vc4: Don't worry about partial Z/S clear if the other is already cleared.

We have to be careful to not smash the value they're clearing to, but
other than that we're fine.  Avoids quad clears in Processing, which likes
to do glClear(Z|S); glClear(Z).

Improves performance of Processing's QuadRendering demo at 5000 quads by
5.46507% +/- 1.35576% (n=15 before, 32 after)

diff --git a/src/gallium/drivers/vc4/vc4_draw.c b/src/gallium/drivers/vc4/vc4_draw.c
index 87c8a5f99c6e4dd5c0c2ad360bb39fbf0c1cd8f5..f44831f1e43fe1d3bd0c75656abef51ece85f28f 100644 (file)
--- a/src/gallium/drivers/vc4/vc4_draw.c
+++ b/src/gallium/drivers/vc4/vc4_draw.c
@@ -514,7 +514,7 @@ vc4_clear(struct pipe_context *pctx, unsigned buffers,
                  */
                 if ((zsclear == PIPE_CLEAR_DEPTH ||
                      zsclear == PIPE_CLEAR_STENCIL) &&
-                    (rsc->initialized_buffers & ~zsclear) &&
+                    (rsc->initialized_buffers & ~(zsclear | job->cleared)) &&
                     util_format_is_depth_and_stencil(vc4->framebuffer.zsbuf->format)) {
                         perf_debug("Partial clear of Z+stencil buffer, "
                                    "drawing a quad instead of fast clearing\n");
@@ -533,8 +533,12 @@ vc4_clear(struct pipe_context *pctx, unsigned buffers,
                 /* Though the depth buffer is stored with Z in the high 24,
                  * for this field we just need to store it in the low 24.
                  */
-                job->clear_depth = util_pack_z(PIPE_FORMAT_Z24X8_UNORM, depth);
-                job->clear_stencil = stencil;
+                if (buffers & PIPE_CLEAR_DEPTH) {
+                        job->clear_depth = util_pack_z(PIPE_FORMAT_Z24X8_UNORM,
+                                                       depth);
+                }
+                if (buffers & PIPE_CLEAR_STENCIL)
+                        job->clear_stencil = stencil;
 
                 rsc->initialized_buffers |= zsclear;
         }