(svga, NULL);
-      ret = SVGA3D_BeginSurfaceCopy(svga->swc,
-                                    &src.base,
-                                    &dst.base,
-                                    &boxes, 1);
-      assert(ret == PIPE_OK);
-   }
+   SVGA_RETRY(svga, SVGA3D_BeginSurfaceCopy(svga->swc,
+                                            &src.base,
+                                            &dst.base,
+                                            &boxes, 1));
+
    *boxes = box;
    SVGA_FIFOCommitAll(svga->swc);
 }
 
 
+/* A helper function to sync up the two surface handles.
+ */
+static void
+svga_texture_copy_handle_resource(struct svga_context *svga,
+                                  struct svga_texture *src_tex,
+                                  struct svga_winsys_surface *dst,
+                                  unsigned int numMipLevels,
+                                  unsigned int numLayers,
+                                  int zslice_pick,
+                                  unsigned int mipoffset,
+                                  unsigned int layeroffset)
+{
+   unsigned int i, j;
+   unsigned int zoffset = 0;
+
+   /* A negative zslice_pick implies zoffset at 0, and depth to copy is
+    * from the depth of the texture at the particular mipmap level.
+    */
+   if (zslice_pick >= 0)
+      zoffset = zslice_pick;
+
+   for (i = 0; i < numMipLevels; i++) {
+      unsigned int miplevel = i + mipoffset;
+
+      for (j = 0; j < numLayers; j++) {
+         if (svga_is_texture_level_defined(src_tex, j+layeroffset, miplevel)) {
+            unsigned depth = (zslice_pick < 0 ?
+                              u_minify(src_tex->b.b.depth0, miplevel) : 1);
+
+            if (src_tex->b.b.nr_samples > 1) {
+               unsigned subResource = j * numMipLevels + i;
+               svga_texture_copy_region(svga, src_tex->handle,
+                                        subResource, 0, 0, zoffset,
+                                        dst, subResource, 0, 0, 0,
+                                        src_tex->b.b.width0, src_tex->b.b.height0, depth);
+            }
+            else {
+               svga_texture_copy_handle(svga,
+                                        src_tex->handle,
+                                        0, 0, zoffset,
+                                        miplevel,
+                                        j + layeroffset,
+                                        dst, 0, 0, 0, i, j,
+                                        u_minify(src_tex->b.b.width0, miplevel),
+                                        u_minify(src_tex->b.b.height0, miplevel),
+                                        depth);
+            }
+         }
+      }
+   }
+}
+
+
 struct svga_winsys_surface *
-svga_texture_view_surface(struct pipe_context *pipe,
+svga_texture_view_surface(struct svga_context *svga,
                           struct svga_texture *tex,
+                          unsigned bind_flags,
+                          SVGA3dSurfaceAllFlags flags,
                           SVGA3dSurfaceFormat format,
                           unsigned start_mip,
                           unsigned num_mip,
-                          int face_pick,
+                          int layer_pick,
+                          unsigned num_layers,
                           int zslice_pick,
+                          boolean cacheable,
                           struct svga_host_surface_cache_key *key) /* OUT */
 {
-   struct svga_screen *ss = svga_screen(pipe->screen);
-   struct svga_winsys_surface *handle;
-   uint32_t i, j;
-   unsigned z_offset = 0;
+   struct svga_screen *ss = svga_screen(svga->pipe.screen);
+   struct svga_winsys_surface *handle = NULL;
+   boolean validated;
+   boolean needCopyResource;
 
-   SVGA_DBG(DEBUG_PERF, 
-            "svga: Create surface view: face %d zslice %d mips %d..%d\n",
-            face_pick, zslice_pick, start_mip, start_mip+num_mip-1);
+   SVGA_DBG(DEBUG_PERF,
+            "svga: Create surface view: layer %d zslice %d mips %d..%d\n",
+            layer_pick, zslice_pick, start_mip, start_mip+num_mip-1);
 
-   key->flags = 0;
+   SVGA_STATS_TIME_PUSH(ss->sws, SVGA_STATS_TIME_EMULATESURFACEVIEW);
+
+   key->flags = flags;
    key->format = format;
    key->numMipLevels = num_mip;
    key->size.width = u_minify(tex->b.b.width0, start_mip);
    key->size.height = u_minify(tex->b.b.height0, start_mip);
    key->size.depth = zslice_pick < 0 ? u_minify(tex->b.b.depth0, start_mip) : 1;
    key->cachable = 1;
-   assert(key->size.depth == 1);
-   
-   if(tex->b.b.target == PIPE_TEXTURE_CUBE && face_pick < 0) {
+   key->arraySize = 1;
+   key->numFaces = 1;
+
+   /* single sample surface can be treated as non-multisamples surface */
+   key->sampleCount = tex->b.b.nr_samples > 1 ? tex->b.b.nr_samples : 0;
+
+   if (key->sampleCount > 1) {
+      assert(ss->sws->have_sm4_1);
+      key->flags |= SVGA3D_SURFACE_MULTISAMPLE;
+   }
+
+   if (tex->b.b.target == PIPE_TEXTURE_CUBE && layer_pick < 0) {
       key->flags |= SVGA3D_SURFACE_CUBEMAP;
       key->numFaces = 6;
-   } else {
-      key->numFaces = 1;
+   } else if (tex->b.b.target == PIPE_TEXTURE_1D_ARRAY ||
+              tex->b.b.target == PIPE_TEXTURE_2D_ARRAY) {
+      key->arraySize = num_layers;
    }
 
-   if(key->format == SVGA3D_FORMAT_INVALID) {
+   if (key->format == SVGA3D_FORMAT_INVALID) {
       key->cachable = 0;
-      return NULL;
+      goto done;
+   }
+
+   if (cacheable && tex->backed_handle &&
+       memcmp(key, &tex->backed_key, sizeof *key) == 0) {
+      handle = tex->backed_handle;
+      needCopyResource = tex->backed_age < tex->age;
+   } else {
+      SVGA_DBG(DEBUG_DMA, "surface_create for texture view\n");
+      handle = svga_screen_surface_create(ss, bind_flags, PIPE_USAGE_DEFAULT,
+                                          &validated, key);
+      needCopyResource = TRUE;
+
+      if (cacheable && !tex->backed_handle) {
+         tex->backed_handle = handle;
+         memcpy(&tex->backed_key, key, sizeof *key);
+      }
    }
 
-   SVGA_DBG(DEBUG_DMA, "surface_create for texture view\n");
-   handle = svga_screen_surface_create(ss, key);
    if (!handle) {
       key->cachable = 0;
-      return NULL;
+      goto done;
    }
 
    SVGA_DBG(DEBUG_DMA, " --> got sid %p (texture view)\n", handle);
 
-   if (face_pick < 0)
-      face_pick = 0;
-
-   if (zslice_pick >= 0)
-       z_offset = zslice_pick;
+   if (layer_pick < 0)
+      layer_pick = 0;
 
-   for (i = 0; i < key->numMipLevels; i++) {
-      for (j = 0; j < key->numFaces; j++) {
-         if(tex->defined[j + face_pick][i + start_mip]) {
-            unsigned depth = (zslice_pick < 0 ?
-                              u_minify(tex->b.b.depth0, i + start_mip) :
-                              1);
-
-            svga_texture_copy_handle(svga_context(pipe),
-                                     tex->handle, 
-                                     0, 0, z_offset, 
-                                     i + start_mip, 
-                                     j + face_pick,
-                                     handle, 0, 0, 0, i, j,
-                                     u_minify(tex->b.b.width0, i + start_mip),
-                                     u_minify(tex->b.b.height0, i + start_mip),
-                                     depth);
-         }
-      }
+   if (needCopyResource) {
+      svga_texture_copy_handle_resource(svga, tex, handle,
+                                        key->numMipLevels,
+                                        key->numFaces * key->arraySize,
+                                        zslice_pick, start_mip, layer_pick);
+      tex->backed_age = tex->age;
    }
 
+done:
+   SVGA_STATS_TIME_POP(ss->sws);
+
    return handle;
 }
 
 
+/**
+ * A helper function to create a surface view.
+ * The view boolean flag specifies whether svga_texture_view_surface()
+ * will be called to create a cloned surface and resource for the view.
+ */
 static struct pipe_surface *
-svga_create_surface(struct pipe_context *pipe,
-                    struct pipe_resource *pt,
-                    const struct pipe_surface *surf_tmpl)
+svga_create_surface_view(struct pipe_context *pipe,
+                         struct pipe_resource *pt,
+                         const struct pipe_surface *surf_tmpl,
+                         boolean view)
 {
+   struct svga_context *svga = svga_context(pipe);
    struct svga_texture *tex = svga_texture(pt);
    struct pipe_screen *screen = pipe->screen;
+   struct svga_screen *ss = svga_screen(screen);
    struct svga_surface *s;
-   unsigned face, zslice;
-   /* XXX surfaces should only be used for rendering purposes nowadays */
-   boolean render = (surf_tmpl->usage & (PIPE_BIND_RENDER_TARGET |
-                                         PIPE_BIND_DEPTH_STENCIL)) ? TRUE : FALSE;
-   boolean view = FALSE;
+   unsigned layer, zslice, bind;
+   unsigned nlayers = 1;
+   SVGA3dSurfaceAllFlags flags = 0;
    SVGA3dSurfaceFormat format;
-
-   assert(surf_tmpl->u.tex.first_layer == surf_tmpl->u.tex.last_layer);
+   struct pipe_surface *retVal = NULL;
 
    s = CALLOC_STRUCT(svga_surface);
    if (!s)
       return NULL;
 
+   SVGA_STATS_TIME_PUSH(ss->sws, SVGA_STATS_TIME_CREATESURFACEVIEW);
+
    if (pt->target == PIPE_TEXTURE_CUBE) {
-         face = surf_tmpl->u.tex.first_layer;
-         zslice = 0;
+      layer = surf_tmpl->u.tex.first_layer;
+      zslice = 0;
+   }
+   else if (pt->target == PIPE_TEXTURE_1D_ARRAY ||
+            pt->target == PIPE_TEXTURE_2D_ARRAY ||
+            pt->target == PIPE_TEXTURE_CUBE_ARRAY) {
+      layer = surf_tmpl->u.tex.first_layer;
+      zslice = 0;
+      nlayers = surf_tmpl->u.tex.last_layer - surf_tmpl->u.tex.first_layer + 1;
    }
    else {
-      face = 0;
+      layer = 0;
       zslice = surf_tmpl->u.tex.first_layer;
    }
 
@@ -214,27 +325,124 @@ svga_create_surface(struct pipe_context *pipe,
    s->base.format = surf_tmpl->format;
    s->base.width = u_minify(pt->width0, surf_tmpl->u.tex.level);
    s->base.height = u_minify(pt->height0, surf_tmpl->u.tex.level);
-   s->base.usage = surf_tmpl->usage;
    s->base.u.tex.level = surf_tmpl->u.tex.level;
    s->base.u.tex.first_layer = surf_tmpl->u.tex.first_layer;
    s->base.u.tex.last_layer = surf_tmpl->u.tex.last_layer;
+   s->view_id = SVGA3D_INVALID_ID;
 
-   if (!render)
-      format = svga_translate_format(surf_tmpl->format);
-   else
-      format = svga_translate_format_render(surf_tmpl->format);
+   s->backed = NULL;
+
+   if (util_format_is_depth_or_stencil(surf_tmpl->format)) {
+      flags = SVGA3D_SURFACE_HINT_DEPTHSTENCIL |
+              SVGA3D_SURFACE_BIND_DEPTH_STENCIL;
+      bind = PIPE_BIND_DEPTH_STENCIL;
+   }
+   else {
+      flags = SVGA3D_SURFACE_HINT_RENDERTARGET |
+              SVGA3D_SURFACE_BIND_RENDER_TARGET;
+      bind = PIPE_BIND_RENDER_TARGET;
+   }
+
+   if (tex->imported) {
+      /* imported resource (a window) */
+      format = tex->key.format;
+      if (util_format_is_srgb(surf_tmpl->format)) {
+         /* sRGB rendering to window */
+         format = svga_linear_to_srgb(format);
+      }
+   }
+   else {
+      format = svga_translate_format(ss, surf_tmpl->format, bind);
+   }
 
    assert(format != SVGA3D_FORMAT_INVALID);
 
-   if (svga_screen(screen)->debug.force_surface_view)
-      view = TRUE;
+   if (view) {
+      SVGA_DBG(DEBUG_VIEWS,
+               "New backed surface view: resource %p, level %u layer %u z %u, %p\n",
+               pt, surf_tmpl->u.tex.level, layer, zslice, s);
+
+      if (svga_have_vgpu10(svga)) {
+         switch (pt->target) {
+         case PIPE_TEXTURE_1D:
+            flags |= SVGA3D_SURFACE_1D;
+            break;
+         case PIPE_TEXTURE_1D_ARRAY:
+            flags |= SVGA3D_SURFACE_1D | SVGA3D_SURFACE_ARRAY;
+            break;
+         case PIPE_TEXTURE_2D_ARRAY:
+            flags |= SVGA3D_SURFACE_ARRAY;
+            break;
+         case PIPE_TEXTURE_3D:
+            flags |= SVGA3D_SURFACE_VOLUME;
+            break;
+         case PIPE_TEXTURE_CUBE:
+            if (nlayers == 6)
+               flags |= SVGA3D_SURFACE_CUBEMAP;
+            break;
+         case PIPE_TEXTURE_CUBE_ARRAY:
+            if (nlayers % 6 == 0)
+               flags |= SVGA3D_SURFACE_CUBEMAP | SVGA3D_SURFACE_ARRAY;
+            break;   
+         default:
+            break;
+         }
+      }
 
-   /* Currently only used for compressed textures */
-   if (render && 
-       format != svga_translate_format(surf_tmpl->format)) {
-      view = TRUE;
+      /* When we clone the surface view resource, use the format used in
+       * the creation of the original resource.
+       */
+      s->handle = svga_texture_view_surface(svga, tex, bind, flags,
+                                            tex->key.format,
+                                            surf_tmpl->u.tex.level, 1,
+                                            layer, nlayers, zslice,
+                                            TRUE, &s->key);
+      if (!s->handle) {
+         FREE(s);
+         goto done;
+      }
+
+      s->key.format = format;
+      s->real_layer = 0;
+      s->real_level = 0;
+      s->real_zslice = 0;
+   } else {
+      SVGA_DBG(DEBUG_VIEWS,
+               "New surface view: resource %p, level %u, layer %u, z %u, %p\n",
+               pt, surf_tmpl->u.tex.level, layer, zslice, s);
+
+      memset(&s->key, 0, sizeof s->key);
+      s->key.format = format;
+      s->handle = tex->handle;
+      s->real_layer = layer;
+      s->real_zslice = zslice;
+      s->real_level = surf_tmpl->u.tex.level;
    }
 
+   svga->hud.num_surface_views++;
+   retVal = &s->base;
+
+done:
+   SVGA_STATS_TIME_POP(ss->sws);
+   return retVal;
+}
+
+
+static struct pipe_surface *
+svga_create_surface(struct pipe_context *pipe,
+                    struct pipe_resource *pt,
+                    const struct pipe_surface *surf_tmpl)
+{
+   struct svga_context *svga = svga_context(pipe);
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surf = NULL;
+   boolean view = FALSE;
+
+   SVGA_STATS_TIME_PUSH(svga_sws(svga), SVGA_STATS_TIME_CREATESURFACE);
+
+   if (svga_screen(screen)->debug.force_surface_view)
+      view = TRUE;
+
    if (surf_tmpl->u.tex.level != 0 &&
        svga_screen(screen)->debug.force_level_surface_view)
       view = TRUE;
@@ -242,85 +450,309 @@ svga_create_surface(struct pipe_context *pipe,
    if (pt->target == PIPE_TEXTURE_3D)
       view = TRUE;
 
-   if (svga_screen(screen)->debug.no_surface_view)
+   if (svga_have_vgpu10(svga) || svga_screen(screen)->debug.no_surface_view)
       view = FALSE;
 
-   if (view) {
-      SVGA_DBG(DEBUG_VIEWS, "svga: Surface view: yes %p, level %u face %u z %u, %p\n",
-               pt, surf_tmpl->u.tex.level, face, zslice, s);
+   surf = svga_create_surface_view(pipe, pt, surf_tmpl, view);
 
-      s->handle = svga_texture_view_surface(NULL, tex, format, surf_tmpl->u.tex.level,
-                                           1, face, zslice, &s->key);
-      s->real_face = 0;
-      s->real_level = 0;
-      s->real_zslice = 0;
-   } else {
-      SVGA_DBG(DEBUG_VIEWS, "svga: Surface view: no %p, level %u, face %u, z %u, %p\n",
-               pt, surf_tmpl->u.tex.level, face, zslice, s);
+   SVGA_STATS_TIME_POP(svga_sws(svga));
 
-      memset(&s->key, 0, sizeof s->key);
-      s->handle = tex->handle;
-      s->real_face = face;
-      s->real_zslice = zslice;
-      s->real_level = surf_tmpl->u.tex.level;
+   return surf;
+}
+
+
+/**
+ * Clone the surface view and its associated resource.
+ */
+static struct svga_surface *
+create_backed_surface_view(struct svga_context *svga, struct svga_surface *s)
+{
+   struct svga_texture *tex = svga_texture(s->base.texture);
+
+   if (!s->backed) {
+      struct pipe_surface *backed_view;
+
+      SVGA_STATS_TIME_PUSH(svga_sws(svga),
+                           SVGA_STATS_TIME_CREATEBACKEDSURFACEVIEW);
+
+      backed_view = svga_create_surface_view(&svga->pipe,
+                                             &tex->b.b,
+                                             &s->base,
+                                             TRUE);
+      if (!backed_view)
+         goto done;
+
+      s->backed = svga_surface(backed_view);
+
+      SVGA_STATS_TIME_POP(svga_sws(svga));
+   }
+   else if (s->backed->age < tex->age) {
+      /*
+       * There is already an existing backing surface, but we still need to
+       * sync the backing resource if the original resource has been modified
+       * since the last copy.
+       */
+      struct svga_surface *bs = s->backed;
+      unsigned int layer, zslice;
+
+      assert(bs->handle);
+
+      switch (tex->b.b.target) {
+      case PIPE_TEXTURE_CUBE:
+      case PIPE_TEXTURE_CUBE_ARRAY:
+      case PIPE_TEXTURE_1D_ARRAY:
+      case PIPE_TEXTURE_2D_ARRAY:
+         layer = s->base.u.tex.first_layer;
+         zslice = 0;
+         break;
+      default:
+         layer = 0;
+         zslice = s->base.u.tex.first_layer;
+      }
+
+      svga_texture_copy_handle_resource(svga, tex, bs->handle,
+                                        bs->key.numMipLevels,
+                                        bs->key.numFaces * bs->key.arraySize,
+                                        zslice, s->base.u.tex.level, layer);
    }
 
-   return &s->base;
+   svga_mark_surface_dirty(&s->backed->base);
+   s->backed->age = tex->age;
+
+done:
+   return s->backed;
 }
 
+/**
+ * Create a DX RenderTarget/DepthStencil View for the given surface,
+ * if needed.
+ */
+struct pipe_surface *
+svga_validate_surface_view(struct svga_context *svga, struct svga_surface *s)
+{
+   enum pipe_error ret = PIPE_OK;
+   enum pipe_shader_type shader;
+
+   assert(svga_have_vgpu10(svga));
+   assert(s);
+
+   SVGA_STATS_TIME_PUSH(svga_sws(svga),
+                        SVGA_STATS_TIME_VALIDATESURFACEVIEW);
+
+   /**
+    * DX spec explicitly specifies that no resource can be bound to a render
+    * target view and a shader resource view simultanously.
+    * So first check if the resource bound to this surface view collides with
+    * a sampler view. If so, then we will clone this surface view and its
+    * associated resource. We will then use the cloned surface view for
+    * render target.
+    */
+   for (shader = PIPE_SHADER_VERTEX; shader <= PIPE_SHADER_TESS_EVAL; shader++) {
+      if (svga_check_sampler_view_resource_collision(svga, s->handle, shader)) {
+         SVGA_DBG(DEBUG_VIEWS,
+                  "same resource used in shaderResource and renderTarget 0x%x\n",
+                  s->handle);
+         s = create_backed_surface_view(svga, s);
+
+         if (s)
+            svga->state.hw_draw.has_backed_views = TRUE;
+
+         /* s may be null here if the function failed */
+         break;
+      }
+   }
+
+   /**
+    * Create an alternate surface view for the specified context if the
+    * view was created for another context.
+    */
+   if (s && s->base.context != &svga->pipe) {
+      struct pipe_surface *surf;
+      surf = svga_create_surface_view(&svga->pipe, s->base.texture, &s->base, FALSE);
+      s = svga_surface(surf);
+   }
+
+   if (s && s->view_id == SVGA3D_INVALID_ID) {
+      SVGA3dResourceType resType;
+      SVGA3dRenderTargetViewDesc desc;
+      struct svga_texture *stex = svga_texture(s->base.texture);
+
+      if (stex->validated == FALSE) {
+         assert(stex->handle);
+
+         /* We are about to render into a surface that has not been validated.
+          * First invalidate the surface so that the device does not
+          * need to update the host-side copy with the invalid
+          * content when the associated mob is first bound to the surface.
+          */
+         SVGA_RETRY(svga, SVGA3D_InvalidateGBSurface(svga->swc, stex->handle));
+         stex->validated = TRUE;
+      }
+
+      desc.tex.mipSlice = s->real_level;
+      desc.tex.firstArraySlice = s->real_layer + s->real_zslice;
+      desc.tex.arraySize =
+         s->base.u.tex.last_layer - s->base.u.tex.first_layer + 1;
+
+      s->view_id = util_bitmask_add(svga->surface_view_id_bm);
+
+      resType = svga_resource_type(s->base.texture->target);
+
+      if (util_format_is_depth_or_stencil(s->base.format)) {
+         ret = SVGA3D_vgpu10_DefineDepthStencilView(svga->swc,
+                                                    s->view_id,
+                                                    s->handle,
+                                                    s->key.format,
+                                                    resType,
+                                                    &desc);
+      }
+      else {
+         SVGA3dSurfaceFormat view_format = s->key.format;
+         const struct svga_texture *stex = svga_texture(s->base.texture);
+
+         /* Can't create RGBA render target view of a RGBX surface so adjust
+          * the view format.  We do something similar for texture samplers in
+          * svga_validate_pipe_sampler_view().
+          */
+         if (view_format == SVGA3D_B8G8R8A8_UNORM &&
+             (stex->key.format == SVGA3D_B8G8R8X8_UNORM ||
+              stex->key.format == SVGA3D_B8G8R8X8_TYPELESS)) {
+            view_format = SVGA3D_B8G8R8X8_UNORM;
+         }
+
+         ret = SVGA3D_vgpu10_DefineRenderTargetView(svga->swc,
+                                                    s->view_id,
+                                                    s->handle,
+                                                    view_format,
+                                                    resType,
+                                                    &desc);
+      }
+
+      if (ret != PIPE_OK) {
+         util_bitmask_clear(svga->surface_view_id_bm, s->view_id);
+         s->view_id = SVGA3D_INVALID_ID;
+         s = NULL;
+      }
+   }
+   
+   SVGA_STATS_TIME_POP(svga_sws(svga));
+
+   return s ? &s->base : NULL;
+}
+
+
 
 static void
 svga_surface_destroy(struct pipe_context *pipe,
                      struct pipe_surface *surf)
 {
+   struct svga_context *svga = svga_context(pipe);
    struct svga_surface *s = svga_surface(surf);
    struct svga_texture *t = svga_texture(surf->texture);
    struct svga_screen *ss = svga_screen(surf->texture->screen);
 
-   if(s->handle != t->handle) {
+   SVGA_STATS_TIME_PUSH(ss->sws, SVGA_STATS_TIME_DESTROYSURFACE);
+
+   /* Destroy the backed view surface if it exists */
+   if (s->backed) {
+      svga_surface_destroy(pipe, &s->backed->base);
+      s->backed = NULL;
+   }
+
+   /* Destroy the surface handle if this is a backed handle and
+    * it is not being cached in the texture.
+    */
+   if (s->handle != t->handle && s->handle != t->backed_handle) {
       SVGA_DBG(DEBUG_DMA, "unref sid %p (tex surface)\n", s->handle);
       svga_screen_surface_destroy(ss, &s->key, &s->handle);
    }
 
+   if (s->view_id != SVGA3D_INVALID_ID) {
+      /* The SVGA3D device will generate a device error if the
+       * render target view or depth stencil view is destroyed from
+       * a context other than the one it was created with.
+       * Similar to shader resource view, in this case, we will skip
+       * the destroy for now.
+       */
+      if (surf->context != pipe) {
+         _debug_printf("context mismatch in %s\n", __func__);
+      }
+      else {
+         assert(svga_have_vgpu10(svga));
+         if (util_format_is_depth_or_stencil(s->base.format)) {
+            SVGA_RETRY(svga, SVGA3D_vgpu10_DestroyDepthStencilView(svga->swc,
+                                                                   s->view_id));
+         }
+         else {
+            SVGA_RETRY(svga, SVGA3D_vgpu10_DestroyRenderTargetView(svga->swc,
+                                                                   s->view_id));
+         }
+         util_bitmask_clear(svga->surface_view_id_bm, s->view_id);
+      }
+   }
+
    pipe_resource_reference(&surf->texture, NULL);
    FREE(surf);
+
+   svga->hud.num_surface_views--;
+   SVGA_STATS_TIME_POP(ss->sws);
 }
 
 
-static INLINE void 
+static void
 svga_mark_surface_dirty(struct pipe_surface *surf)
 {
    struct svga_surface *s = svga_surface(surf);
+   struct svga_texture *tex = svga_texture(surf->texture);
 
-   if(!s->dirty) {
-      struct svga_texture *tex = svga_texture(surf->texture);
-
+   if (!s->dirty) {
       s->dirty = TRUE;
 
       if (s->handle == tex->handle) {
          /* hmm so 3d textures always have all their slices marked ? */
-         if (surf->texture->target == PIPE_TEXTURE_CUBE)
-            tex->defined[surf->u.tex.first_layer][surf->u.tex.level] = TRUE;
-         else
-            tex->defined[0][surf->u.tex.level] = TRUE;
+         svga_define_texture_level(tex, surf->u.tex.first_layer,
+                                   surf->u.tex.level);
       }
       else {
          /* this will happen later in svga_propagate_surface */
       }
    }
+
+   /* Increment the view_age and texture age for this surface's mipmap
+    * level so that any sampler views into the texture are re-validated too.
+    * Note: we age the texture for backed surface view only when the
+    *       backed surface is propagated to the original surface.
+    */
+   if (s->handle == tex->handle)
+      svga_age_texture_view(tex, surf->u.tex.level);
 }
 
 
-void svga_mark_surfaces_dirty(struct svga_context *svga)
+void
+svga_mark_surfaces_dirty(struct svga_context *svga)
 {
    unsigned i;
+   struct svga_hw_clear_state *hw = &svga->state.hw_clear;
+
+   if (svga_have_vgpu10(svga)) {
 
-   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
-      if (svga->curr.framebuffer.cbufs[i])
-         svga_mark_surface_dirty(svga->curr.framebuffer.cbufs[i]);
+      /* For VGPU10, mark the dirty bit in the rendertarget/depth stencil view surface.
+       * This surface can be the backed surface.
+       */
+      for (i = 0; i < hw->num_rendertargets; i++) {
+         if (hw->rtv[i])
+            svga_mark_surface_dirty(hw->rtv[i]);
+      }
+      if (hw->dsv)
+         svga_mark_surface_dirty(hw->dsv);
+   } else {
+      for (i = 0; i < svga->curr.framebuffer.nr_cbufs; i++) {
+         if (svga->curr.framebuffer.cbufs[i])
+            svga_mark_surface_dirty(svga->curr.framebuffer.cbufs[i]);
+      }
+      if (svga->curr.framebuffer.zsbuf)
+         svga_mark_surface_dirty(svga->curr.framebuffer.zsbuf);
    }
-   if (svga->curr.framebuffer.zsbuf)
-      svga_mark_surface_dirty(svga->curr.framebuffer.zsbuf);
 }
 
 
@@ -329,55 +761,215 @@ void svga_mark_surfaces_dirty(struct svga_context *svga)
  * pipe is optional context to inline the blit command in.
  */
 void
-svga_propagate_surface(struct pipe_context *pipe, struct pipe_surface *surf)
+svga_propagate_surface(struct svga_context *svga, struct pipe_surface *surf,
+                       boolean reset)
 {
    struct svga_surface *s = svga_surface(surf);
    struct svga_texture *tex = svga_texture(surf->texture);
    struct svga_screen *ss = svga_screen(surf->texture->screen);
-   unsigned zslice, face;
 
    if (!s->dirty)
       return;
 
-   if (surf->texture->target == PIPE_TEXTURE_CUBE) {
-      zslice = 0;
-      face = surf->u.tex.first_layer;
-   }
-   else {
-      zslice = surf->u.tex.first_layer;
-      face = 0;
-   }
+   SVGA_STATS_TIME_PUSH(ss->sws, SVGA_STATS_TIME_PROPAGATESURFACE);
+
+   /* Reset the dirty flag if specified. This is to ensure that
+    * the dirty flag will not be reset and stay unset when the backing
+    * surface is still being bound and rendered to.
+    * The reset flag will be set to TRUE when the surface is propagated
+    * and will be unbound.
+    */
+   s->dirty = !reset;
 
-   s->dirty = FALSE;
    ss->texture_timestamp++;
-   tex->view_age[surf->u.tex.level] = ++(tex->age);
+   svga_age_texture_view(tex, surf->u.tex.level);
 
    if (s->handle != tex->handle) {
-      SVGA_DBG(DEBUG_VIEWS, "svga: Surface propagate: tex %p, level %u, from %p\n", tex, surf->u.tex.level, surf);
-      svga_texture_copy_handle(svga_context(pipe),
-                               s->handle, 0, 0, 0, s->real_level, s->real_face,
-                               tex->handle, 0, 0, zslice, surf->u.tex.level, face,
-                               u_minify(tex->b.b.width0, surf->u.tex.level),
-                               u_minify(tex->b.b.height0, surf->u.tex.level), 1);
-      tex->defined[face][surf->u.tex.level] = TRUE;
+      unsigned zslice, layer;
+      unsigned nlayers = 1;
+      unsigned i;
+      unsigned numMipLevels = tex->b.b.last_level + 1;
+      unsigned srcLevel = s->real_level;
+      unsigned dstLevel = surf->u.tex.level;
+      unsigned width = u_minify(tex->b.b.width0, dstLevel);
+      unsigned height = u_minify(tex->b.b.height0, dstLevel);
+
+      if (surf->texture->target == PIPE_TEXTURE_CUBE) {
+         zslice = 0;
+         layer = surf->u.tex.first_layer;
+      }
+      else if (surf->texture->target == PIPE_TEXTURE_1D_ARRAY ||
+               surf->texture->target == PIPE_TEXTURE_2D_ARRAY ||
+               surf->texture->target == PIPE_TEXTURE_CUBE_ARRAY) {
+         zslice = 0;
+         layer = surf->u.tex.first_layer;
+         nlayers = surf->u.tex.last_layer - surf->u.tex.first_layer + 1;
+      }
+      else {
+         zslice = surf->u.tex.first_layer;
+         layer = 0;
+      }
+
+      SVGA_DBG(DEBUG_VIEWS,
+               "Propagate surface %p to resource %p, level %u\n",
+               surf, tex, surf->u.tex.level);
+
+      if (svga_have_vgpu10(svga)) {
+         unsigned srcSubResource, dstSubResource;
+
+         for (i = 0; i < nlayers; i++) {
+            srcSubResource = (s->real_layer + i) * numMipLevels + srcLevel;
+            dstSubResource = (layer + i) * numMipLevels + dstLevel;
+
+            svga_texture_copy_region(svga,
+                                     s->handle, srcSubResource, 0, 0, 0,
+                                     tex->handle, dstSubResource, 0, 0, zslice,
+                                     width, height, 1);
+            svga_define_texture_level(tex, layer + i, dstLevel);
+         }
+      }
+      else {
+         for (i = 0; i < nlayers; i++) {
+            svga_texture_copy_handle(svga,
+                                     s->handle, 0, 0, 0, srcLevel,
+                                     s->real_layer + i,
+                                     tex->handle, 0, 0, zslice, dstLevel,
+                                     layer + i,
+                                     width, height, 1);
+        
+            svga_define_texture_level(tex, layer + i, dstLevel);
+         }
+      }
+
+      /* Sync the surface view age with the texture age */
+      s->age = tex->age;
+
+      /* If this backed surface is cached in the texture,
+       * update the backed age as well.
+       */
+      if (tex->backed_handle == s->handle) {
+         tex->backed_age = tex->age;
+      }
+   }
+
+   SVGA_STATS_TIME_POP(ss->sws);
+}
+
+
+/**
+ * If any of the render targets are in backing texture views, propagate any
+ * changes to them back to the original texture.
+ */
+void
+svga_propagate_rendertargets(struct svga_context *svga)
+{
+   unsigned i;
+
+   /* Early exit if there is no backing texture views in use */
+   if (!svga->state.hw_draw.has_backed_views)
+      return;
+
+   /* Note that we examine the svga->state.hw_draw.framebuffer surfaces,
+    * not the svga->curr.framebuffer surfaces, because it's the former
+    * surfaces which may be backing surface views (the actual render targets).
+    */
+   for (i = 0; i < svga->state.hw_clear.num_rendertargets; i++) {
+      struct pipe_surface *s = svga->state.hw_clear.rtv[i];
+      if (s) {
+         svga_propagate_surface(svga, s, FALSE);
+      }
+   }
+
+   if (svga->state.hw_clear.dsv) {
+      svga_propagate_surface(svga, svga->state.hw_clear.dsv, FALSE);
    }
 }
 
+
 /**
  * Check if we should call svga_propagate_surface on the surface.
  */
 boolean
-svga_surface_needs_propagation(struct pipe_surface *surf)
+svga_surface_needs_propagation(const struct pipe_surface *surf)
 {
-   struct svga_surface *s = svga_surface(surf);
+   const struct svga_surface *s = svga_surface_const(surf);
    struct svga_texture *tex = svga_texture(surf->texture);
 
    return s->dirty && s->handle != tex->handle;
 }
 
 
+static void
+svga_get_sample_position(struct pipe_context *context,
+                         unsigned sample_count, unsigned sample_index,
+                         float *pos_out)
+{
+   /* We can't actually query the device to learn the sample positions.
+    * These were grabbed from nvidia's driver.
+    */
+   static const float pos1[1][2] = {
+      { 0.5, 0.5 }
+   };
+   static const float pos2[2][2] = {
+      { 0.75, 0.75 },
+      { 0.25, 0.25 }
+   };
+   static const float pos4[4][2] = {
+      { 0.375000, 0.125000 },
+      { 0.875000, 0.375000 },
+      { 0.125000, 0.625000 },
+      { 0.625000, 0.875000 }
+   };
+   static const float pos8[8][2] = {
+      { 0.562500, 0.312500 },
+      { 0.437500, 0.687500 },
+      { 0.812500, 0.562500 },
+      { 0.312500, 0.187500 },
+      { 0.187500, 0.812500 },
+      { 0.062500, 0.437500 },
+      { 0.687500, 0.937500 },
+      { 0.937500, 0.062500 }
+   };
+   static const float pos16[16][2] = {
+      { 0.187500, 0.062500 },
+      { 0.437500, 0.187500 },
+      { 0.062500, 0.312500 },
+      { 0.312500, 0.437500 },
+      { 0.687500, 0.062500 },
+      { 0.937500, 0.187500 },
+      { 0.562500, 0.312500 },
+      { 0.812500, 0.437500 },
+      { 0.187500, 0.562500 },
+      { 0.437500, 0.687500 },
+      { 0.062500, 0.812500 },
+      { 0.312500, 0.937500 },
+      { 0.687500, 0.562500 },
+      { 0.937500, 0.687500 },
+      { 0.562500, 0.812500 },
+      { 0.812500, 0.937500 }
+   };
+   const float (*positions)[2];
+
+   switch (sample_count) {
+   case 2:
+      positions = pos2;
+      break;
+   case 4:
+      positions = pos4;
+      break;
+   case 8:
+      positions = pos8;
+      break;
+   case 16:
+      positions = pos16;
+      break;
+   default:
+      positions = pos1;
+   }
 
-
+   pos_out[0] = positions[sample_index][0];
+   pos_out[1] = positions[sample_index][1];
+}
 
 
 void
@@ -385,5 +977,5 @@ svga_init_surface_functions(struct svga_context *svga)
 {
    svga->pipe.create_surface = svga_create_surface;
    svga->pipe.surface_destroy = svga_surface_destroy;
+   svga->pipe.get_sample_position = svga_get_sample_position;
 }
-