| Issue |
166890
|
| Summary |
Missed Optimization: fold (arg1 + 4095 + arg0) u< 4096 (with nuw/nsw + same-sign) into a zero-sum test icmp ult (arg0 + arg1), 1
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
Leo0506
|
```
define i1 @src(i64 %arg0, i64 %arg1) {
%v0 = add nuw nsw i64 %arg1, 4095
%v1 = add nuw nsw i64 %v0, %arg0
%v2 = icmp samesign ult i64 %v1, 4096
ret i1 %v2
}
define i1 @tgt(i64 %arg0, i64 %arg1) {
%v0 = add nuw nsw i64 %arg0, %arg1
%v1 = icmp ult i64 %v0, 1
ret i1 %v1
}
```
alive2: https://alive2.llvm.org/ce/z/PNCxyX
godbolt: https://godbolt.org/z/eWvbejrG5
Pattern found in: https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/linux/optimized/iov_iter.ll
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
