https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119917
--- Comment #3 from Jin Haobo <a1343922569 at outlook dot com> --- Thank you for your confirmation. I understand that the char/short cases are treated differently due to integer promotion. While I acknowledge that 8/16-bit unsigned integers undergo integer promotion which changes the exact semantics of the operation (as shown in your example with the char case), the mathematical result should still be equivalent for all valid inputs. The expression `a > 0xFF / b` should yield the same boolean result as checking if `a * b > 0xFF`, so a similar optimization strategy could theoretically be applied. However, even if there are concerns about optimizing the 8/16-bit cases due to promotion semantics, the 32/64-bit cases should clearly be optimized. In these cases: 1. The explicit zero check (`if (b == 0) return false;`) already guarantees that division-by-zero cannot occur. 2. There is no integer promotion affecting the semantics. 3. The compiler already recognizes the pattern as a multiplication overflow check for these sizes (as evidenced by the `mul` instruction followed by `seto`). The redundant zero check in the assembly output for 32/64-bit cases is unnecessary and impacts performance, especially in performance-critical code paths. I hope this specific optimization for int and long long types can be implemented even if the smaller types require further analysis. Thank you for your understanding!
