GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-14 04:35:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated the cuda tests to use the new GpuDevice constructor

Browse Source
This commit is contained in:
Benoit Steiner 2015-07-15 12:39:26 -07:00
parent e892524efe
commit 8315e025e1
3 changed files with 4 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
unsupported/test/cxx11_tensor_contract_cuda.cpp
View File

@ -52,8 +52,7 @@ static void test_cuda_contraction(int m_size, int k_size, int n_size)
cudaMemcpy(d_t_left, t_left.data(), t_left_bytes, cudaMemcpyHostToDevice); cudaMemcpy(d_t_left, t_left.data(), t_left_bytes, cudaMemcpyHostToDevice);
cudaMemcpy(d_t_right, t_right.data(), t_right_bytes, cudaMemcpyHostToDevice); cudaMemcpy(d_t_right, t_right.data(), t_right_bytes, cudaMemcpyHostToDevice);
cudaStream_t stream; Eigen::CudaStreamDevice stream;
assert(cudaStreamCreate(&stream) == cudaSuccess);
Eigen::GpuDevice gpu_device(&stream); Eigen::GpuDevice gpu_device(&stream);
Eigen::TensorMap<Eigen::Tensor<float, 2, DataLayout> > Eigen::TensorMap<Eigen::Tensor<float, 2, DataLayout> >

5
unsupported/test/cxx11_tensor_device.cpp
View File

@ -77,14 +77,11 @@ struct GPUContext {
assert(cudaMalloc((void**)(&kernel_3d_), 8*sizeof(float)) == cudaSuccess); assert(cudaMalloc((void**)(&kernel_3d_), 8*sizeof(float)) == cudaSuccess);
float kernel_3d_val[] = {3.14f, -1.0f, 2.7f, -0.3f, 0.2f, -0.7f, 7.0f, -0.5f}; float kernel_3d_val[] = {3.14f, -1.0f, 2.7f, -0.3f, 0.2f, -0.7f, 7.0f, -0.5f};
assert(cudaMemcpy(kernel_3d_, kernel_3d_val, 8*sizeof(float), cudaMemcpyHostToDevice) == cudaSuccess); assert(cudaMemcpy(kernel_3d_, kernel_3d_val, 8*sizeof(float), cudaMemcpyHostToDevice) == cudaSuccess);
assert(cudaStreamCreate(&stream_) == cudaSuccess);
} }
~GPUContext() { ~GPUContext() {
assert(cudaFree(kernel_1d_) == cudaSuccess); assert(cudaFree(kernel_1d_) == cudaSuccess);
assert(cudaFree(kernel_2d_) == cudaSuccess); assert(cudaFree(kernel_2d_) == cudaSuccess);
assert(cudaFree(kernel_3d_) == cudaSuccess); assert(cudaFree(kernel_3d_) == cudaSuccess);
assert(cudaStreamDestroy(stream_) == cudaSuccess);
} }
const Eigen::GpuDevice& device() const { return gpu_device_; } const Eigen::GpuDevice& device() const { return gpu_device_; }
@ -105,7 +102,7 @@ struct GPUContext {
float* kernel_2d_; float* kernel_2d_;
float* kernel_3d_; float* kernel_3d_;
cudaStream_t stream_; Eigen::CudaStreamDevice stream_;
Eigen::GpuDevice gpu_device_; Eigen::GpuDevice gpu_device_;
}; };

3
unsupported/test/cxx11_tensor_reduction_cuda.cpp
View File

@ -19,7 +19,8 @@
template<int DataLayout> template<int DataLayout>
static void test_full_reductions() { static void test_full_reductions() {
Eigen::GpuDevice gpu_device; Eigen::CudaStreamDevice stream;
Eigen::GpuDevice gpu_device(&stream);
const int num_rows = internal::random<int>(1024, 5*1024); const int num_rows = internal::random<int>(1024, 5*1024);
const int num_cols = internal::random<int>(1024, 5*1024); const int num_cols = internal::random<int>(1024, 5*1024);