[Mesa-dev] [PATCH 3/4] radeonsi/gfx9: don't ever flush the TC metadata cache

Marek Olšák maraeo at gmail.com
Tue Jun 20 17:06:30 UTC 2017


From: Marek Olšák <marek.olsak at amd.com>

The closed Vulkan driver doesn't do it either.

Also remove some old comments that aren't useful.
---
 src/gallium/drivers/radeonsi/si_state_draw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 85ceaca..332e0c4 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -964,36 +964,29 @@ void si_emit_cache_flush(struct si_context *sctx)
 			break;
 		case SI_CONTEXT_FLUSH_AND_INV_DB:
 			cb_db_event = V_028A90_FLUSH_AND_INV_DB_DATA_TS;
 			break;
 		default:
 			/* both CB & DB */
 			cb_db_event = V_028A90_CACHE_FLUSH_AND_INV_TS_EVENT;
 		}
 
 		/* TC    | TC_WB         = invalidate L2 data
-		 * TC_MD | TC_WB         = invalidate L2 metadata
+		 * TC_MD | TC_WB         = invalidate L2 metadata (DCC, etc.)
 		 * TC    | TC_WB | TC_MD = invalidate L2 data & metadata
-		 *
-		 * The metadata cache must always be invalidated for coherency
-		 * between CB/DB and shaders. (metadata = HTILE, CMASK, DCC)
-		 *
-		 * TC must be invalidated on GFX9 only if the CB/DB surface is
-		 * not pipe-aligned. If the surface is RB-aligned, it might not
-		 * strictly be pipe-aligned since RB alignment takes precendence.
 		 */
-		tc_flags = EVENT_TC_WB_ACTION_ENA |
-			   EVENT_TC_MD_ACTION_ENA;
+		tc_flags = 0;
 
 		/* Ideally flush TC together with CB/DB. */
 		if (rctx->flags & SI_CONTEXT_INV_GLOBAL_L2) {
 			tc_flags |= EVENT_TC_ACTION_ENA |
+				    EVENT_TC_WB_ACTION_ENA |
 				    EVENT_TCL1_ACTION_ENA;
 
 			/* Clear the flags. */
 			rctx->flags &= ~(SI_CONTEXT_INV_GLOBAL_L2 |
 					 SI_CONTEXT_WRITEBACK_GLOBAL_L2 |
 					 SI_CONTEXT_INV_VMEM_L1);
 			sctx->b.num_L2_invalidates++;
 		}
 
 		/* Do the flush (enqueue the event and wait for it). */
-- 
2.7.4



More information about the mesa-dev mailing list