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 vector of random numbers on the host
std::vector<int> host_keys(PERF_N);
std::generate(host_keys.begin(), host_keys.end(), rand);
std::vector<long> host_values(PERF_N);
std::copy(host_keys.begin(), host_keys.end(), host_values.begin());
// create vector on the device and copy the data
boost::compute::vector<int> device_keys(PERF_N, context);
boost::compute::vector<long> device_values(PERF_N, context);
perf_timer t;
for(size_t trial = 0; trial < PERF_TRIALS; trial++){
boost::compute::copy(
host_keys.begin(), host_keys.end(), device_keys.begin(), queue
);
boost::compute::copy(
host_values.begin(), host_values.end(), device_values.begin(), queue
);
t.start();
// sort vector
boost::compute::sort_by_key(
device_keys.begin(), device_keys.end(), device_values.begin(), queue
);
queue.finish();
t.stop();
}
std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
// verify keys are sorted
if(!boost::compute::is_sorted(device_keys.begin(), device_keys.end(), queue)){
std::cout << "ERROR: is_sorted() returned false for the keys" << std::endl;
return -1;
}
// verify values are sorted
if(!boost::compute::is_sorted(device_values.begin(), device_values.end(), queue)){
std::cout << "ERROR: is_sorted() returned false for the values" << 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 vector of keys and random values
std::vector<int> host_keys(PERF_N);
std::vector<int> host_values(PERF_N);
std::generate(host_keys.begin(), host_keys.end(), UniqueKey);
std::generate(host_values.begin(), host_values.end(), rand_int);
// create vectors for keys and values on the device and copy the data
boost::compute::vector<int> device_keys(PERF_N, context);
boost::compute::vector<int> device_values(PERF_N,context);
boost::compute::copy(
host_keys.begin(),
host_keys.end(),
device_keys.begin(),
queue
);
boost::compute::copy(
host_values.begin(),
host_values.end(),
device_values.begin(),
queue
);
// vectors for the results
boost::compute::vector<int> device_keys_results(PERF_N, context);
boost::compute::vector<int> device_values_results(PERF_N,context);
typedef boost::compute::vector<int>::iterator iterType;
std::pair<iterType, iterType> result(
device_keys_results.begin(),
device_values_results.begin()
);
// reduce by key
perf_timer t;
for(size_t trial = 0; trial < PERF_TRIALS; trial++) {
t.start();
result = boost::compute::reduce_by_key(device_keys.begin(),
device_keys.end(),
device_values.begin(),
device_keys_results.begin(),
device_values_results.begin(),
queue);
t.stop();
}
std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
size_t result_size = std::distance(device_keys_results.begin(), result.first);
if(result_size != static_cast<size_t>(host_keys[PERF_N-1] + 1)) {
std::cout << "ERROR: "
<< "wrong number of keys" << result_size << "\n" << (host_keys[PERF_N-1] + 1)
<< std::endl;
return -1;
}
return 0;
}
