[PATCH] D99249: [PassManager] Run additional LICM before LoopRotate

[Nikita Popov via Phabricator via llvm-commits]

nikic added a comment.

In D99249#2661952 <https://reviews.llvm.org/D99249#2661952>, @asbirlea wrote:

> Most performance results look ok. There are a few that are mixed (+ & -), one regression that's still investigated that so far seems architecture specific and not related to this patch, a couple of noticeable improvements and a few marginal improvements.
> Please note I have only tested the new pass manager.
> So, green light from my side assuming the other reviewers are also on board with this.

Thanks for running the tests. Sounds like the effect here is relatively minor, but at least it's not negative.

I think at this point there is a pretty clear motivation for why we want to run LICM before LoopRotate, but the motivation for the LICM-LoopRotate-LICM sequence is still somewhat murky to me. Do we have any PhaseOrdering tests that would regress without the second LICM run?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99249/new/

https://reviews.llvm.org/D99249