| Issue |
167090
|
| Summary |
Missed Optimization: Fold chained umin+add clamps into a single cap
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
Leo0506
|
```
define i32 @src(i32 %i16arg0) {
%v0 = tail call i32 @llvm.umin.i32(i32 %i16arg0, i32 16)
%v1 = add nuw nsw i32 %v0, 16
%v2 = tail call i32 @llvm.umin.i32(i32 %i16arg0, i32 %v1)
%v3 = add nuw nsw i32 %v2, 16
%v4 = tail call i32 @llvm.umin.i32(i32 %i16arg0, i32 %v3)
ret i32 %v4
}
define i32 @tgt(i32 %i16arg0) {
%v0 = tail call i32 @llvm.umin.i32(i32 %i16arg0, i32 48)
ret i32 %v0
}
```
alive2: https://alive2.llvm.org/ce/z/XB4jkJ
godbolt: https://godbolt.org/z/j975765ce
Pattern found in: https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/ffmpeg/optimized/qdmc.ll
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
