| Issue |
166975
|
| Summary |
[X86][Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - allow AVX512 kmov intrinsics to be used in constexp
|
| Labels |
good first issue,
backend:X86,
clang:headers,
constexpr,
clang:bytecode
|
| Assignees |
|
| Reporter |
RKSimon
|
These wrap the __builtin_ia32_kmov* builtins, which are just NOPs used to create mask predicates
```
unsigned int _cvtmask16_u32 (__mmask16 a)
unsigned int _cvtmask32_u32 (__mmask32 a)
unsigned __int64 _cvtmask64_u64 (__mmask64 a)
unsigned int _cvtmask8_u32 (__mmask8 a)
__mmask64 _cvtu64_mask64 (unsigned __int64 a)
__mmask16 _cvtu32_mask16 (unsigned int a)
__mmask32 _cvtu32_mask32 (unsigned int a)
__mmask8 _cvtu32_mask8 (unsigned int a)
```
We still need very basic test coverage just to prove they work in constexpr
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
