[llvm] [Bitcode] Fix nondeterministic phi instruction order (PR #151006)
Nikita Popov via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 30 09:13:24 PDT 2025
nikic wrote:
> I'm a little concerned that the comments suggest that the `OM.lookup(` calls are supposed to succeed... but they must be failing if two entries are comparing equal. So maybe this algorithm isn't working the way it's supposed to. @nikic, do you remember how this works?
The preceding loop only puts entries into List which have an OrderMap entry. So these lookups are indeed not supposed to fail.
TBH it's not immediately obvious under which circumstances the stable sort would make a difference here.
https://github.com/llvm/llvm-project/pull/151006
More information about the llvm-commits
mailing list