int main(int argc, char *argv[])
{
perf_parse_args(argc, argv);
std::cout << "size: " << PERF_N << std::endl;
boost::compute::device device = boost::compute::system::default_device();
boost::compute::context context(device);
boost::compute::command_queue queue(context, device);
std::vector<int> v1 = generate_random_vector<int>(PERF_N / 2);
std::vector<int> v2 = generate_random_vector<int>(PERF_N / 2);
std::vector<int> v3(PERF_N);
std::sort(v1.begin(), v1.end());
std::sort(v2.begin(), v2.end());
boost::compute::vector<int> gpu_v1(PERF_N / 2, context);
boost::compute::vector<int> gpu_v2(PERF_N / 2, context);
boost::compute::vector<int> gpu_v3(PERF_N, context);
boost::compute::copy(v1.begin(), v1.end(), gpu_v1.begin(), queue);
boost::compute::copy(v2.begin(), v2.end(), gpu_v2.begin(), queue);
perf_timer t;
for(size_t trial = 0; trial < PERF_TRIALS; trial++){
t.start();
boost::compute::merge(gpu_v1.begin(), gpu_v1.end(),
gpu_v2.begin(), gpu_v2.end(),
gpu_v3.begin(),
queue
);
t.stop();
}
std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
std::vector<int> check_v3(PERF_N);
boost::compute::copy(gpu_v3.begin(), gpu_v3.end(), check_v3.begin(), queue);
queue.finish();
std::merge(v1.begin(), v1.end(), v2.begin(), v2.end(), v3.begin());
bool ok = std::equal(check_v3.begin(), check_v3.end(), v3.begin());
if(!ok){
std::cerr << "ERROR: merged ranges different" << std::endl;
return -1;
}
return 0;
}
int main(int argc, char *argv[])
{
perf_parse_args(argc, argv);
std::cout << "size: " << PERF_N << std::endl;
// setup context and queue for the default device
boost::compute::device device = boost::compute::system::default_device();
boost::compute::context context(device);
boost::compute::command_queue queue(context, device);
std::cout << "device: " << device.name() << std::endl;
// create vectors of random numbers on the host
std::vector<int> v1(std::floor(PERF_N / 2.0));
std::vector<int> v2(std::ceil(PERF_N / 2.0));
std::generate(v1.begin(), v1.end(), rand_int);
std::generate(v2.begin(), v2.end(), rand_int);
std::sort(v1.begin(), v1.end());
std::sort(v2.begin(), v2.end());
// create vectors on the device and copy the data
boost::compute::vector<int> gpu_v1(std::floor(PERF_N / 2.0), context);
boost::compute::vector<int> gpu_v2(std::ceil(PERF_N / 2.0), context);
boost::compute::copy(
v1.begin(), v1.end(), gpu_v1.begin(), queue
);
boost::compute::copy(
v2.begin(), v2.end(), gpu_v2.begin(), queue
);
boost::compute::vector<int> gpu_v3(PERF_N, context);
boost::compute::vector<int>::iterator gpu_v3_end;
perf_timer t;
for(size_t trial = 0; trial < PERF_TRIALS; trial++){
t.start();
gpu_v3_end = boost::compute::set_intersection(
gpu_v1.begin(), gpu_v1.end(),
gpu_v2.begin(), gpu_v2.end(),
gpu_v3.begin(), queue
);
queue.finish();
t.stop();
}
std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
std::cout << "size: " << std::distance(gpu_v3.begin(), gpu_v3_end) << std::endl;
return 0;
}
