radeonsi: fix automatic DCC retiling after compute image stores

Only internal compute shaders use DCC stores, so the TODOs are not
critical yet.

Fixes: 1d64a1045ea - radeonsi: enable dcc image stores on gfx10+

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10261>
(cherry picked from commit a1653854f50aa429cb1a5e0ece063900cc988554)

4 files changed, 22 insertions, 0 deletions

diff --git a/src/gallium/drivers/radeonsi/si_compute.c b/src/gallium/drivers/radeonsi/si_compute.c
index 60977d35f1f..48ec79ac5a1 100644
--- a/src/gallium/drivers/radeonsi/si_compute.c
+++ b/src/gallium/drivers/radeonsi/si_compute.c
@@ -1004,6 +1004,18 @@ static void si_launch_grid(struct pipe_context *ctx, const struct pipe_grid_info
       si_log_compute_state(sctx, sctx->log);
    }
 
+   /* Mark displayable DCC as dirty for bound images. */
+   unsigned display_dcc_store_mask = sctx->images[PIPE_SHADER_COMPUTE].display_dcc_store_mask &
+                               BITFIELD_MASK(program->sel.info.base.num_images);
+   while (display_dcc_store_mask) {
+      struct si_texture *tex = (struct si_texture *)
+         sctx->images[PIPE_SHADER_COMPUTE].views[u_bit_scan(&display_dcc_store_mask)].resource;
+
+      si_mark_display_dcc_dirty(sctx, tex);
+   }
+
+   /* TODO: Bindless images don't set displayable_dcc_dirty after image stores. */
+
    sctx->compute_is_busy = true;
    sctx->num_compute_calls++;
 
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index c3a81daa434..5337ad36d8e 100644
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -686,6 +686,7 @@ static void si_disable_shader_image(struct si_context *ctx, unsigned shader, uns
 
       memcpy(descs->list + desc_slot * 8, null_image_descriptor, 8 * 4);
       images->enabled_mask &= ~(1u << slot);
+      images->display_dcc_store_mask &= ~(1u << slot);
       ctx->descriptors_dirty |= 1u << si_sampler_and_image_descriptors_idx(shader);
    }
 }
@@ -791,6 +792,7 @@ static void si_set_shader_image(struct si_context *ctx, unsigned shader, unsigne
 
    if (res->b.b.target == PIPE_BUFFER || view->shader_access & SI_IMAGE_ACCESS_AS_BUFFER) {
       images->needs_color_decompress_mask &= ~(1 << slot);
+      images->display_dcc_store_mask &= ~(1u << slot);
       res->bind_history |= PIPE_BIND_SHADER_IMAGE;
    } else {
       struct si_texture *tex = (struct si_texture *)res;
@@ -802,6 +804,11 @@ static void si_set_shader_image(struct si_context *ctx, unsigned shader, unsigne
          images->needs_color_decompress_mask &= ~(1 << slot);
       }
 
+      if (tex->surface.display_dcc_offset && view->access & PIPE_IMAGE_ACCESS_WRITE)
+         images->display_dcc_store_mask |= 1u << slot;
+      else
+         images->display_dcc_store_mask &= ~(1u << slot);
+
       if (vi_dcc_enabled(tex, level) && p_atomic_read(&tex->framebuffers_bound))
          ctx->need_check_render_feedback = true;
    }
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 27b4b8af050..442671a8c08 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -718,6 +718,7 @@ struct si_images {
    struct pipe_image_view views[SI_NUM_IMAGES];
    uint32_t needs_color_decompress_mask;
    unsigned enabled_mask;
+   unsigned display_dcc_store_mask;
 };
 
 struct si_framebuffer {
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.cpp b/src/gallium/drivers/radeonsi/si_state_draw.cpp
index 63644d02925..c54f14da805 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.cpp
+++ b/src/gallium/drivers/radeonsi/si_state_draw.cpp
@@ -2200,6 +2200,8 @@ static void si_draw_vbo(struct pipe_context *ctx,
          sctx->num_prim_restart_calls++;
    }
 
+   /* TODO: Set displayable_dcc_dirty if image stores are used. */
+
    DRAW_CLEANUP;
 }