devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Clang-format tests, examples, libraries, benchmarks, etc.

Browse Source
This commit is contained in:
Antonio Sánchez
2023-12-05 21:22:55 +00:00
committed by Rasmus Munk Larsen
parent 3252ecc7a4
commit 46e9cdb7fe
876 changed files with 33453 additions and 37795 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
unsupported/test/cxx11_tensor_image_op_sycl.cpp
View File

@@ -25,59 +25,55 @@ using Eigen::SyclDevice;
using Eigen::Tensor;
using Eigen::TensorMap;
using Eigen::Tensor;
using Eigen::RowMajor;
using Eigen::Tensor;
template <typename DataType, int DataLayout, typename IndexType>
static void test_image_op_sycl(const Eigen::SyclDevice &sycl_device)
{
static void test_image_op_sycl(const Eigen::SyclDevice& sycl_device) {
IndexType sizeDim1 = 245;
IndexType sizeDim2 = 343;
IndexType sizeDim3 = 577;
array<IndexType, 3> input_range ={{sizeDim1, sizeDim2, sizeDim3}};
array<IndexType, 3> slice_range ={{sizeDim1-1, sizeDim2, sizeDim3}};
array<IndexType, 3> input_range = {{sizeDim1, sizeDim2, sizeDim3}};
array<IndexType, 3> slice_range = {{sizeDim1 - 1, sizeDim2, sizeDim3}};
Tensor<DataType, 3,DataLayout, IndexType> tensor1(input_range);
Tensor<DataType, 3,DataLayout, IndexType> tensor2(input_range);
Tensor<DataType, 3, DataLayout, IndexType> tensor1(input_range);
Tensor<DataType, 3, DataLayout, IndexType> tensor2(input_range);
Tensor<DataType, 3, DataLayout, IndexType> tensor3(slice_range);
Tensor<DataType, 3, DataLayout, IndexType> tensor3_cpu(slice_range);
typedef Eigen::DSizes<IndexType, 3> Index3;
Index3 strides1(1L,1L, 1L);
Index3 strides1(1L, 1L, 1L);
Index3 indicesStart1(1L, 0L, 0L);
Index3 indicesStop1(sizeDim1, sizeDim2, sizeDim3);
Index3 strides2(1L,1L, 1L);
Index3 strides2(1L, 1L, 1L);
Index3 indicesStart2(0L, 0L, 0L);
Index3 indicesStop2(sizeDim1-1, sizeDim2, sizeDim3);
Eigen::DSizes<IndexType, 3> sizes(sizeDim1-1,sizeDim2,sizeDim3);
Index3 indicesStop2(sizeDim1 - 1, sizeDim2, sizeDim3);
Eigen::DSizes<IndexType, 3> sizes(sizeDim1 - 1, sizeDim2, sizeDim3);
tensor1.setRandom();
tensor2.setRandom();
DataType* gpu_data1 = static_cast<DataType*>(sycl_device.allocate(tensor1.size()*sizeof(DataType)));
DataType* gpu_data2 = static_cast<DataType*>(sycl_device.allocate(tensor2.size()*sizeof(DataType)));
DataType* gpu_data3 = static_cast<DataType*>(sycl_device.allocate(tensor3.size()*sizeof(DataType)));
DataType* gpu_data1 = static_cast<DataType*>(sycl_device.allocate(tensor1.size() * sizeof(DataType)));
DataType* gpu_data2 = static_cast<DataType*>(sycl_device.allocate(tensor2.size() * sizeof(DataType)));
DataType* gpu_data3 = static_cast<DataType*>(sycl_device.allocate(tensor3.size() * sizeof(DataType)));
TensorMap<Tensor<DataType, 3, DataLayout, IndexType>> gpu1(gpu_data1, input_range);
TensorMap<Tensor<DataType, 3, DataLayout, IndexType>> gpu2(gpu_data2, input_range);
TensorMap<Tensor<DataType, 3, DataLayout, IndexType>> gpu3(gpu_data3, slice_range);
sycl_device.memcpyHostToDevice(gpu_data1, tensor1.data(),(tensor1.size())*sizeof(DataType));
sycl_device.memcpyHostToDevice(gpu_data2, tensor2.data(),(tensor2.size())*sizeof(DataType));
gpu3.device(sycl_device)= gpu1.slice(indicesStart1, sizes) - gpu2.slice(indicesStart2, sizes);
sycl_device.memcpyDeviceToHost(tensor3.data(), gpu_data3,(tensor3.size())*sizeof(DataType));
sycl_device.memcpyHostToDevice(gpu_data1, tensor1.data(), (tensor1.size()) * sizeof(DataType));
sycl_device.memcpyHostToDevice(gpu_data2, tensor2.data(), (tensor2.size()) * sizeof(DataType));
gpu3.device(sycl_device) = gpu1.slice(indicesStart1, sizes) - gpu2.slice(indicesStart2, sizes);
sycl_device.memcpyDeviceToHost(tensor3.data(), gpu_data3, (tensor3.size()) * sizeof(DataType));
tensor3_cpu = tensor1.stridedSlice(indicesStart1,indicesStop1,strides1) - tensor2.stridedSlice(indicesStart2,indicesStop2,strides2);
tensor3_cpu = tensor1.stridedSlice(indicesStart1, indicesStop1, strides1) -
tensor2.stridedSlice(indicesStart2, indicesStop2, strides2);
for (IndexType i = 0; i <slice_range[0] ; ++i) {
for (IndexType i = 0; i < slice_range[0]; ++i) {
for (IndexType j = 0; j < slice_range[1]; ++j) {
for (IndexType k = 0; k < slice_range[2]; ++k) {
VERIFY_IS_EQUAL(tensor3_cpu(i,j,k), tensor3(i,j,k));
VERIFY_IS_EQUAL(tensor3_cpu(i, j, k), tensor3(i, j, k));
}
}
}
@@ -86,19 +82,19 @@ static void test_image_op_sycl(const Eigen::SyclDevice &sycl_device)
sycl_device.deallocate(gpu_data3);
}
template<typename DataType, typename dev_Selector> void sycl_computing_test_per_device(dev_Selector s){
template <typename DataType, typename dev_Selector>
void sycl_computing_test_per_device(dev_Selector s) {
QueueInterface queueInterface(s);
auto sycl_device = Eigen::SyclDevice(&queueInterface);
test_image_op_sycl<DataType, RowMajor, int64_t>(sycl_device);
}
EIGEN_DECLARE_TEST(cxx11_tensor_image_op_sycl) {
for (const auto& device :Eigen::get_sycl_supported_devices()) {
for (const auto& device : Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_computing_test_per_device<half>(device));
CALL_SUBTEST(sycl_computing_test_per_device<float>(device));
CALL_SUBTEST(sycl_computing_test_per_device<float>(device));
#ifdef EIGEN_SYCL_DOUBLE_SUPPORT
CALL_SUBTEST(sycl_computing_test_per_device<double>(device));
CALL_SUBTEST(sycl_computing_test_per_device<double>(device));
#endif
}
}