[PATCH] D115611: [X86][BF16] delete `typedef unsigned short __bfloat16`

[Phoebe Wang via Phabricator via cfe-commits]

pengfei requested changes to this revision.
pengfei added a comment.
This revision now requires changes to proceed.

I would suggest we drop the change. Sorry for my fickleness :(



================
Comment at: clang/lib/Headers/avx512bf16intrin.h:32
 /// \param __A
 ///    A bfloat data.
 /// \returns A float data whose sign field and exponent field keep unchanged,
----------------
Sorry, when I reviewed the doxygen, I found a new problem. If we want to use `unsigned short`, we may need to change here too. But `unsigned short` is not clear to user since they actually want to convert a bfloat type instead of integer.
On the other hand, the double underscore naming conversion is reserved for compiler use and we have already used lots of these terminologies for X86 intrinsics. A much similar one is `__mask16`. So I think using `__bfloat16` here is consistent with the existing types we have defined.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D115611/new/

https://reviews.llvm.org/D115611