intel/brw: Eliminate top-level FIND_LIVE_CHANNEL & BROADCAST once

brw_fs_opt_eliminate_find_live_channel eliminates FIND_LIVE_CHANNEL
outside of control flow.  None of our optimization passes generate
additional cases of that instruction, so once it's gone, we shouldn't
ever have to run the pass again.  Moving it out of the loop should
save a bit of CPU time.

While we're at it, also clean adjacent BROADCAST instructions that
consume the result of our FIND_LIVE_CHANNEL.  Without this, we have
to perform copy propagation to get the MOV 0 immediate into the
BROADCAST, then algebraic to turn it into a MOV, which enables more
copy propagation...not to mention CSE gets involved.  Since this
FIND_LIVE_CHANNEL + BROADCAST pattern from emit_uniformize() is
really common, and it's trivial to clean up, we can do that.  This
lets the initial copy prop in the loop see MOV instead of BROADCAST.

Zero impact on fossil-db, but less work in the optimization loop.

Together with the previous patches, this cuts compile time in
Borderlands 3 on Alchemist by -1.38539% +/- 0.1632% (n = 24).

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28286>

1 files changed, 22 insertions, 1 deletions

diff --git a/src/intel/compiler/brw_fs_opt.cpp b/src/intel/compiler/brw_fs_opt.cpp
index 7835d060084..e9abeef4098 100644
--- a/src/intel/compiler/brw_fs_opt.cpp
+++ b/src/intel/compiler/brw_fs_opt.cpp
@@ -56,6 +56,8 @@ brw_fs_optimize(fs_visitor &s)
 
    OPT(brw_fs_opt_remove_extra_rounding_modes);
 
+   OPT(brw_fs_opt_eliminate_find_live_channel);
+
    do {
       progress = false;
       pass_num = 0;
@@ -71,7 +73,6 @@ brw_fs_optimize(fs_visitor &s)
       OPT(brw_fs_opt_dead_control_flow_eliminate);
       OPT(brw_fs_opt_saturate_propagation);
       OPT(brw_fs_opt_register_coalesce);
-      OPT(brw_fs_opt_eliminate_find_live_channel);
 
       OPT(brw_fs_opt_compact_virtual_grfs);
    } while (progress);
@@ -430,6 +431,26 @@ brw_fs_opt_eliminate_find_live_channel(fs_visitor &s)
             inst->sources = 1;
             inst->force_writemask_all = true;
             progress = true;
+
+            /* emit_uniformize() frequently emits FIND_LIVE_CHANNEL paired
+             * with a BROADCAST.  Save some work for opt_copy_propagation
+             * and opt_algebraic by trivially cleaning up both together.
+             */
+            assert(!inst->next->is_tail_sentinel());
+            fs_inst *bcast = (fs_inst *) inst->next;
+
+            /* Ignore stride when comparing */
+            if (bcast->opcode == SHADER_OPCODE_BROADCAST &&
+                inst->dst.file == VGRF &&
+                inst->dst.file == bcast->src[1].file &&
+                inst->dst.nr == bcast->src[1].nr &&
+                inst->dst.offset == bcast->src[1].offset) {
+               bcast->opcode = BRW_OPCODE_MOV;
+               if (!is_uniform(bcast->src[0]))
+                  bcast->src[0] = component(bcast->src[0], 0);
+               bcast->sources = 1;
+               bcast->force_writemask_all = true;
+            }
          }
          break;