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

Adding Memset; optimising MecopyDeviceToHost by removing double copying;

Browse Source
This commit is contained in:
Mehdi Goli
2016-11-10 18:45:12 +00:00
parent 75c080b176
commit 2e704d4257
4 changed files with 78 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
unsupported/test/cxx11_tensor_device_sycl.cpp
View File

@@ -19,10 +19,23 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
#include<stdint.h>
void test_device_sycl(const Eigen::SyclDevice &sycl_device) {
std::cout <<"Helo from ComputeCpp: the requested device exists and the device name is : "
<< sycl_device.m_queue.get_device(). template get_info<cl::sycl::info::device::name>() <<std::endl;;
int sizeDim1 = 100;
array<int, 1> tensorRange = {{sizeDim1}};
Tensor<int, 1> in(tensorRange);
Tensor<int, 1> in1(tensorRange);
memset(in1.data(), 1,in1.dimensions().TotalSize()*sizeof(int));
int * gpu_in_data = static_cast<int*>(sycl_device.allocate(in.dimensions().TotalSize()*sizeof(int)));
sycl_device.memset(gpu_in_data, 1,in.dimensions().TotalSize()*sizeof(int) );
sycl_device.memcpyDeviceToHost(in.data(), gpu_in_data, in.dimensions().TotalSize()*sizeof(int) );
for (int i=0; i<in.dimensions().TotalSize(); i++)
VERIFY_IS_APPROX(in(i), in1(i));
sycl_device.deallocate(gpu_in_data);
}
void test_cxx11_tensor_device_sycl() {
cl::sycl::gpu_selector s;