diff options
| author | spupyrev <spupyrev@fb.com> | 2022-03-17 09:51:07 -0700 |
|---|---|---|
| committer | spupyrev <spupyrev@fb.com> | 2022-03-22 12:42:42 -0700 |
| commit | 4609f60ebc57d0bd63f97cae11978b96c53204d6 (patch) | |
| tree | 4ba1c90cb6e2893f8b3296bb036a209267fe2b24 /clang/lib/FrontendTool/ExecuteCompilerInvocation.cpp | |
| parent | 03949165cd55958350f3494c92aebfa255086819 (diff) | |
| download | llvm-4609f60ebc57d0bd63f97cae11978b96c53204d6.zip llvm-4609f60ebc57d0bd63f97cae11978b96c53204d6.tar.gz llvm-4609f60ebc57d0bd63f97cae11978b96c53204d6.tar.bz2 | |
[BOLT] Avoid pointless loop rotation
It seems the earlier implementation does not follow the description
in LoopRotationPass.h: It rotates loops even if they are already laid out
correctly. The diff adjusts the behaviour.
Given that the impact of LoopInversionPass is minor, this change won't
yield significant perf differences. Tested on clang-10: there seems to be a
0.1%-0.3% cpu win and a small reduction of branch misses.
**Before:**
BOLT-INFO: 120 Functions were reordered by LoopInversionPass
**After:**
BOLT-INFO: 79 Functions were reordered by LoopInversionPass
Reviewed By: yota9
Differential Revision: https://reviews.llvm.org/D121921
Diffstat (limited to 'clang/lib/FrontendTool/ExecuteCompilerInvocation.cpp')
0 files changed, 0 insertions, 0 deletions