[llvm-branch-commits] [llvm] [CodeGen] Limit number of analyzed predecessors (PR #142584)

[via llvm-branch-commits]

https://github.com/spupyrev approved this pull request.

I don't have objections but maybe wait what others say.

My assumption is that this is really needed to speed up some artificial IRs which are not (or rarely) observed in practice; for such cases it doesn't matter too much what this pass produces. However, if that's not the case and we do care about the result, then we probably should just optimize the function, instead of adding arbitrary thresholds.

https://github.com/llvm/llvm-project/pull/142584