[PATCH] D62363: [X86] Enable intrinsics that convert float and bf16 data to each other

[Simon Pilgrim via Phabricator via cfe-commits]

RKSimon added inline comments.


================
Comment at: lib/Headers/avx512bf16intrin.h:52
+///
+/// \headerfile<x86intrin.h>
+///
----------------
\headerfile <x86intrin.h>


================
Comment at: lib/Headers/avx512bf16intrin.h:54
+///
+/// This intrinsic is lowered to  LLVM IR.
+///
----------------
Please can you write a better description?


================
Comment at: lib/Headers/avx512bf16intrin.h:66
+///
+/// \headerfile<x86intrin.h>
+///
----------------
\headerfile <x86intrin.h>


================
Comment at: lib/Headers/avx512bf16intrin.h:98
+  return (__m512bh)__builtin_ia32_selectw_512(
+      (__mmask32)__U, (__v32hi)_mm512_cvtne2ps_pbh(__A, __B), (__v32hi)__W);
 }
----------------
style/clang-format changes like these should be separated into their own patch


Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D62363/new/

https://reviews.llvm.org/D62363