diff options
| author |  Benoit Steiner <benoit.steiner.goog@gmail.com> | 2016-06-09 08:22:27 -0700 |
|---|---|---|
| committer | Benoit Steiner <benoit.steiner.goog@gmail.com> | 2016-06-09 08:22:27 -0700 |
| commit | aa33446dace833fbf06632e586c80119b3d8ac11 (patch) | |
| tree | d7bc59330e69e19de37da9fffc64016ee0febec0 /unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h | |
| parent | 15890c304edbccedc8a989468ed3fc475f428059 (diff) | |
Improved support for vectorization of 16-bit floats
Diffstat (limited to 'unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h')
| -rw-r--r-- | unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h b/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h index b1645d56f..82a905a65 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h @@ -55,6 +55,11 @@ struct PacketType { // For CUDA packet types when using a GpuDevice #if defined(EIGEN_USE_GPU) && defined(__CUDACC__) template <> + struct PacketType<half, GpuDevice> { + typedef half2 type; + static const int size = 2; + }; +template <> struct PacketType<float, GpuDevice> { typedef float4 type; static const int size = 4; |