radeonsi: flush if sampler views and images use too much memory

Reviewed-by: Nicolai Hähnle <[email protected]>

2 files changed, 63 insertions, 19 deletions

diff --git a/src/gallium/drivers/radeon/r600_cs.h b/src/gallium/drivers/radeon/r600_cs.h
index 72a91bf56aa..988774aac8f 100644
--- a/src/gallium/drivers/radeon/r600_cs.h
+++ b/src/gallium/drivers/radeon/r600_cs.h
@@ -54,6 +54,40 @@ static inline unsigned radeon_add_to_buffer_list(struct r600_common_context *rct
 				      rbo->domains, priority) * 4;
 }
 
+/**
+ * Same as above, but also checks memory usage and flushes the context
+ * accordingly.
+ *
+ * When this SHOULD NOT be used:
+ *
+ * - if r600_context_add_resource_size has been called for the buffer
+ *   followed by *_need_cs_space for checking the memory usage
+ *
+ * - if r600_need_dma_space has been called for the buffer
+ *
+ * - when emitting state packets and draw packets (because preceding packets
+ *   can't be re-emitted at that point)
+ *
+ * - if shader resource "enabled_mask" is not up-to-date or there is
+ *   a different constraint disallowing a context flush
+ */
+static inline unsigned
+radeon_add_to_buffer_list_check_mem(struct r600_common_context *rctx,
+				    struct r600_ring *ring,
+				    struct r600_resource *rbo,
+				    enum radeon_bo_usage usage,
+				    enum radeon_bo_priority priority,
+				    bool check_mem)
+{
+	if (check_mem &&
+	    !rctx->ws->cs_memory_below_limit(ring->cs,
+					     rctx->vram + rbo->vram_usage,
+					     rctx->gtt + rbo->gart_usage))
+		ring->flush(rctx, RADEON_FLUSH_ASYNC, NULL);
+
+	return radeon_add_to_buffer_list(rctx, ring, rbo, usage, priority);
+}
+
 static inline void r600_emit_reloc(struct r600_common_context *rctx,
 				   struct r600_ring *ring, struct r600_resource *rbo,
 				   enum radeon_bo_usage usage,
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index 1cbb86a4b1e..ea585845435 100644
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -296,9 +296,12 @@ static void si_release_sampler_views(struct si_sampler_views *views)
 static void si_sampler_view_add_buffer(struct si_context *sctx,
 				       struct pipe_resource *resource,
 				       enum radeon_bo_usage usage,
-				       bool is_stencil_sampler)
+				       bool is_stencil_sampler,
+				       bool check_mem)
 {
 	struct r600_resource *rres;
+	struct r600_texture *rtex;
+	enum radeon_bo_priority priority;
 
 	if (!resource)
 		return;
@@ -311,17 +314,23 @@ static void si_sampler_view_add_buffer(struct si_context *sctx,
 	}
 
 	rres = (struct r600_resource*)resource;
-	radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, rres, usage,
-				  r600_get_sampler_view_priority(rres));
+	priority = r600_get_sampler_view_priority(rres);
 
-	if (resource->target != PIPE_BUFFER) {
-		struct r600_texture *rtex = (struct r600_texture*)resource;
+	radeon_add_to_buffer_list_check_mem(&sctx->b, &sctx->b.gfx,
+					    rres, usage, priority,
+					    check_mem);
 
-		if (rtex->dcc_separate_buffer)
-			radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
-						  rtex->dcc_separate_buffer, usage,
-						  RADEON_PRIO_DCC);
-	}
+	if (resource->target == PIPE_BUFFER)
+		return;
+
+	/* Now add separate DCC if it's present. */
+	rtex = (struct r600_texture*)resource;
+	if (!rtex->dcc_separate_buffer)
+		return;
+
+	radeon_add_to_buffer_list_check_mem(&sctx->b, &sctx->b.gfx,
+					    rtex->dcc_separate_buffer, usage,
+					    RADEON_PRIO_DCC, check_mem);
 }
 
 static void si_sampler_views_begin_new_cs(struct si_context *sctx,
@@ -336,7 +345,7 @@ static void si_sampler_views_begin_new_cs(struct si_context *sctx,
 
 		si_sampler_view_add_buffer(sctx, sview->base.texture,
 					   RADEON_USAGE_READ,
-					   sview->is_stencil_sampler);
+					   sview->is_stencil_sampler, false);
 	}
 }
 
@@ -403,7 +412,7 @@ static void si_set_sampler_view(struct si_context *sctx,
 
 		si_sampler_view_add_buffer(sctx, view->texture,
 					   RADEON_USAGE_READ,
-					   rview->is_stencil_sampler);
+					   rview->is_stencil_sampler, true);
 
 		pipe_sampler_view_reference(&views->views[slot], view);
 		memcpy(desc, rview->state, 8*4);
@@ -565,7 +574,7 @@ si_image_views_begin_new_cs(struct si_context *sctx, struct si_images_info *imag
 		assert(view->resource);
 
 		si_sampler_view_add_buffer(sctx, view->resource,
-					   RADEON_USAGE_READWRITE, false);
+					   RADEON_USAGE_READWRITE, false, false);
 	}
 }
 
@@ -624,7 +633,7 @@ static void si_set_shader_image(struct si_context *ctx,
 		util_copy_image_view(&images->views[slot], view);
 
 	si_sampler_view_add_buffer(ctx, &res->b.b,
-				   RADEON_USAGE_READWRITE, false);
+				   RADEON_USAGE_READWRITE, false, true);
 
 	if (res->b.b.target == PIPE_BUFFER) {
 		if (view->access & PIPE_IMAGE_ACCESS_WRITE)
@@ -1506,9 +1515,10 @@ static void si_invalidate_buffer(struct pipe_context *ctx, struct pipe_resource
 				sctx->descriptors_dirty |=
 					1u << si_sampler_descriptors_idx(shader);
 
-				radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
-						      rbuffer, RADEON_USAGE_READ,
-						      RADEON_PRIO_SAMPLER_BUFFER);
+				radeon_add_to_buffer_list_check_mem(&sctx->b, &sctx->b.gfx,
+								    rbuffer, RADEON_USAGE_READ,
+								    RADEON_PRIO_SAMPLER_BUFFER,
+								    true);
 			}
 		}
 	}
@@ -1534,10 +1544,10 @@ static void si_invalidate_buffer(struct pipe_context *ctx, struct pipe_resource
 				sctx->descriptors_dirty |=
 					1u << si_image_descriptors_idx(shader);
 
-				radeon_add_to_buffer_list(
+				radeon_add_to_buffer_list_check_mem(
 					&sctx->b, &sctx->b.gfx, rbuffer,
 					RADEON_USAGE_READWRITE,
-					RADEON_PRIO_SAMPLER_BUFFER);
+					RADEON_PRIO_SAMPLER_BUFFER, true);
 			}
 		}
 	}