GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-05-03 01:04:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix remaining CUDA >= 300 checks

Browse Source
This commit is contained in:
Igor Babuschkin 2016-08-18 17:18:30 +01:00
parent 1569a7d7ab
commit 18c67df31c
1 changed files with 7 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h
View File

@ -41,9 +41,6 @@ __device__ EIGEN_ALWAYS_INLINE void atomicReduce(T* output, T accum, R& reducer)
return; return;
} }
} }
#else
assert(0 && "Shouldn't be called on unsupported device");
#endif
} }
else if (sizeof(T) == 8) { else if (sizeof(T) == 8) {
unsigned long long oldval = *reinterpret_cast<unsigned long long*>(output); unsigned long long oldval = *reinterpret_cast<unsigned long long*>(output);
@ -65,6 +62,9 @@ __device__ EIGEN_ALWAYS_INLINE void atomicReduce(T* output, T accum, R& reducer)
else { else {
assert(0 && "Wordsize not supported"); assert(0 && "Wordsize not supported");
} }
#else
assert(0 && "Shouldn't be called on unsupported device");
#endif
} }
// We extend atomicExch to support extra data types // We extend atomicExch to support extra data types
@ -373,6 +373,7 @@ template <int NumPerThread, typename Self,
typename Reducer, typename Index> typename Reducer, typename Index>
__global__ void InnerReductionKernel(Reducer reducer, const Self input, Index num_coeffs_to_reduce, Index num_preserved_coeffs, __global__ void InnerReductionKernel(Reducer reducer, const Self input, Index num_coeffs_to_reduce, Index num_preserved_coeffs,
typename Self::CoeffReturnType* output) { typename Self::CoeffReturnType* output) {
#if __CUDA_ARCH__ >= 300
typedef typename Self::CoeffReturnType Type; typedef typename Self::CoeffReturnType Type;
eigen_assert(blockDim.y == 1); eigen_assert(blockDim.y == 1);
eigen_assert(blockDim.z == 1); eigen_assert(blockDim.z == 1);
@ -433,6 +434,9 @@ __global__ void InnerReductionKernel(Reducer reducer, const Self input, Index nu
} }
} }
} }
#else
assert(0 && "Shouldn't be called on unsupported device");
#endif
} }
#ifdef EIGEN_HAS_CUDA_FP16 #ifdef EIGEN_HAS_CUDA_FP16