[all-commits] [llvm/llvm-project] 4609f6: [BOLT] Avoid pointless loop rotation

[spupyrev via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 4609f60ebc57d0bd63f97cae11978b96c53204d6
      https://github.com/llvm/llvm-project/commit/4609f60ebc57d0bd63f97cae11978b96c53204d6
  Author: spupyrev <spupyrev at fb.com>
  Date:   2022-03-22 (Tue, 22 Mar 2022)

  Changed paths:
    M bolt/lib/Passes/LoopInversionPass.cpp
    M bolt/test/X86/loop-inversion-pass.s

  Log Message:
  -----------
  [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