[PATCH] D53486: [libcxx] Only define __libcpp_is_floating_point<_Float16> for Clang
Marshall Clow via Phabricator
reviews at reviews.llvm.org
Wed Oct 24 08:30:08 PDT 2018
mclow.lists added a comment.
In https://reviews.llvm.org/D53486#1274294, @ldionne wrote:
> Question: @mclow.lists if `__fp16` supported all the proper operations required to be "a floating point type", would there be pushback against defining `is_floating_point` for `__fp16` as a library extension?
I can only speak for myself here - not other people, but I have two concerns; one rooted in the standard, the other more personal.
- The standard makes no provisions for implementations to define additional floating point types. (Contrast this with integral types, where the standard goes on an on about implementation-defined integral types)
- Adding support for another FP type is a lot of work. (again, refer to P0192R4) There's a lot of other stuff that I'd rather be doing.
Repository:
rCXX libc++
https://reviews.llvm.org/D53486
More information about the libcxx-commits
mailing list