diff options
author | Benoit Steiner <benoit.steiner.goog@gmail.com> | 2016-01-08 13:53:40 -0800 |
---|---|---|
committer | Benoit Steiner <benoit.steiner.goog@gmail.com> | 2016-01-08 13:53:40 -0800 |
commit | 53749ff4152191d2f1bd56090a14f6474fe059c2 (patch) | |
tree | db47f940c3d31de9131dbc5a2c0019d6bf2b3bf7 /unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h | |
parent | 6639b7d6e86ef36f6f78cf51e36efa5a004154eb (diff) |
Prevent nvcc from miscompiling the cuda metakernel. Unfortunately this reintroduces some compulation warnings but it's much better than having to deal with random assertion failures.
Diffstat (limited to 'unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h')
-rw-r--r-- | unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h b/unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h index b046607b9..558d0c83d 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h @@ -115,11 +115,11 @@ struct FullReducer<Self, Op, GpuDevice, Vectorizable> { internal::is_same<typename Self::CoeffReturnType, float>::value; template <typename OutputType> - EIGEN_DEVICE_FUNC static void run(const Self& self, Op& reducer, const GpuDevice& device, OutputType* output) { + static void run(const Self& self, Op& reducer, const GpuDevice& device, OutputType* output) { assert(false && "Should only be called on floats"); } - EIGEN_DEVICE_FUNC static void run(const Self& self, Op& reducer, const GpuDevice& device, float* output) { + static void run(const Self& self, Op& reducer, const GpuDevice& device, float* output) { typedef typename Self::Index Index; const Index num_coeffs = array_prod(self.m_impl.dimensions()); |