Link Copied
Hi,
Thanks for reaching out to us.
>>Is there any specific oneAPI macro..
Please find the sample code and steps we have followed to target specific devices from our end.
vec_add.cpp:
#include <CL/sycl.hpp>
#include <iostream>
#include <vector>
using namespace sycl;
#define size 1024
int main() {
std::vector<int> A(size, size), B(size, size), C(size, 0);
{
queue cpuQ(cpu_selector{});
queue gpuQ(gpu_selector{});
range<1> R(size);
buffer<int,1> buffA(A.data(), R);
buffer<int,1> buffB(B.data(), R);
buffer<int,1> buffC(C.data(), R);
#ifdef CPU
std::cout<<"Running on: "<<cpuQ.get_device().get_info<sycl::info::device::name>()<<"\n";
cpuQ.submit([&](handler &cgh) {
auto acc_buffA = buffA.get_access<access::mode::read>(cgh);
auto acc_buffB = buffB.get_access<access::mode::read>(cgh);
auto acc_buffC = buffC.get_access<access::mode::write>(cgh);
cgh.parallel_for(R, [=](id<1> i) {
acc_buffC[i] = acc_buffA[i] + acc_buffB[i];
}
);
});
#elif GPU
std::cout<<"Running on: "<<gpuQ.get_device().get_info<sycl::info::device::name>()<<"\n";
gpuQ.submit([&](handler &cgh) {
auto acc_buffA = buffA.get_access<access::mode::read>(cgh);
auto acc_buffB = buffB.get_access<access::mode::read>(cgh);
auto acc_buffC = buffC.get_access<access::mode::write>(cgh);
cgh.parallel_for(R, [=](id<1> i) {
acc_buffC[i] = acc_buffA[i] + acc_buffB[i];
}
);
});
#endif
}
std::vector<int> vecValidate(size, 2*size);
(C==vecValidate) ? std::cout << "Success\n" : std::cout<<"Failure\n";
return 0;
}
Command to execute:
dpcpp vec_add.cpp -DCPU -o cpuadd && ./cpuadd
dpcpp vec_add.cpp -DGPU -o gpuadd && ./gpuadd
Please find the attached screenshot for more details:
We use the -D option to define a macro name while compiling the source code.
Please refer to the below link for more details:
We can target specific backend using the SYCL_DEVICE_FILTER environment variable.
Syntax:
export SYCL_DEVICE_FILTER=backend:device_type:device_num
Possible values of backend are:
host
level_zero
opencl
Possible values of device_type are:
host
cpu
gpu
acc
device_num is an integer that indexes the enumeration of devices from the sycl-ls utility tool and will return all devices with an index from all different backends.
Please find below link for more details:
Thanks & Regards,
Noorjahan
