Currently, whether we fold a shift or extension operation into a compare depends on whether the foldable operand happens to land on the RHS. Fix it so we consistently put foldable operations on the RHS.
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
Comment Actions
Found an issue in test/CodeGen/AArch64/and-mask-removal.ll I somehow missed; I'll post a revised version soon.