| Issue |
167059
|
| Summary |
Missed Optimization: Fold smin(32 - a, b - a) > 0 → (a < 32) && (a < b)
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
Leo0506
|
```
define i1 @src(i16 %arg0, i16 %arg1) {
%v0 = sub nsw i16 %arg1, %arg0
%v1 = sub nsw i16 32, %arg0
%v2 = call i16 @llvm.smin.i16(i16 %v1, i16 %v0)
%v3 = icmp sgt i16 %v2, 0
ret i1 %v3
}
define i1 @tgt(i16 %arg0, i16 %arg1) {
%v0 = icmp slt i16 %arg0, 32
%v1 = icmp slt i16 %arg0, %arg1
%v2 = and i1 %v0, %v1
ret i1 %v2
}
```
alive2: https://alive2.llvm.org/ce/z/zwo7PE
godbolt: https://godbolt.org/z/7eeWc1fx4
Pattern found in: https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/ffmpeg/optimized/pngenc.ll
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
