r600/sb: add gcm support to avoid clause between lds read/queue read

You have to schedule LDS_READ_RET _, x and MOV reg, LDS_OQ_A_POP
in the same basic block/clause. This makes sure once we've issues
and MOV we don't add another block until we balance it with an
LDS read.

Acked-By: Roland Scheidegger <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>

2 files changed, 17 insertions, 2 deletions

diff --git a/src/gallium/drivers/r600/sb/sb_gcm.cpp b/src/gallium/drivers/r600/sb/sb_gcm.cpp
index fbebe3427d4..7776a10fc86 100644
--- a/src/gallium/drivers/r600/sb/sb_gcm.cpp
+++ b/src/gallium/drivers/r600/sb/sb_gcm.cpp
@@ -366,6 +366,9 @@ void gcm::bu_sched_bb(bb_node* bb) {
 				continue;
 			}
 
+			if (sq != SQ_ALU && outstanding_lds_oq)
+				continue;
+
 			if (!bu_ready_next[sq].empty())
 				bu_ready[sq].splice(bu_ready[sq].end(), bu_ready_next[sq]);
 
@@ -388,7 +391,7 @@ void gcm::bu_sched_bb(bb_node* bb) {
 				}
 
 				// simple heuristic to limit register pressure,
-				if (sq == SQ_ALU && live_count > rp_threshold &&
+				if (sq == SQ_ALU && live_count > rp_threshold && !outstanding_lds_oq &&
 						(!bu_ready[SQ_TEX].empty() ||
 						 !bu_ready[SQ_VTX].empty() ||
 						 !bu_ready_next[SQ_TEX].empty() ||
@@ -423,6 +426,12 @@ void gcm::bu_sched_bb(bb_node* bb) {
 						check_alu_ready_count(24))
 					break;
 
+
+				if (sq == SQ_ALU && n->consumes_lds_oq() &&
+				    (bu_ready[SQ_TEX].size() || bu_ready[SQ_VTX].size() || bu_ready[SQ_GDS].size())) {
+					GCM_DUMP( sblog << "switching scheduling due to lds op\n"; );
+					break;
+				}
 				bu_ready[sq].pop_front();
 
 				if (sq != SQ_CF) {
@@ -513,6 +522,10 @@ void gcm::bu_schedule(container_node* c, node* n) {
 
 	assert(op_map[n].bottom_bb == bu_bb);
 
+	if (n->produces_lds_oq())
+		outstanding_lds_oq--;
+	if (n->consumes_lds_oq())
+		outstanding_lds_oq++;
 	bu_release_defs(n->src, true);
 	bu_release_defs(n->dst, false);
 
diff --git a/src/gallium/drivers/r600/sb/sb_pass.h b/src/gallium/drivers/r600/sb/sb_pass.h
index e878f8c70ca..b5818039c2e 100644
--- a/src/gallium/drivers/r600/sb/sb_pass.h
+++ b/src/gallium/drivers/r600/sb/sb_pass.h
@@ -223,6 +223,7 @@ class gcm : public pass {
 	sched_queue ready;
 	sched_queue ready_above;
 
+	unsigned outstanding_lds_oq;
 	container_node pending;
 
 	struct op_info {
@@ -263,7 +264,8 @@ public:
 
 	gcm(shader &sh) : pass(sh),
 		bu_ready(), bu_ready_next(), bu_ready_early(),
-		ready(), op_map(), uses(), nuc_stk(1), ucs_level(),
+		ready(), outstanding_lds_oq(),
+		op_map(), uses(), nuc_stk(1), ucs_level(),
 		bu_bb(), pending_defs(), pending_nodes(), cur_sq(),
 		live(), live_count(), pending_exec_mask_update() {}