[llvm] [InstCombine] Relax the one-use constraints for `icmp pred (binop X, Z), (binop Y, Z)` (PR #76384)

Yingwei Zheng via llvm-commits llvm-commits at lists.llvm.org
Tue Dec 26 05:34:21 PST 2023


dtcxzyw wrote:

> This patch breaks the alias analysis and causes some regressions. We should do the simplification when at least one binop is one-use.

Done. All regressions have been fixed except for https://github.com/dtcxzyw/llvm-opt-benchmark/pull/49#issuecomment-1869546202. I will post a patch for clang later.


https://github.com/llvm/llvm-project/pull/76384


More information about the llvm-commits mailing list