GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-09-12 09:23:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

Adding extra test for non-fixed size to broadcast; Replacing stcl with sycl.

Browse Source
This commit is contained in:
Mehdi Goli 2016-11-14 18:13:53 +00:00
parent f8ca893976
commit 05e8c2a1d9
2 changed files with 48 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
unsupported/Eigen/CXX11/src/Tensor/TensorMorphing.h
View File

@ -503,11 +503,11 @@ struct TensorEvaluator<const TensorSlicingOp<StartIndices, Sizes, ArgType>, Devi
} }
return NULL; return NULL;
} }
/// used by stcl /// used by sycl
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const TensorEvaluator<ArgType, Device>& impl() const{ EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const TensorEvaluator<ArgType, Device>& impl() const{
return m_impl; return m_impl;
} }
/// used by stcl /// used by sycl
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const StartIndices& startIndices() const{ EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const StartIndices& startIndices() const{
return m_offsets; return m_offsets;
} }

48
unsupported/test/cxx11_tensor_broadcast_sycl.cpp
View File

@ -25,7 +25,7 @@ using Eigen::SyclDevice;
using Eigen::Tensor; using Eigen::Tensor;
using Eigen::TensorMap; using Eigen::TensorMap;
static void test_broadcast_sycl(const Eigen::SyclDevice &sycl_device){ static void test_broadcast_sycl_fixed(const Eigen::SyclDevice &sycl_device){
// BROADCAST test: // BROADCAST test:
array<int, 4> in_range = {{2, 3, 5, 7}}; array<int, 4> in_range = {{2, 3, 5, 7}};
@ -48,7 +48,49 @@ static void test_broadcast_sycl(const Eigen::SyclDevice &sycl_device){
float * gpu_out_data = static_cast<float*>(sycl_device.allocate(out.dimensions().TotalSize()*sizeof(float))); float * gpu_out_data = static_cast<float*>(sycl_device.allocate(out.dimensions().TotalSize()*sizeof(float)));
TensorMap<TensorFixedSize<float, Sizes<2, 3, 5, 7>>> gpu_in(gpu_in_data, in_range); TensorMap<TensorFixedSize<float, Sizes<2, 3, 5, 7>>> gpu_in(gpu_in_data, in_range);
//TensorMap<Tensor<float, 4>> gpu_in(gpu_in_data, in_range); TensorMap<Tensor<float, 4>> gpu_out(gpu_out_data, out_range);
sycl_device.memcpyHostToDevice(gpu_in_data, input.data(),(input.dimensions().TotalSize())*sizeof(float));
gpu_out.device(sycl_device) = gpu_in.broadcast(broadcasts);
sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(float));
for (int i = 0; i < 4; ++i) {
for (int j = 0; j < 9; ++j) {
for (int k = 0; k < 5; ++k) {
for (int l = 0; l < 28; ++l) {
VERIFY_IS_APPROX(input(i%2,j%3,k%5,l%7), out(i,j,k,l));
}
}
}
}
printf("Broadcast Test with fixed size Passed\n");
sycl_device.deallocate(gpu_in_data);
sycl_device.deallocate(gpu_out_data);
}
static void test_broadcast_sycl(const Eigen::SyclDevice &sycl_device){
// BROADCAST test:
array<int, 4> in_range = {{2, 3, 5, 7}};
array<int, 4> broadcasts = {{2, 3, 1, 4}};
array<int, 4> out_range; // = in_range * broadcasts
for (size_t i = 0; i < out_range.size(); ++i)
out_range[i] = in_range[i] * broadcasts[i];
Tensor<float, 4> input(in_range);
Tensor<float, 4> out(out_range);
for (size_t i = 0; i < in_range.size(); ++i)
VERIFY_IS_EQUAL(out.dimension(i), out_range[i]);
for (int i = 0; i < input.size(); ++i)
input(i) = static_cast<float>(i);
float * gpu_in_data = static_cast<float*>(sycl_device.allocate(input.dimensions().TotalSize()*sizeof(float)));
float * gpu_out_data = static_cast<float*>(sycl_device.allocate(out.dimensions().TotalSize()*sizeof(float)));
TensorMap<Tensor<float, 4>> gpu_in(gpu_in_data, in_range);
TensorMap<Tensor<float, 4>> gpu_out(gpu_out_data, out_range); TensorMap<Tensor<float, 4>> gpu_out(gpu_out_data, out_range);
sycl_device.memcpyHostToDevice(gpu_in_data, input.data(),(input.dimensions().TotalSize())*sizeof(float)); sycl_device.memcpyHostToDevice(gpu_in_data, input.data(),(input.dimensions().TotalSize())*sizeof(float));
gpu_out.device(sycl_device) = gpu_in.broadcast(broadcasts); gpu_out.device(sycl_device) = gpu_in.broadcast(broadcasts);
@ -68,8 +110,10 @@ static void test_broadcast_sycl(const Eigen::SyclDevice &sycl_device){
sycl_device.deallocate(gpu_out_data); sycl_device.deallocate(gpu_out_data);
} }
void test_cxx11_tensor_broadcast_sycl() { void test_cxx11_tensor_broadcast_sycl() {
cl::sycl::gpu_selector s; cl::sycl::gpu_selector s;
Eigen::SyclDevice sycl_device(s); Eigen::SyclDevice sycl_device(s);
CALL_SUBTEST(test_broadcast_sycl_fixed(sycl_device));
CALL_SUBTEST(test_broadcast_sycl(sycl_device)); CALL_SUBTEST(test_broadcast_sycl(sycl_device));
} }