[llvm] [polly] [SCEV] Introduce SCEVUse, use it instead of const SCEV * (NFCI) (WIP). (PR #91961)
Ramkumar Ramachandra via llvm-commits
llvm-commits at lists.llvm.org
Wed Feb 12 14:45:32 PST 2025
artagnon wrote:
Reverse ping. Do we have any viable alternative to resolve issues pertaining to the global wrap flags in SCEV? Several of them are quite complicated and exotic, but they result in real miscompiles. I think the compile-time regression is the cost to pay for a fundamental oversight in SCEV's design, and I think this patch pay for itself in terms of codegen improvements (although there's a long tail of follow-up patches required). Consider putting up an RFC for this patch, so that we get consensus on the compile-time cost, and get volunteers to help out with follow-ups if it lands? I think you can motivate it by scraping issues for long-standing miscompiles that it would eventually fix.
https://github.com/llvm/llvm-project/pull/91961
More information about the llvm-commits
mailing list