| Issue |
79266
|
| Summary |
[InstCombine] Missed optimization: fold `((select C1, C2) | A) ^ C3` to `(select C1^C3, C2^C3) | A)`
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
XChy
|
Alive2 proof: https://alive2.llvm.org/ce/z/oEmX9u
### Motivating example
```llvm
define i32 @src(i32 %a, i1 %c) {
entry:
%s = select i1 %c, i32 0, i32 4
%shl = shl i32 %a, 4
%or = or disjoint i32 %s, %shl
%xor = xor i32 %or, 4
ret i32 %xor
}
```
can be folded to
```llvm
define i32 @tgt(i32 %a, i1 %c) {
entry:
%s = select i1 %c, i32 4, i32 0
%shl = shl i32 %a, 4
%or = or disjoint i32 %s, %shl
ret i32 %or
}
```
Since `or disjoint` here can be regarded as `xor`, `((select C1, C2) | A) ^ C3` can be `(select C1, C2) ^ C3 ^ A)` and then we get `(select C1^C3, C2^C3) | A)`.
### Real-world motivation
This snippet of IR is derived from [zstd/lib/compress/zstd_compress_literals.c@ZSTD_compressRleLiteralsBlock](https://github.com/facebook/zstd/blob/1a860c8737f42b60e4e80b70d987df8583c977b8/lib/compress/zstd_compress_literals.c#L213) (after O3 pipeline).
The example above is a reduced version. If you're interested in the original suboptimal IR and optimal IR, contact me to get them, please.
**Let me know if you can confirm that it's an optimization opportunity, thanks.**
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
