freedreno/a6xx: Drop the WFI in constant uploads.

Now that the bin vs render constlen is fixed, we can skip these waits.

Improves webgl aquarium performance at 10k fish from 27fps to 33.
Some highlights from renderdoc-traces:

traces/android/minecraft.rdc:             +17.1% +/-   3.4%
traces/glmark2/ideas-speed=duration.rdc:  +11.6% +/-   2.4%
traces/android/candycrush.rdc:            +5.4%  +/-   1.1%
traces/android/clashofclans.rdc:          +4.4%  +/-   1.3%

Reviewed-by: Rob Clark <robdclark@gmail.com>

diff --git a/src/gallium/drivers/freedreno/a6xx/fd6_emit.c b/src/gallium/drivers/freedreno/a6xx/fd6_emit.c
index 5e8a65c9dd73159fc5fba37a13b333ac35396b28..e75bb369bba85b6884ffcf1d1d8431d58874d0d8 100644 (file)
--- a/src/gallium/drivers/freedreno/a6xx/fd6_emit.c
+++ b/src/gallium/drivers/freedreno/a6xx/fd6_emit.c
@@ -934,7 +934,6 @@ fd6_emit_state(struct fd_ringbuffer *ring, struct fd6_emit *emit)
                struct fd_ringbuffer *vsconstobj = fd_submit_new_ringbuffer(
                                ctx->batch->submit, 0x1000, FD_RINGBUFFER_STREAMING);
 
-               OUT_WFI5(vsconstobj);
                ir3_emit_vs_consts(vp, vsconstobj, ctx, emit->info);
                fd6_emit_add_group(emit, vsconstobj, FD6_GROUP_VS_CONST, 0x7);
                fd_ringbuffer_del(vsconstobj);
@@ -944,7 +943,6 @@ fd6_emit_state(struct fd_ringbuffer *ring, struct fd6_emit *emit)
                struct fd_ringbuffer *fsconstobj = fd_submit_new_ringbuffer(
                                ctx->batch->submit, 0x1000, FD_RINGBUFFER_STREAMING);
 
-               OUT_WFI5(fsconstobj);
                ir3_emit_fs_consts(fp, fsconstobj, ctx);
                fd6_emit_add_group(emit, fsconstobj, FD6_GROUP_FS_CONST, 0x6);
                fd_ringbuffer_del(fsconstobj);