This is tracking issue to refactor float16 operations math functions currently implemented in LLVM libc to be header-only. part of: #147386 @bassiounix