Issue 152490
Summary [Headers][X86] Allow SSE2/AVX2/AVX512F/AVX512BW/AVX512DQ integer arithmetic intrinsics to be used in constexpr
Labels new issue
Assignees
Reporter RKSimon
    All of these intrinsics wrap generic implementations which can already be used in constexpr
```
_mm_add_epi8 _mm_add_epi16 _mm_add_epi32 _mm_add_epi64
_mm_sub_epi8 _mm_sub_epi16 _mm_sub_epi32 _mm_sub_epi64
_mm_mullo_epi16 _mm_mullo_epi32

_mm256_add_epi8 _mm256_add_epi16 _mm256_add_epi32 _mm256_add_epi64
_mm256_sub_epi8 _mm256_sub_epi16 _mm256_sub_epi32 _mm256_sub_epi64
_mm256_mullo_epi16 _mm256_mullo_epi32

_mm512_add_epi8 _mm512_add_epi16 _mm512_add_epi32 _mm512_add_epi64
_mm512_sub_epi8 _mm512_sub_epi16 _mm512_sub_epi32 _mm512_sub_epi64
_mm512_mullo_epi16 _mm512_mullo_epi32
_mm_mullo_epi64 _mm256_mullo_epi64 _mm512_mullo_epi64
```
(ignore the AVX512 mask/maskz variants for now)

You can use https://github.com/llvm/llvm-project/pull/152296 for reference
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs

Reply via email to