| Issue |
176880
|
| Summary |
[X86] Failure to fold input shuffle into pclmulqdq instruction
|
| Labels |
backend:X86,
missed-optimization
|
| Assignees |
RKSimon
|
| Reporter |
RKSimon
|
```ll
define <2 x i64> @test_shuffle_pclmulqdq_v2i64(<2 x i64> %a0, <2 x i64> %a1) {
%s0 = shufflevector <2 x i64> %a0, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
%s1 = shufflevector <2 x i64> %a1, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
%clmul = call <2 x i64> @llvm.x86.pclmulqdq(<2 x i64> %s0, <2 x i64> %s1, i8 1)
ret <2 x i64> %clmul
}
```
llc -mcpu=tigerlake
```asm
test_shuffle_pclmulqdq_v2i64: # @test_shuffle_pclmulqdq_v2i64
vpbroadcastq %xmm0, %xmm0
vpshufd $78, %xmm1, %xmm1 # xmm1 = xmm1[2,3,0,1]
vpclmulqdq $1, %xmm1, %xmm0, %xmm0
retq
```
This should fold to:
```asm
test_shuffle_pclmulqdq_v2i64: # @test_shuffle_pclmulqdq_v2i64
vpclmulqdq $16, %xmm1, %xmm0, %xmm0
retq
```
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
