> Presumably the same is true for LLVM IR I think it depends on the flags the multiplication operation is emitted with; namely, the nsw and nuw (no signed/unsi

> Presumably the same is true for LLVM IR I think it depends on the flags the multiplication operation is emitted with; namely, the nsw and nuw (no signed/unsigned wrap) would denote whether the optimization LLVM does is correct. If rust emits those flags then that would be a Rust bug; if Rust does not, then it might be an LLVM bug (I'm not sure what LLVM's semantics are for regular imul without flags).