freedreno/ir3/postsched: try to avoid (sy) syncs

Similar to avoidance of `(ss)` syncs, it turns out to be helpful to
avoid `(sy)` syncs as well.  This helps us turn an tex, (sy)alu, tex,
(sy)alu sequence into tex, tex, (sy)alu, alu, which is a big win in
gfxbench gl_fill2.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4923>

diff --git a/src/freedreno/ir3/ir3_postsched.c b/src/freedreno/ir3/ir3_postsched.c
index 4535459efcf623a994bc7e5eeee8965acee4ad39..238d937c8053121992c1704c471564267b749933 100644 (file)
--- a/src/freedreno/ir3/ir3_postsched.c
+++ b/src/freedreno/ir3/ir3_postsched.c
@@ -62,6 +62,7 @@ struct ir3_postsched_ctx {
        bool error;
 
        int sfu_delay;
+       int tex_delay;
 };
 
 struct ir3_postsched_node {
@@ -92,6 +93,12 @@ schedule(struct ir3_postsched_ctx *ctx, struct ir3_instruction *instr)
 
        list_addtail(&instr->node, &instr->block->instr_list);
 
+       struct ir3_postsched_node *n = instr->data;
+       dag_prune_head(ctx->dag, &n->dag);
+
+       if (is_meta(instr) && (instr->opc != OPC_META_TEX_PREFETCH))
+               return;
+
        if (is_sfu(instr)) {
                ctx->sfu_delay = 8;
        } else if (check_src_cond(instr, is_sfu)) {
@@ -100,8 +107,13 @@ schedule(struct ir3_postsched_ctx *ctx, struct ir3_instruction *instr)
                ctx->sfu_delay--;
        }
 
-       struct ir3_postsched_node *n = instr->data;
-       dag_prune_head(ctx->dag, &n->dag);
+       if (is_tex_or_prefetch(instr)) {
+               ctx->tex_delay = 10;
+       } else if (check_src_cond(instr, is_tex_or_prefetch)) {
+               ctx->tex_delay = 0;
+       } else if (ctx->tex_delay > 0) {
+               ctx->tex_delay--;
+       }
 }
 
 static void
@@ -135,6 +147,11 @@ would_sync(struct ir3_postsched_ctx *ctx, struct ir3_instruction *instr)
                        return true;
        }
 
+       if (ctx->tex_delay) {
+               if (check_src_cond(instr, is_tex_or_prefetch))
+                       return true;
+       }
+
        return false;
 }