| Issue |
143636
|
| Summary |
Missed optimization: fold ceil(x/16) == 0 to x == 0
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
Cancelll
|
```llvm
define i1 @src(i32 %arg0) {
%1 = add i32 %arg0, 1
%2 = lshr i32 %1, 4
%3 = and i32 %1, 15
%4 = icmp ne i32 %3, 0
%5 = zext i1 %4 to i32
%6 = add nuw nsw i32 %2, %5
%7 = icmp eq i32 %6, 0
ret i1 %7
}
define i1 @tgt(i32 %arg0) {
%5 = icmp eq i32 %arg0, -1
ret i1 %5
}
```
(x >> 4) + ((x & 0xF) ≠ 0) is equivalent to ceil(x/16).
ceil(x/16) equals to 0 only when x is 0.
Alive2: https://alive2.llvm.org/ce/z/E85DRW
godbolt: https://godbolt.org/z/Y5frz97Mh
The pattern is found in projects including wasmi, coreutils, ruff. An example:
https://github.com/dtcxzyw/llvm-opt-benchmark/blob/013f9638352b38eb8f26daddafeeb0d87fc6510b/bench/coreutils-rs/optimized/50z74lntrf1cjp27.ll#L1210-L1216
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
