st/mesa: implement accelerated stencil blitting using shader stencil export

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

4 files changed, 41 insertions, 12 deletions

diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c
index a50b79a..1486779 100644
--- a/src/mesa/state_tracker/st_cb_blit.c
+++ b/src/mesa/state_tracker/st_cb_blit.c
@@ -242,7 +242,7 @@ st_BlitFramebuffer(struct gl_context *ctx,
                           srcX0, srcY0, srcX1, srcY1,
                           srcAtt->Zoffset + srcAtt->CubeMapFace,
                           dstSurf, dstX0, dstY0, dstX1, dstY1,
-                          0.0, pFilter, TGSI_WRITEMASK_XYZW);
+                          0.0, pFilter, TGSI_WRITEMASK_XYZW, 0);
       }
       else {
          struct st_renderbuffer *srcRb =
@@ -257,7 +257,7 @@ st_BlitFramebuffer(struct gl_context *ctx,
                           srcX0, srcY0, srcX1, srcY1,
                           srcSurf->u.tex.first_layer,
                           dstSurf, dstX0, dstY0, dstX1, dstY1,
-                          0.0, pFilter, TGSI_WRITEMASK_XYZW);
+                          0.0, pFilter, TGSI_WRITEMASK_XYZW, 0);
       }
    }
 
@@ -281,6 +281,13 @@ st_BlitFramebuffer(struct gl_context *ctx,
       struct pipe_surface *dstDepthSurf =
          dstDepthRb ? dstDepthRb->surface : NULL;
 
+      struct st_renderbuffer *srcStencilRb =
+         st_renderbuffer(readFB->Attachment[BUFFER_STENCIL].Renderbuffer);
+      struct st_renderbuffer *dstStencilRb =
+         st_renderbuffer(drawFB->Attachment[BUFFER_STENCIL].Renderbuffer);
+      struct pipe_surface *dstStencilSurf =
+         dstStencilRb ? dstStencilRb->surface : NULL;
+
       if ((mask & depthStencil) == depthStencil &&
           st_is_depth_stencil_combined(srcDepth, srcStencil) &&
           st_is_depth_stencil_combined(dstDepth, dstStencil)) {
@@ -294,7 +301,15 @@ st_BlitFramebuffer(struct gl_context *ctx,
                           srcX0, srcY0, srcX1, srcY1,
                           srcDepthRb->surface->u.tex.first_layer,
                           dstDepthSurf, dstX0, dstY0, dstX1, dstY1,
-                          0.0, pFilter, 0);
+                          0.0, pFilter, 0,
+                          BLIT_WRITEMASK_Z |
+                          (st->has_stencil_export ? BLIT_WRITEMASK_STENCIL
+                                                  : 0));
+
+         if (!st->has_stencil_export) {
+            _mesa_problem(ctx, "st_BlitFramebuffer(STENCIL) "
+                               "software fallback not implemented");
+         }
       }
       else {
          /* blitting depth and stencil separately */
@@ -305,12 +320,22 @@ st_BlitFramebuffer(struct gl_context *ctx,
                              srcX0, srcY0, srcX1, srcY1,
                              srcDepthRb->surface->u.tex.first_layer,
                              dstDepthSurf, dstX0, dstY0, dstX1, dstY1,
-                             0.0, pFilter, 0);
+                             0.0, pFilter, 0, BLIT_WRITEMASK_Z);
          }
 
          if (mask & GL_STENCIL_BUFFER_BIT) {
-            /* blit stencil only */
-            _mesa_problem(ctx, "st_BlitFramebuffer(STENCIL) not completed");
+            if (st->has_stencil_export) {
+               util_blit_pixels(st->blit, srcStencilRb->texture,
+                                srcStencilRb->surface->u.tex.level,
+                                srcX0, srcY0, srcX1, srcY1,
+                                srcStencilRb->surface->u.tex.first_layer,
+                                dstStencilSurf, dstX0, dstY0, dstX1, dstY1,
+                                0.0, pFilter, 0, BLIT_WRITEMASK_STENCIL);
+            }
+            else {
+               _mesa_problem(ctx, "st_BlitFramebuffer(STENCIL) "
+                                  "software fallback not implemented");
+            }
          }
       }
    }
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index a1d6951..a7f57b9 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -944,7 +944,7 @@ st_CopyTexSubImage(struct gl_context *ctx, GLuint dims,
    struct pipe_screen *screen = pipe->screen;
    enum pipe_format dest_format, src_format;
    GLboolean matching_base_formats;
-   GLuint format_writemask, sample_count;
+   GLuint color_writemask, zs_writemask, sample_count;
    struct pipe_surface *dest_surface = NULL;
    GLboolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP);
    struct pipe_surface surf_tmpl;
@@ -1025,15 +1025,17 @@ st_CopyTexSubImage(struct gl_context *ctx, GLuint dims,
    }
 
    if (texBaseFormat == GL_DEPTH_COMPONENT) {
-      format_writemask = TGSI_WRITEMASK_XYZW;
+      color_writemask = 0;
+      zs_writemask = BLIT_WRITEMASK_Z;
       dst_usage = PIPE_BIND_DEPTH_STENCIL;
    }
    else {
-      format_writemask = compatible_src_dst_formats(ctx, &strb->Base, texImage);
+      color_writemask = compatible_src_dst_formats(ctx, &strb->Base, texImage);
+      zs_writemask = 0;
       dst_usage = PIPE_BIND_RENDER_TARGET;
    }
 
-   if (!format_writemask ||
+   if ((!color_writemask && !zs_writemask) ||
        !screen->is_format_supported(screen, src_format,
                                     PIPE_TEXTURE_2D, sample_count,
                                     PIPE_BIND_SAMPLER_VIEW) ||
@@ -1076,7 +1078,7 @@ st_CopyTexSubImage(struct gl_context *ctx, GLuint dims,
                     destX, destY,
                     destX + width, destY + height,
                     0.0, PIPE_TEX_MIPFILTER_NEAREST,
-                    format_writemask);
+                    color_writemask, zs_writemask);
    pipe_surface_reference(&dest_surface, NULL);
 
    /* Restore conditional rendering state. */
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index 132dcc0..117ea90 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -187,6 +187,8 @@ st_create_context_priv( struct gl_context *ctx, struct pipe_context *pipe )
    st->pixel_xfer.cache = _mesa_new_program_cache();
 
    st->force_msaa = st_get_msaa();
+   st->has_stencil_export =
+      screen->get_param(screen, PIPE_CAP_SHADER_STENCIL_EXPORT);
 
    /* GL limits and extensions */
    st_init_limits(st);
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index 734b4d9..a3f44b3 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -81,7 +81,7 @@ struct st_context
    struct draw_stage *rastpos_stage;  /**< For glRasterPos */
    GLboolean clamp_frag_color_in_shader;
    GLboolean clamp_vert_color_in_shader;
-
+   boolean has_stencil_export; /**< can do shader stencil export? */
 
    /* On old libGL's for linux we need to invalidate the drawables
     * on glViewpport calls, this is set via a option.