intel: Stop creating the wrapped stencil irb.

There were only two places it was really used at this point, which was
in the batchbuffer emit of the separate stencil packets for gen6/7.
Just write in the ->stencil_mt reference in those two places and ditch
all this flailing around with allocation and refcounts.

v2: Fix separate stencil on gen7.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

src/mesa/drivers/dri/i965/brw_misc_state.c

@@ -208,6 +208,7 @@ brw_depthbuffer_format(struct brw_context *brw)
 
    if (!drb &&
        (srb = intel_get_renderbuffer(fb, BUFFER_STENCIL)) &&
+       !srb->mt->stencil_mt &&
        srb->Base.Format == MESA_FORMAT_S8_Z24) {
       drb = srb;
    }
@@ -239,8 +240,10 @@ static void emit_depthbuffer(struct brw_context *brw)
    /* _NEW_BUFFERS */
    struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
    struct intel_renderbuffer *stencil_irb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
+   struct intel_mipmap_tree *stencil_mt = NULL;
    struct intel_region *hiz_region = NULL;
    unsigned int len;
+   bool separate_stencil = false;
 
    if (depth_irb &&
        depth_irb->mt &&
@@ -256,17 +259,21 @@ static void emit_depthbuffer(struct brw_context *brw)
       intel_emit_depth_stall_flushes(intel);
    }
 
-   /*
-    * If either depth or stencil buffer has packed depth/stencil format,
-    * then don't use separate stencil. Emit only a depth buffer.
+   /* Find the real separate stencil mt if present. */
+   if (stencil_irb) {
+      stencil_mt = stencil_irb->mt;
+      if (stencil_mt->stencil_mt)
+	 stencil_mt = stencil_mt->stencil_mt;
+
+      if (stencil_mt->format == MESA_FORMAT_S8)
+	 separate_stencil = true;
+   }
+
+   /* If there's a packed depth/stencil bound to stencil only, we need to
+    * emit the packed depth/stencil buffer packet.
     */
-   if (depth_irb && depth_irb->Base.Format == MESA_FORMAT_S8_Z24) {
-      stencil_irb = NULL;
-   } else if (!depth_irb && stencil_irb
-	      && stencil_irb->Base.Format == MESA_FORMAT_S8_Z24) {
+   if (!depth_irb && stencil_irb && !separate_stencil)
       depth_irb = stencil_irb;
-      stencil_irb = NULL;
-   }
 
    if (intel->gen >= 6)
       len = 7;
@@ -275,7 +282,7 @@ static void emit_depthbuffer(struct brw_context *brw)
    else
       len = 5;
 
-   if (!depth_irb && !stencil_irb) {
+   if (!depth_irb && !separate_stencil) {
       BEGIN_BATCH(len);
       OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
       OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
@@ -292,7 +299,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 
       ADVANCE_BATCH();
 
-   } else if (!depth_irb && stencil_irb) {
+   } else if (!depth_irb && separate_stencil) {
       /*
        * There exists a separate stencil buffer but no depth buffer.
        *
@@ -317,10 +324,9 @@ static void emit_depthbuffer(struct brw_context *brw)
        * Section 7.5.5.1.1 3DSTATE_DEPTH_BUFFER, Bit 1.27 Tiled Surface:
        *     [DevGT+]: This field must be set to TRUE.
        */
-      struct intel_region *region = stencil_irb->mt->region;
+      struct intel_region *region = stencil_mt->region;
 
       assert(intel->has_separate_stencil);
-      assert(stencil_irb->Base.Format == MESA_FORMAT_S8);
 
       BEGIN_BATCH(len);
       OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
@@ -346,7 +352,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       uint32_t tile_x, tile_y, offset;
 
       /* If using separate stencil, hiz must be enabled. */
-      assert(!stencil_irb || hiz_region);
+      assert(!separate_stencil || hiz_region);
 
       offset = intel_renderbuffer_tile_offsets(depth_irb, &tile_x, &tile_y);
 
@@ -381,7 +387,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       ADVANCE_BATCH();
    }
 
-   if (hiz_region || stencil_irb) {
+   if (hiz_region || separate_stencil) {
       /*
        * In the 3DSTATE_DEPTH_BUFFER batch emitted above, the 'separate
        * stencil enable' and 'hiz enable' bits were set. Therefore we must
@@ -408,8 +414,8 @@ static void emit_depthbuffer(struct brw_context *brw)
       }
 
       /* Emit stencil buffer. */
-      if (stencil_irb) {
-	 struct intel_region *region = stencil_irb->mt->region;
+      if (separate_stencil) {
+	 struct intel_region *region = stencil_mt->region;
 	 BEGIN_BATCH(3);
 	 OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
 	 OUT_BATCH(region->pitch * region->cpp - 1);

src/mesa/drivers/dri/i965/gen7_misc_state.c

@@ -38,20 +38,32 @@ static void emit_depthbuffer(struct brw_context *brw)
    /* _NEW_BUFFERS */
    struct intel_renderbuffer *drb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
    struct intel_renderbuffer *srb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
+   struct intel_mipmap_tree *depth_mt = NULL, *stencil_mt = NULL;
 
-   intel_emit_depth_stall_flushes(intel);
+   if (drb)
+      depth_mt = drb->mt;
+
+   if (srb) {
+      stencil_mt = srb->mt;
+      if (stencil_mt->stencil_mt)
+	 stencil_mt = stencil_mt->stencil_mt;
+
+      assert(stencil_mt->format == MESA_FORMAT_S8);
+   }
 
    /* Gen7 doesn't support packed depth/stencil */
-   assert(srb == NULL || srb != drb);
+   assert(stencil_mt == NULL || depth_mt != stencil_mt);
+
+   intel_emit_depth_stall_flushes(intel);
 
-   if (drb == NULL) {
+   if (depth_mt == NULL) {
       uint32_t dw1 = BRW_DEPTHFORMAT_D32_FLOAT << 18;
       uint32_t dw3 = 0;
 
-      if (srb == NULL) {
+      if (stencil_mt == NULL) {
 	 dw1 |= (BRW_SURFACE_NULL << 29);
       } else {
-	 struct intel_region *region = srb->mt->region;
+	 struct intel_region *region = stencil_mt->region;
 
 	 /* _NEW_STENCIL: enable stencil buffer writes */
 	 dw1 |= ((ctx->Stencil.WriteMask != 0) << 27);
@@ -71,7 +83,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       OUT_BATCH(0);
       ADVANCE_BATCH();
    } else {
-      struct intel_region *region = drb->mt->region;
+      struct intel_region *region = depth_mt->region;
       uint32_t tile_x, tile_y, offset;
 
       offset = intel_renderbuffer_tile_offsets(drb, &tile_x, &tile_y);
@@ -84,7 +96,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       OUT_BATCH(((region->pitch * region->cpp) - 1) |
 		(brw_depthbuffer_format(brw) << 18) |
 		(0 << 22) /* no HiZ buffer */ |
-		((srb != NULL && ctx->Stencil.WriteMask != 0) << 27) |
+		((stencil_mt != NULL && ctx->Stencil.WriteMask != 0) << 27) |
 		((ctx->Depth.Mask != 0) << 28) |
 		(BRW_SURFACE_2D << 29));
       OUT_RELOC(region->bo,
@@ -104,7 +116,7 @@ static void emit_depthbuffer(struct brw_context *brw)
    OUT_BATCH(0);
    ADVANCE_BATCH();
 
-   if (srb == NULL) {
+   if (stencil_mt == NULL) {
       BEGIN_BATCH(3);
       OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
       OUT_BATCH(0);
@@ -113,8 +125,8 @@ static void emit_depthbuffer(struct brw_context *brw)
    } else {
       BEGIN_BATCH(3);
       OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
-      OUT_BATCH(srb->mt->region->pitch * srb->mt->region->cpp - 1);
-      OUT_RELOC(srb->mt->region->bo,
+      OUT_BATCH(stencil_mt->region->pitch * stencil_mt->region->cpp - 1);
+      OUT_RELOC(stencil_mt->region->bo,
 	        I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
 		0);
       ADVANCE_BATCH();

src/mesa/drivers/dri/intel/intel_fbo.c

@@ -81,9 +81,12 @@ struct intel_region*
 intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex)
 {
    struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex);
-   if (irb && irb->mt)
-      return irb->mt->region;
-   else
+   if (irb && irb->mt) {
+      if (attIndex == BUFFER_STENCIL && irb->mt->stencil_mt)
+	 return irb->mt->stencil_mt->region;
+      else
+	 return irb->mt->region;
+   } else
       return NULL;
 }
 
@@ -111,7 +114,6 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb)
    intel_miptree_release(&irb->mt);
 
    _mesa_reference_renderbuffer(&irb->wrapped_depth, NULL);
-   _mesa_reference_renderbuffer(&irb->wrapped_stencil, NULL);
 
    free(irb);
 }
@@ -260,24 +262,19 @@ intel_alloc_renderbuffer_storage(struct gl_context * ctx, struct gl_renderbuffer
 
    if (irb->mt->stencil_mt) {
       bool ok;
-      struct intel_renderbuffer *depth_irb, *stencil_irb;
+      struct intel_renderbuffer *depth_irb;
 
       /* The RB got allocated as separate stencil.  Hook up our wrapped
-       * renderbuffers so that consumers of intel_get_renderbuffer() end up
-       * with pointers to the separate pieces.
+       * renderbuffer so that consumers of intel_get_renderbuffer(BUFFER_DEPTH)
+       * end up with pointers to the separate depth.
        */
       if (!irb->wrapped_depth) {
 	 _mesa_reference_renderbuffer(&irb->wrapped_depth,
 				      intel_new_renderbuffer(ctx, ~0));
       }
-      if (!irb->wrapped_stencil) {
-	 _mesa_reference_renderbuffer(&irb->wrapped_stencil,
-				      intel_new_renderbuffer(ctx, ~0));
-      }
 
       depth_irb = intel_renderbuffer(irb->wrapped_depth);
-      stencil_irb = intel_renderbuffer(irb->wrapped_stencil);
-      if (!depth_irb || !stencil_irb) {
+      if (!depth_irb) {
 	 intel_miptree_release(&irb->mt);
 	 return false;
       }
@@ -288,13 +285,6 @@ intel_alloc_renderbuffer_storage(struct gl_context * ctx, struct gl_renderbuffer
 					     MESA_FORMAT_X8_Z24,
 					     GL_DEPTH_COMPONENT24);
       assert(ok);
-
-      ok = intel_renderbuffer_update_wrapper(intel, stencil_irb,
-					     irb->mt->stencil_mt,
-					     0, 0, /* level, layer */
-					     MESA_FORMAT_S8,
-					     GL_STENCIL_INDEX);
-      assert(ok);
    }
 
    return true;
@@ -541,32 +531,25 @@ intel_renderbuffer_update_wrapper(struct intel_context *intel,
    irb->mt_level = level;
    irb->mt_layer = layer;
 
+   intel_miptree_reference(&irb->mt, mt);
    if (mt->stencil_mt && _mesa_is_depthstencil_format(rb->InternalFormat)) {
-      assert((irb->wrapped_depth == NULL) == (irb->wrapped_stencil == NULL));
-
       struct intel_renderbuffer *depth_irb;
-      struct intel_renderbuffer *stencil_irb;
 
       if (!irb->wrapped_depth) {
 	 depth_irb = intel_renderbuffer_wrap_miptree(intel,
 	                                             mt, level, layer,
 	                                             MESA_FORMAT_X8_Z24,
 	                                             GL_DEPTH_COMPONENT24);
-	 stencil_irb = intel_renderbuffer_wrap_miptree(intel,
-	                                               mt->stencil_mt,
-	                                               level, layer,
-	                                               MESA_FORMAT_S8,
-	                                               GL_STENCIL_INDEX8);
 	 _mesa_reference_renderbuffer(&irb->wrapped_depth, &depth_irb->Base);
-	 _mesa_reference_renderbuffer(&irb->wrapped_stencil, &stencil_irb->Base);
 
-	 if (!irb->wrapped_depth || !irb->wrapped_stencil)
+	 if (!irb->wrapped_depth) {
+	    intel_miptree_release(&irb->mt);
 	    return false;
+	 }
       } else {
 	 bool ok = true;
 
 	 depth_irb = intel_renderbuffer(irb->wrapped_depth);
-	 stencil_irb = intel_renderbuffer(irb->wrapped_stencil);
 
 	 ok &= intel_renderbuffer_update_wrapper(intel,
 	                                         depth_irb,
@@ -574,20 +557,12 @@ intel_renderbuffer_update_wrapper(struct intel_context *intel,
 	                                         level, layer,
 	                                         MESA_FORMAT_X8_Z24,
 	                                         GL_DEPTH_COMPONENT24);
-	 ok &= intel_renderbuffer_update_wrapper(intel,
-	                                         stencil_irb,
-	                                         mt->stencil_mt,
-	                                         level, layer,
-	                                         MESA_FORMAT_S8,
-	                                         GL_STENCIL_INDEX8);
-	 if (!ok)
+	 if (!ok) {
+	    intel_miptree_release(&irb->mt);
 	    return false;
+	 }
       }
-
-      intel_miptree_reference(&depth_irb->mt->stencil_mt, stencil_irb->mt);
-      intel_miptree_reference(&irb->mt, depth_irb->mt);
    } else {
-      intel_miptree_reference(&irb->mt, mt);
       intel_renderbuffer_set_draw_offset(irb);
 
       if (mt->hiz_mt == NULL &&
@@ -857,8 +832,11 @@ intel_validate_framebuffer(struct gl_context *ctx, struct gl_framebuffer *fb)
 
    if (depthRb)
       depth_mt = depthRb->mt;
-   if (stencilRb)
+   if (stencilRb) {
       stencil_mt = stencilRb->mt;
+      if (stencil_mt->stencil_mt)
+	 stencil_mt = stencil_mt->stencil_mt;
+   }
 
    if (depth_mt && stencil_mt) {
       if (depth_mt == stencil_mt) {

src/mesa/drivers/dri/intel/intel_fbo.h

@@ -72,7 +72,6 @@ struct intel_renderbuffer
     * are the real renderbuffers.
     */
    struct gl_renderbuffer *wrapped_depth;
-   struct gl_renderbuffer *wrapped_stencil;
    /** \} */
 
    GLuint draw_x, draw_y; /**< Offset of drawing within the region */
@@ -134,11 +133,6 @@ intel_get_renderbuffer(struct gl_framebuffer *fb, gl_buffer_index attIndex)
 	 irb = intel_renderbuffer(irb->wrapped_depth);
       }
       break;
-   case BUFFER_STENCIL:
-      if (irb->wrapped_stencil) {
-	 irb = intel_renderbuffer(irb->wrapped_stencil);
-      }
-      break;
    default:
       break;
    }

src/mesa/drivers/dri/intel/intel_mipmap_tree.h

@@ -149,8 +149,7 @@ struct intel_mipmap_tree
     * two miptrees for storing the data.  If the depthstencil texture or rb is
     * MESA_FORMAT_Z32_FLOAT_X24S8, then mt->format will be
     * MESA_FORMAT_Z32_FLOAT, otherwise for MESA_FORMAT_S8_Z24 objects it will be
-    * MESA_FORMAT_S8_Z24 (textures) or MESA_FORMAT_X8_Z24 (renderbuffers).
-    * mt->stencil_mt->format will always be MESA_FORMAT_S8.
+    * MESA_FORMAT_S8_Z24.
     */
    gl_format format;