Mesa (main): v3d: enable early Z/S clears

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: 07cfa2bd96e58e059a35ac2a4e03f7e093144bf4
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=07cfa2bd96e58e059a35ac2a4e03f7e093144bf4

Author: Juan A. Suarez Romero <jasuarez at igalia.com>
Date:   Tue Apr 19 16:52:44 2022 +0200

v3d: enable early Z/S clears

This performance optimization can be enabled if we are clearing Z/S
buffer, and not storing or loading it.

v2:
 - Add assertion on depth/stencil job loads (Iago)

Signed-off-by: Juan A. Suarez Romero <jasuarez at igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral at igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16166>

---

 src/gallium/drivers/v3d/v3d_context.h |  5 +++++
 src/gallium/drivers/v3d/v3dx_rcl.c    | 14 ++++++++++----
 2 files changed, 15 insertions(+), 4 deletions(-)

diff --git a/src/gallium/drivers/v3d/v3d_context.h b/src/gallium/drivers/v3d/v3d_context.h
index d8c2191074e..5bfe53ee6fc 100644
--- a/src/gallium/drivers/v3d/v3d_context.h
+++ b/src/gallium/drivers/v3d/v3d_context.h
@@ -470,6 +470,11 @@ struct v3d_job {
          */
         enum v3d_ez_state first_ez_state;
 
+        /**
+         * If this job has been configured to use early Z/S clear.
+         */
+        bool early_zs_clear;
+
         /**
          * Number of draw calls (not counting full buffer clears) queued in
          * the current job.
diff --git a/src/gallium/drivers/v3d/v3dx_rcl.c b/src/gallium/drivers/v3d/v3dx_rcl.c
index 088d454e95a..7c85525bc60 100644
--- a/src/gallium/drivers/v3d/v3dx_rcl.c
+++ b/src/gallium/drivers/v3d/v3dx_rcl.c
@@ -245,6 +245,7 @@ v3d_rcl_emit_loads(struct v3d_job *job, struct v3d_cl *cl, int layer)
         if ((loads_pending & PIPE_CLEAR_DEPTHSTENCIL) &&
             (V3D_VERSION >= 40 ||
              (job->zsbuf && job->zsbuf->texture->nr_samples > 1))) {
+                assert(!job->early_zs_clear);
                 struct pipe_surface *src = job->bbuf ? job->bbuf : job->zsbuf;
                 struct v3d_resource *rsc = v3d_resource(src->texture);
 
@@ -345,6 +346,7 @@ v3d_rcl_emit_stores(struct v3d_job *job, struct v3d_cl *cl, int layer)
 
         if (job->store & PIPE_CLEAR_DEPTHSTENCIL && job->zsbuf &&
             !(V3D_VERSION < 40 && job->zsbuf->texture->nr_samples <= 1)) {
+                assert(!job->early_zs_clear);
                 struct v3d_resource *rsc = v3d_resource(job->zsbuf->texture);
                 if (rsc->separate_stencil) {
                         if (job->store & PIPE_CLEAR_DEPTH) {
@@ -418,7 +420,7 @@ v3d_rcl_emit_stores(struct v3d_job *job, struct v3d_cl *cl, int layer)
          */
         if (job->clear) {
                 cl_emit(cl, CLEAR_TILE_BUFFERS, clear) {
-                        clear.clear_z_stencil_buffer = true;
+                        clear.clear_z_stencil_buffer = !job->early_zs_clear;
                         clear.clear_all_render_targets = true;
                 }
         }
@@ -657,7 +659,7 @@ emit_render_layer(struct v3d_job *job, uint32_t layer)
                 }
                 if (i == 0 || do_double_initial_tile_clear(job)) {
                         cl_emit(&job->rcl, CLEAR_TILE_BUFFERS, clear) {
-                                clear.clear_z_stencil_buffer = true;
+                                clear.clear_z_stencil_buffer = !job->early_zs_clear;
                                 clear.clear_all_render_targets = true;
                         }
                 }
@@ -724,9 +726,13 @@ v3dX(emit_rcl)(struct v3d_job *job)
                         struct v3d_surface *surf = v3d_surface(job->zsbuf);
                         config.internal_depth_type = surf->internal_type;
                 }
-#endif /* V3D_VERSION >= 40 */
 
-                /* XXX: Early D/S clear */
+                job->early_zs_clear = (job->clear & PIPE_CLEAR_DEPTHSTENCIL) &&
+                        !(job->load & PIPE_CLEAR_DEPTHSTENCIL) &&
+                        !(job->store & PIPE_CLEAR_DEPTHSTENCIL);
+
+                config.early_depth_stencil_clear = job->early_zs_clear;
+#endif /* V3D_VERSION >= 40 */
 
                 switch (job->first_ez_state) {
                 case V3D_EZ_UNDECIDED: