[PATCH] D104569: [SimplifyCFG] Fix SimplifyBranchOnICmpChain to be undef/poison safe.

Hyeongyu Kim via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Jun 22 18:57:31 PDT 2021


hyeongyukim added a comment.

In D104569#2831268 <https://reviews.llvm.org/D104569#2831268>, @aqjune wrote:

> IIUC, the result of `isGuaranteedNotToBePoison` is suboptimal because `mustTriggerUB` does not consider `br poison` as UB.
>
> Could you try D92739 <https://reviews.llvm.org/D92739> and see whether the unnecessary freeze instructions disappear?
>
> BTW, it is possible that the redundant freeze instructions actually don't matter; compiling benchmarks and seeing assembly diffs will be helpful for determining this.

When I applied D92739 <https://reviews.llvm.org/D92739>, redundant freeze instructions was removed.
Also, a small difference was found when comparing assembly diffs of llvm testsuite. 
Of the 5,200 examples, 121 changes were observed.F17548189: diff_list.txt <https://reviews.llvm.org/F17548189>

Example of assembly diff is below. F17548121: diff_example.txt <https://reviews.llvm.org/F17548121>

Although it is a small difference, I think the difference that D92739 <https://reviews.llvm.org/D92739> makes is meaningful.
It seems better to solve D92739 <https://reviews.llvm.org/D92739> first and then solve this issue again. Is it okay to solve D92739 <https://reviews.llvm.org/D92739> first?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D104569/new/

https://reviews.llvm.org/D104569



More information about the llvm-commits mailing list