radeonsi: fix read from compute / write from draw sync

A compute dispatch should see the result of a previous draw command.
radeonsi was missing this implicit sync, causing rendering artifacts:
the compute shader was reading from a texture still being written to
by the previous draw.

Framebuffer BOs are marked with RADEON_USAGE_NEEDS_IMPLICIT_SYNC,
so compute jobs will sync.

v2: use RADEON_USAGE_NEEDS_IMPLICIT_SYNC
v3: unconditionally make CB coherent after a flush

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com> (v3)
Reviewed-by: Marek Olšák <marek.olsak@amd.com> (v3)
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4032
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2878
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/1336
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8869>

5 files changed, 67 insertions, 4 deletions

diff --git a/src/gallium/drivers/radeon/radeon_winsys.h b/src/gallium/drivers/radeon/radeon_winsys.h
index 9f83c6ad770..44dbf532a10 100644
--- a/src/gallium/drivers/radeon/radeon_winsys.h
+++ b/src/gallium/drivers/radeon/radeon_winsys.h
@@ -98,7 +98,12 @@ enum radeon_bo_usage
   /* The winsys ensures that the CS submission will be scheduled after
    * previously flushed CSs referencing this BO in a conflicting way.
    */
-  RADEON_USAGE_SYNCHRONIZED = 8
+  RADEON_USAGE_SYNCHRONIZED = 8,
+
+  /* When used, an implicit sync is done to make sure a compute shader
+   * will read the written values from a previous draw.
+   */
+  RADEON_USAGE_NEEDS_IMPLICIT_SYNC = 16,
 };
 
 enum radeon_map_flags
diff --git a/src/gallium/drivers/radeonsi/si_compute.c b/src/gallium/drivers/radeonsi/si_compute.c
index 0f1ece6965d..e2752368439 100644
--- a/src/gallium/drivers/radeonsi/si_compute.c
+++ b/src/gallium/drivers/radeonsi/si_compute.c
@@ -823,6 +823,47 @@ static void si_emit_dispatch_packets(struct si_context *sctx, const struct pipe_
    radeon_end();
 }
 
+static bool si_check_needs_implicit_sync(struct si_context *sctx)
+{
+   /* If the compute shader is going to read from a texture/image written by a
+    * previous draw, we must wait for its completion before continuing.
+    * Buffers and image stores (from the draw) are not taken into consideration
+    * because that's the app responsibility.
+    *
+    * The OpenGL 4.6 spec says:
+    *
+    *    buffer object and texture stores performed by shaders are not
+    *    automatically synchronized
+    */
+   struct si_shader_info *info = &sctx->cs_shader_state.program->sel.info;
+   struct si_samplers *samplers = &sctx->samplers[PIPE_SHADER_COMPUTE];
+   unsigned mask = samplers->enabled_mask & info->base.textures_used;
+
+   while (mask) {
+      int i = u_bit_scan(&mask);
+      struct si_sampler_view *sview = (struct si_sampler_view *)samplers->views[i];
+
+      struct si_resource *res = si_resource(sview->base.texture);
+      if (sctx->ws->cs_is_buffer_referenced(&sctx->gfx_cs, res->buf,
+                                            RADEON_USAGE_NEEDS_IMPLICIT_SYNC))
+         return true;
+   }
+
+   struct si_images *images = &sctx->images[PIPE_SHADER_COMPUTE];
+   mask = u_bit_consecutive(0, info->base.num_images);
+
+   while (mask) {
+      int i = u_bit_scan(&mask);
+      struct pipe_image_view *sview = &images->views[i];
+
+      struct si_resource *res = si_resource(sview->resource);
+      if (sctx->ws->cs_is_buffer_referenced(&sctx->gfx_cs, res->buf,
+                                            RADEON_USAGE_NEEDS_IMPLICIT_SYNC))
+         return true;
+   }
+   return false;
+}
+
 static void si_launch_grid(struct pipe_context *ctx, const struct pipe_grid_info *info)
 {
    struct si_context *sctx = (struct si_context *)ctx;
@@ -849,6 +890,11 @@ static void si_launch_grid(struct pipe_context *ctx, const struct pipe_grid_info
       if (sctx->last_num_draw_calls != sctx->num_draw_calls) {
          si_update_fb_dirtiness_after_rendering(sctx);
          sctx->last_num_draw_calls = sctx->num_draw_calls;
+
+         if (sctx->force_cb_shader_coherent || si_check_needs_implicit_sync(sctx))
+            si_make_CB_shader_coherent(sctx, 0,
+                                       sctx->framebuffer.CB_has_shader_readable_metadata,
+                                       sctx->framebuffer.all_DCC_pipe_aligned);
       }
 
       si_decompress_textures(sctx, 1 << PIPE_SHADER_COMPUTE);
diff --git a/src/gallium/drivers/radeonsi/si_gfx_cs.c b/src/gallium/drivers/radeonsi/si_gfx_cs.c
index a63e7db6deb..200dceb6003 100644
--- a/src/gallium/drivers/radeonsi/si_gfx_cs.c
+++ b/src/gallium/drivers/radeonsi/si_gfx_cs.c
@@ -551,6 +551,13 @@ void si_begin_new_gfx_cs(struct si_context *ctx, bool first_cs)
       ctx->index_ring_base = ctx->index_ring_size_per_ib;
 
    ctx->index_ring_offset = 0;
+
+   /* All buffer references are removed on a flush, so si_check_needs_implicit_sync
+    * cannot determine if si_make_CB_shader_coherent() needs to be called.
+    * ctx->force_cb_shader_coherent will be cleared by the first call to
+    * si_make_CB_shader_coherent.
+    */
+   ctx->force_cb_shader_coherent = true;
 }
 
 void si_emit_surface_sync(struct si_context *sctx, struct radeon_cmdbuf *cs, unsigned cp_coher_cntl)
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 27da51eaff8..2aa13948fba 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -1266,6 +1266,8 @@ struct si_context {
    struct list_head shader_query_buffers;
    unsigned num_active_shader_queries;
 
+   bool force_cb_shader_coherent;
+
    /* Statistics gathering for the DCC enablement heuristic. It can't be
     * in si_texture because si_texture can be shared by multiple
     * contexts. This is for back buffers only. We shouldn't get too many
@@ -1745,6 +1747,7 @@ static inline void si_make_CB_shader_coherent(struct si_context *sctx, unsigned
                                               bool shaders_read_metadata, bool dcc_pipe_aligned)
 {
    sctx->flags |= SI_CONTEXT_FLUSH_AND_INV_CB | SI_CONTEXT_INV_VCACHE;
+   sctx->force_cb_shader_coherent = false;
 
    if (sctx->chip_class >= GFX10) {
       if (sctx->screen->info.tcc_harvested)
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index fb90fc80511..95f5a9834ee 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -2955,17 +2955,19 @@ static void si_emit_framebuffer_state(struct si_context *sctx)
 
       tex = (struct si_texture *)cb->base.texture;
       radeon_add_to_buffer_list(
-         sctx, &sctx->gfx_cs, &tex->buffer, RADEON_USAGE_READWRITE,
+         sctx, &sctx->gfx_cs, &tex->buffer, RADEON_USAGE_READWRITE | RADEON_USAGE_NEEDS_IMPLICIT_SYNC,
          tex->buffer.b.b.nr_samples > 1 ? RADEON_PRIO_COLOR_BUFFER_MSAA : RADEON_PRIO_COLOR_BUFFER);
 
       if (tex->cmask_buffer && tex->cmask_buffer != &tex->buffer) {
-         radeon_add_to_buffer_list(sctx, &sctx->gfx_cs, tex->cmask_buffer, RADEON_USAGE_READWRITE,
+         radeon_add_to_buffer_list(sctx, &sctx->gfx_cs, tex->cmask_buffer,
+                                   RADEON_USAGE_READWRITE | RADEON_USAGE_NEEDS_IMPLICIT_SYNC,
                                    RADEON_PRIO_SEPARATE_META);
       }
 
       if (tex->dcc_separate_buffer)
          radeon_add_to_buffer_list(sctx, &sctx->gfx_cs, tex->dcc_separate_buffer,
-                                   RADEON_USAGE_READWRITE, RADEON_PRIO_SEPARATE_META);
+                                   RADEON_USAGE_READWRITE | RADEON_USAGE_NEEDS_IMPLICIT_SYNC,
+                                   RADEON_PRIO_SEPARATE_META);
 
       /* Compute mutable surface parameters. */
       cb_color_base = tex->buffer.gpu_address >> 8;