[PATCH] D100794: [HIP] Support overloaded math functions for hipRTC

Yaxun Liu via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Thu Apr 22 12:33:03 PDT 2021


yaxunl added inline comments.


================
Comment at: clang/lib/Headers/__clang_hip_cmath.h:309
+};
+template <> struct is_arithmetic<_Float16> {
+  enum { value = 1 };
----------------
This changes the overloading behavior of math functions with _Float16 arguments since previously we use std::numeric_limits<T>::is_specialized to determine whether a type is an arithmetic, which is not specialized for _Float16.

To make this patch NFC for non-hipRTC, I would like to split this part to another patch. This facilitates debugging if issues arise


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D100794/new/

https://reviews.llvm.org/D100794



More information about the cfe-commits mailing list