diff options
| author |  Rasmus Larsen <rmlarsen@google.com> | 2019-05-24 16:05:31 +0000 |
|---|---|---|
| committer | Rasmus Larsen <rmlarsen@google.com> | 2019-05-24 16:05:31 +0000 |
| commit | 36e0a2b93fd4cd4490afe1f271d617831074c809 (patch) | |
| tree | 262bf3089f137ff65c1b838a081b99d529fab933 | |
| parent | c5019f722bc76537ec56947e75fc4169deceffcd (diff) | |
| parent | 2c389301610a0c60a5760fec7a362803a680ff60 (diff) | |
Merged in deven-amd/eigen-hip-fix-190524 (pull request PR-649)
fix for HIP build errors that were introduced by a commit earlier this week
| -rw-r--r-- | Eigen/src/Core/arch/GPU/Half.h | 2 | ||||
| -rw-r--r-- | Eigen/src/Core/arch/GPU/PacketMathHalf.h | 2 |
2 files changed, 2 insertions, 2 deletions
diff --git a/Eigen/src/Core/arch/GPU/Half.h b/Eigen/src/Core/arch/GPU/Half.h index 7f1d7e5b8..5affe8b92 100644 --- a/Eigen/src/Core/arch/GPU/Half.h +++ b/Eigen/src/Core/arch/GPU/Half.h @@ -318,7 +318,7 @@ EIGEN_STRONG_INLINE __device__ bool operator >= (const half& a, const half& b) { // We need to provide emulated *host-side* FP16 operators for clang. #pragma push_macro("EIGEN_DEVICE_FUNC") #undef EIGEN_DEVICE_FUNC -#if defined(EIGEN_HAS_CUDA_FP16) +#if defined(EIGEN_HAS_GPU_FP16) #define EIGEN_DEVICE_FUNC __host__ #else // both host and device need emulated ops. #define EIGEN_DEVICE_FUNC __host__ __device__ diff --git a/Eigen/src/Core/arch/GPU/PacketMathHalf.h b/Eigen/src/Core/arch/GPU/PacketMathHalf.h index 3da8c1f3e..723506c8b 100644 --- a/Eigen/src/Core/arch/GPU/PacketMathHalf.h +++ b/Eigen/src/Core/arch/GPU/PacketMathHalf.h @@ -46,7 +46,7 @@ template<> struct packet_traits<Eigen::half> : default_packet_traits template<> struct unpacket_traits<half2> { typedef Eigen::half type; enum {size=2, alignment=Aligned16, vectorizable=true, masked_load_available=false, masked_store_available=false}; typedef half2 half; }; template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE half2 pset1<half2>(const Eigen::half& from) { -#if !defined(EIGEN_CUDA_ARCH) +#if !defined(EIGEN_CUDA_ARCH) && !defined(EIGEN_HIP_DEVICE_COMPILE) half2 r; r.x = from; r.y = from; |