[PATCH] D132342: [X86][AVX512FP16] Relax limitation to AVX512FP16 intrinsics. NFCI

Simon Pilgrim via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon Aug 22 09:33:43 PDT 2022


RKSimon added inline comments.


================
Comment at: clang/lib/Headers/avx512fp16intrin.h:13
 
+#ifdef __SSE2__
+
----------------
pengfei wrote:
> RKSimon wrote:
> > Doesn't this have to be the general case like in other places in the headers?
> > ```
> > #if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
> >     defined(__SSE2__)
> > ```
> I don't think so. `!(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) || defined(__AVX512FP16__)` has already been checked in immintrin.h
> 
> So if we are preprocessing here, the above condition must be true. Since `AVX512FP16` implies `SSE2`, the new condition is constant ture.
got it - cheers


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D132342/new/

https://reviews.llvm.org/D132342



More information about the cfe-commits mailing list