radeonsi: make si_cp_wait_mem more configurable

Tested-by: Dieter Nützel <[email protected]>

5 files changed, 8 insertions, 8 deletions

diff --git a/src/gallium/drivers/radeonsi/si_fence.c b/src/gallium/drivers/radeonsi/si_fence.c
index d385f445774..b6920c95e34 100644
--- a/src/gallium/drivers/radeonsi/si_fence.c
+++ b/src/gallium/drivers/radeonsi/si_fence.c
@@ -160,13 +160,11 @@ unsigned si_cp_write_fence_dwords(struct si_screen *screen)
 	return dwords;
 }
 
-void si_cp_wait_mem(struct si_context *ctx,
+void si_cp_wait_mem(struct si_context *ctx, struct radeon_cmdbuf *cs,
 		    uint64_t va, uint32_t ref, uint32_t mask, unsigned flags)
 {
-	struct radeon_cmdbuf *cs = ctx->gfx_cs;
-
 	radeon_emit(cs, PKT3(PKT3_WAIT_REG_MEM, 5, 0));
-	radeon_emit(cs, WAIT_REG_MEM_EQUAL | WAIT_REG_MEM_MEM_SPACE(1) | flags);
+	radeon_emit(cs, WAIT_REG_MEM_MEM_SPACE(1) | flags);
 	radeon_emit(cs, va);
 	radeon_emit(cs, va >> 32);
 	radeon_emit(cs, ref); /* reference value */
diff --git a/src/gallium/drivers/radeonsi/si_perfcounter.c b/src/gallium/drivers/radeonsi/si_perfcounter.c
index fc2c58854bc..f15f310dd11 100644
--- a/src/gallium/drivers/radeonsi/si_perfcounter.c
+++ b/src/gallium/drivers/radeonsi/si_perfcounter.c
@@ -701,7 +701,7 @@ static void si_pc_emit_stop(struct si_context *sctx,
 			  EOP_DST_SEL_MEM, EOP_INT_SEL_NONE,
 			  EOP_DATA_SEL_VALUE_32BIT,
 			  buffer, va, 0, SI_NOT_QUERY);
-	si_cp_wait_mem(sctx, va, 0, 0xffffffff, 0);
+	si_cp_wait_mem(sctx, cs, va, 0, 0xffffffff, WAIT_REG_MEM_EQUAL);
 
 	radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
 	radeon_emit(cs, EVENT_TYPE(V_028A90_PERFCOUNTER_SAMPLE) | EVENT_INDEX(0));
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 4512d2529af..957629e4633 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -1213,7 +1213,7 @@ void si_cp_release_mem(struct si_context *ctx,
 		       struct r600_resource *buf, uint64_t va,
 		       uint32_t new_fence, unsigned query_type);
 unsigned si_cp_write_fence_dwords(struct si_screen *screen);
-void si_cp_wait_mem(struct si_context *ctx,
+void si_cp_wait_mem(struct si_context *ctx, struct radeon_cmdbuf *cs,
 		      uint64_t va, uint32_t ref, uint32_t mask, unsigned flags);
 void si_init_fence_functions(struct si_context *ctx);
 void si_init_screen_fence_functions(struct si_screen *screen);
diff --git a/src/gallium/drivers/radeonsi/si_query.c b/src/gallium/drivers/radeonsi/si_query.c
index 093643bf684..a1563de6970 100644
--- a/src/gallium/drivers/radeonsi/si_query.c
+++ b/src/gallium/drivers/radeonsi/si_query.c
@@ -1576,7 +1576,8 @@ static void si_query_hw_get_result_resource(struct si_context *sctx,
 			va = qbuf->buf->gpu_address + qbuf->results_end - query->result_size;
 			va += params.fence_offset;
 
-			si_cp_wait_mem(sctx, va, 0x80000000, 0x80000000, 0);
+			si_cp_wait_mem(sctx, sctx->gfx_cs, va, 0x80000000,
+				       0x80000000, WAIT_REG_MEM_EQUAL);
 		}
 
 		sctx->b.launch_grid(&sctx->b, &grid);
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index cfd904e621c..6454491457b 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -1056,7 +1056,8 @@ void si_emit_cache_flush(struct si_context *sctx)
 				  EOP_DATA_SEL_VALUE_32BIT,
 				  sctx->wait_mem_scratch, va,
 				  sctx->wait_mem_number, SI_NOT_QUERY);
-		si_cp_wait_mem(sctx, va, sctx->wait_mem_number, 0xffffffff, 0);
+		si_cp_wait_mem(sctx, cs, va, sctx->wait_mem_number, 0xffffffff,
+			       WAIT_REG_MEM_EQUAL);
 	}
 
 	/* Make sure ME is idle (it executes most packets) before continuing.