[PATCH] D53486: [libcxx] Only define __libcpp_is_floating_point<_Float16> for Clang

[Louis Dionne via Phabricator]

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

After considering the different options and the problems with `__fp16`, we think it would be better to revert r333103 altogether. `__fp16` is not well-behaved on x86 (can't pass it as a function parameter, for example), and so trying to "fully" support it (in the sense of "floating point types") would be messy. Furthermore, the original use case can be solved with a much narrower workaround.

In the longer term, we look forward to p0192 solving this problem and paving the way for "implementation defined floating point types." This would probably require an additional paper.


Repository:
  rCXX libc++

https://reviews.llvm.org/D53486