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

113
unsupported/test/cxx11_tensor_broadcast_sycl.cpp
View File

@@ -26,47 +26,44 @@ using Eigen::Tensor;
using Eigen::TensorMap;
template <typename DataType, int DataLayout, typename IndexType>
static void test_broadcast_sycl_fixed(const Eigen::SyclDevice &sycl_device){
static void test_broadcast_sycl_fixed(const Eigen::SyclDevice& sycl_device) {
// BROADCAST test:
IndexType inDim1=2;
IndexType inDim2=3;
IndexType inDim3=5;
IndexType inDim4=7;
IndexType bDim1=2;
IndexType bDim2=3;
IndexType bDim3=1;
IndexType bDim4=4;
array<IndexType, 4> in_range = {{inDim1, inDim2, inDim3, inDim4}};
IndexType inDim1 = 2;
IndexType inDim2 = 3;
IndexType inDim3 = 5;
IndexType inDim4 = 7;
IndexType bDim1 = 2;
IndexType bDim2 = 3;
IndexType bDim3 = 1;
IndexType bDim4 = 4;
array<IndexType, 4> in_range = {{inDim1, inDim2, inDim3, inDim4}};
array<IndexType, 4> broadcasts = {{bDim1, bDim2, bDim3, bDim4}};
array<IndexType, 4> out_range; // = in_range * broadcasts
for (size_t i = 0; i < out_range.size(); ++i)
out_range[i] = in_range[i] * broadcasts[i];
for (size_t i = 0; i < out_range.size(); ++i) out_range[i] = in_range[i] * broadcasts[i];
Tensor<DataType, 4, DataLayout, IndexType> input(in_range);
Tensor<DataType, 4, DataLayout, IndexType> input(in_range);
Tensor<DataType, 4, DataLayout, IndexType> out(out_range);
for (size_t i = 0; i < in_range.size(); ++i)
VERIFY_IS_EQUAL(out.dimension(i), out_range[i]);
for (size_t i = 0; i < in_range.size(); ++i) VERIFY_IS_EQUAL(out.dimension(i), out_range[i]);
for (IndexType i = 0; i < input.size(); ++i) input(i) = static_cast<DataType>(i);
for (IndexType i = 0; i < input.size(); ++i)
input(i) = static_cast<DataType>(i);
DataType * gpu_in_data = static_cast<DataType*>(sycl_device.allocate(input.dimensions().TotalSize()*sizeof(DataType)));
DataType * gpu_out_data = static_cast<DataType*>(sycl_device.allocate(out.dimensions().TotalSize()*sizeof(DataType)));
DataType* gpu_in_data =
static_cast<DataType*>(sycl_device.allocate(input.dimensions().TotalSize() * sizeof(DataType)));
DataType* gpu_out_data =
static_cast<DataType*>(sycl_device.allocate(out.dimensions().TotalSize() * sizeof(DataType)));
TensorMap<TensorFixedSize<DataType, Sizes<2, 3, 5, 7>, DataLayout, IndexType>> gpu_in(gpu_in_data, in_range);
TensorMap<Tensor<DataType, 4, DataLayout, IndexType>> gpu_out(gpu_out_data, out_range);
sycl_device.memcpyHostToDevice(gpu_in_data, input.data(),(input.dimensions().TotalSize())*sizeof(DataType));
sycl_device.memcpyHostToDevice(gpu_in_data, input.data(), (input.dimensions().TotalSize()) * sizeof(DataType));
gpu_out.device(sycl_device) = gpu_in.broadcast(broadcasts);
sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(DataType));
sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data, (out.dimensions().TotalSize()) * sizeof(DataType));
for (IndexType i = 0; i < inDim1*bDim1; ++i) {
for (IndexType j = 0; j < inDim2*bDim2; ++j) {
for (IndexType k = 0; k < inDim3*bDim3; ++k) {
for (IndexType l = 0; l < inDim4*bDim4; ++l) {
VERIFY_IS_APPROX(input(i%2,j%3,k%5,l%7), out(i,j,k,l));
for (IndexType i = 0; i < inDim1 * bDim1; ++i) {
for (IndexType j = 0; j < inDim2 * bDim2; ++j) {
for (IndexType k = 0; k < inDim3 * bDim3; ++k) {
for (IndexType l = 0; l < inDim4 * bDim4; ++l) {
VERIFY_IS_APPROX(input(i % 2, j % 3, k % 5, l % 7), out(i, j, k, l));
}
}
}
@@ -77,47 +74,44 @@ static void test_broadcast_sycl_fixed(const Eigen::SyclDevice &sycl_device){
}
template <typename DataType, int DataLayout, typename IndexType>
static void test_broadcast_sycl(const Eigen::SyclDevice &sycl_device){
static void test_broadcast_sycl(const Eigen::SyclDevice& sycl_device) {
// BROADCAST test:
IndexType inDim1=2;
IndexType inDim2=3;
IndexType inDim3=5;
IndexType inDim4=7;
IndexType bDim1=2;
IndexType bDim2=3;
IndexType bDim3=1;
IndexType bDim4=4;
array<IndexType, 4> in_range = {{inDim1, inDim2, inDim3, inDim4}};
IndexType inDim1 = 2;
IndexType inDim2 = 3;
IndexType inDim3 = 5;
IndexType inDim4 = 7;
IndexType bDim1 = 2;
IndexType bDim2 = 3;
IndexType bDim3 = 1;
IndexType bDim4 = 4;
array<IndexType, 4> in_range = {{inDim1, inDim2, inDim3, inDim4}};
array<IndexType, 4> broadcasts = {{bDim1, bDim2, bDim3, bDim4}};
array<IndexType, 4> out_range; // = in_range * broadcasts
for (size_t i = 0; i < out_range.size(); ++i)
out_range[i] = in_range[i] * broadcasts[i];
for (size_t i = 0; i < out_range.size(); ++i) out_range[i] = in_range[i] * broadcasts[i];
Tensor<DataType, 4, DataLayout, IndexType> input(in_range);
Tensor<DataType, 4, DataLayout, IndexType> input(in_range);
Tensor<DataType, 4, DataLayout, IndexType> out(out_range);
for (size_t i = 0; i < in_range.size(); ++i)
VERIFY_IS_EQUAL(out.dimension(i), out_range[i]);
for (size_t i = 0; i < in_range.size(); ++i) VERIFY_IS_EQUAL(out.dimension(i), out_range[i]);
for (IndexType i = 0; i < input.size(); ++i) input(i) = static_cast<DataType>(i);
for (IndexType i = 0; i < input.size(); ++i)
input(i) = static_cast<DataType>(i);
DataType* gpu_in_data =
static_cast<DataType*>(sycl_device.allocate(input.dimensions().TotalSize() * sizeof(DataType)));
DataType* gpu_out_data =
static_cast<DataType*>(sycl_device.allocate(out.dimensions().TotalSize() * sizeof(DataType)));
DataType * gpu_in_data = static_cast<DataType*>(sycl_device.allocate(input.dimensions().TotalSize()*sizeof(DataType)));
DataType * gpu_out_data = static_cast<DataType*>(sycl_device.allocate(out.dimensions().TotalSize()*sizeof(DataType)));
TensorMap<Tensor<DataType, 4, DataLayout, IndexType>> gpu_in(gpu_in_data, in_range);
TensorMap<Tensor<DataType, 4, DataLayout, IndexType>> gpu_in(gpu_in_data, in_range);
TensorMap<Tensor<DataType, 4, DataLayout, IndexType>> gpu_out(gpu_out_data, out_range);
sycl_device.memcpyHostToDevice(gpu_in_data, input.data(),(input.dimensions().TotalSize())*sizeof(DataType));
sycl_device.memcpyHostToDevice(gpu_in_data, input.data(), (input.dimensions().TotalSize()) * sizeof(DataType));
gpu_out.device(sycl_device) = gpu_in.broadcast(broadcasts);
sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(DataType));
sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data, (out.dimensions().TotalSize()) * sizeof(DataType));
for (IndexType i = 0; i < inDim1*bDim1; ++i) {
for (IndexType j = 0; j < inDim2*bDim2; ++j) {
for (IndexType k = 0; k < inDim3*bDim3; ++k) {
for (IndexType l = 0; l < inDim4*bDim4; ++l) {
VERIFY_IS_APPROX(input(i%inDim1,j%inDim2,k%inDim3,l%inDim4), out(i,j,k,l));
for (IndexType i = 0; i < inDim1 * bDim1; ++i) {
for (IndexType j = 0; j < inDim2 * bDim2; ++j) {
for (IndexType k = 0; k < inDim3 * bDim3; ++k) {
for (IndexType l = 0; l < inDim4 * bDim4; ++l) {
VERIFY_IS_APPROX(input(i % inDim1, j % inDim2, k % inDim3, l % inDim4), out(i, j, k, l));
}
}
}
@@ -127,7 +121,8 @@ static void test_broadcast_sycl(const Eigen::SyclDevice &sycl_device){
sycl_device.deallocate(gpu_out_data);
}
template<typename DataType> void sycl_broadcast_test_per_device(const cl::sycl::device& d){
template <typename DataType>
void sycl_broadcast_test_per_device(const cl::sycl::device& d) {
std::cout << "Running on " << d.template get_info<cl::sycl::info::device::name>() << std::endl;
QueueInterface queueInterface(d);
auto sycl_device = Eigen::SyclDevice(&queueInterface);
@@ -138,7 +133,7 @@ template<typename DataType> void sycl_broadcast_test_per_device(const cl::sycl::
}
EIGEN_DECLARE_TEST(cxx11_tensor_broadcast_sycl) {
for (const auto& device :Eigen::get_sycl_supported_devices()) {
for (const auto& device : Eigen::get_sycl_supported_devices()) {
CALL_SUBTEST(sycl_broadcast_test_per_device<half>(device));
CALL_SUBTEST(sycl_broadcast_test_per_device<float>(device));
}