diff options
Diffstat (limited to 'unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h')
| -rw-r--r-- | unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h b/unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h index 97f4b34b3..71061293b 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h @@ -323,10 +323,12 @@ template <int B, int N, typename S, typename R, typename I> __global__ void FullReductionKernel(R, const S, I, typename S::CoeffReturnType*); +#ifdef EIGEN_HAS_CUDA_FP16 template <typename S, typename R, typename I> __global__ void ReductionInitKernelHalfFloat(R, const S, I, half2*); template <int B, int N, typename S, typename R, typename I> __global__ void FullReductionKernelHalfFloat(R, const S, I, half*, half2*); +#endif template <int NPT, typename S, typename R, typename I> __global__ void InnerReductionKernel(R, const S, I, I, typename S::CoeffReturnType*); @@ -624,8 +626,10 @@ struct TensorEvaluator<const TensorReductionOp<Op, Dims, ArgType>, Device> #endif #if defined(EIGEN_USE_GPU) && defined(__CUDACC__) template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernel(R, const S, I, typename S::CoeffReturnType*); +#ifdef EIGEN_HAS_CUDA_FP16 template <typename S, typename R, typename I> friend void internal::ReductionInitKernelHalfFloat(R, const S, I, half2*); template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernelHalfFloat(R, const S, I, half*, half2*); +#endif template <int NPT, typename S, typename R, typename I> friend void internal::InnerReductionKernel(R, const S, I, I, typename S::CoeffReturnType*); template <int NPT, typename S, typename R, typename I> friend void internal::OuterReductionKernel(R, const S, I, I, typename S::CoeffReturnType*); #endif |