| Issue |
142711
|
| Summary |
Missed Optimization: fold fcmp ord + select before fcmp oeq (x, C) into a single fcmp oeq (x, C)
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
Cancelll
|
```llvm
define i1 @src(ptr %0) {
%2 = getelementptr inbounds nuw i8, ptr %0, i64 48
%3 = load double, ptr %2, align 8
%4 = getelementptr inbounds nuw i8, ptr %0, i64 24
%5 = load double, ptr %4, align 8
%6 = fadd double %3, %5
%7 = fcmp ord double %6, 0.000000e+00
%8 = select i1 %7, double %6, double 0.000000e+00
%9 = fcmp oeq double %8, 1.000000e+00
ret i1 %9
}
define i1 @tgt(ptr %0) {
%2 = getelementptr inbounds nuw i8, ptr %0, i64 48
%3 = load double, ptr %2, align 8
%4 = getelementptr inbounds nuw i8, ptr %0, i64 24
%5 = load double, ptr %4, align 8
%6 = fadd double %3, %5
%7 = fcmp oeq double %6, 1.000000e+00
ret i1 %7
}
```
Alive2: https://alive2.llvm.org/ce/z/2T-D5v
Found this pattern in https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/typst-rs/optimized/4m3ebbqd1xx21e5m.ll
godbolt: https://godbolt.org/z/vMhzj9eP3
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
