fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-23 04:40:09 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: Fix optimizing branching sequence with s_and_saveexec.

Browse source 
This optimization was broken for two reasons:
- s_and_saveexec has two operands, the copy value and exec
- s_and_saveexec has an exec read, so exec_write_used will always be true
  before we find branch_exec_val_idx

Foz-DB Navi21:
Totals from 31453 (23.31% of 134913) affected shaders:
CodeSize: 204831260 -> 204831156 (-0.00%)
Instrs: 38157117 -> 38157091 (-0.00%)
Latency: 533708882 -> 531211721 (-0.47%); split: -0.47%, +0.00%
InvThroughput: 107088408 -> 106719188 (-0.34%); split: -0.35%, +0.00%
Copies: 2326179 -> 2502490 (+7.58%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18049>
This commit is contained in:
Georg Lehmann 2022-08-14 13:56:30 +02:00 • committed by Marge Bot
parent 5ffc73896f
commit 410eff4d2f
1 changed files with 8 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
src/amd/compiler/aco_ssa_elimination.cpp
View file

@ -544,7 +544,7 @@ eliminate_useless_exec_writes_in_block(ssa_elimination_ctx& ctx, Block& block)
/* For a newly encountered exec write, clear the used flag. */
if (writes_exec) {
if (!logical_end_found && branch_reads_exec && instr->operands.size() == 1) {
if (!logical_end_found && branch_reads_exec && instr->operands.size()) {
/* We are in a branch that jumps according to exec.
* We just found the instruction that copies to exec before the branch.
*/
@ -560,13 +560,16 @@ eliminate_useless_exec_writes_in_block(ssa_elimination_ctx& ctx, Block& block)
}
exec_write_used = false;
}
if (branch_exec_tempid && !exec_write_used && instr->definitions.size() &&
} else if (branch_exec_tempid && instr->definitions.size() &&
instr->definitions[0].tempId() == branch_exec_tempid) {
/* We just found the instruction that produces the exec mask that is copied. */
assert(branch_exec_val_idx == -1);
branch_exec_val_idx = i;
} else if (branch_exec_tempid && branch_exec_val_idx == -1 && needs_exec) {
/* There is an instruction that needs the original exec mask before
* branch_exec_val_idx was found, so we can't optimize the branching sequence. */
branch_exec_copy_idx = -1;
branch_exec_tempid = 0;
}
/* If the current instruction needs exec, mark it as used. */