[libc-commits] [libc] [libc][math][c23] Add exp2m1f16 C23 math function (PR #105690)

[via libc-commits]

lntue wrote:

> > > That doesn't work on the GPU targets yet https://godbolt.org/z/6n3svs1ne. But I guess NVPTX doesn't have `fenv` access and I never got around to merging the AMDGPU version.
> > 
> > 
> > Correct, but you shouldn't design around working around unimplemented features. Full GPU strictfp support isn't coming anytime soon. It is still wrong to rely on this check without it
> 
> The other really broken thing is the -fexperimental-strict-floating-point flag doesn't do anything other than for the hardcoded targets in clang, making it useless as a bringup aid

We can omit the gpu targets for now.  I'm going to add `LIBC_MATH_DEFAULT_ROUNDING_ONLY` option to cleanly omit all the directional rounding codes for targets requesting it.

https://github.com/llvm/llvm-project/pull/105690