#include <iostream>

#include <CL/sycl.hpp>

using namespace cl::sycl;

auto main(int argc, char **argv) -> int {

  queue q;

  std::cout << "Using device: " << q.get_device().get_info<info::device::name>()
            << "\n";

  int hostArray[42];
  auto deviceArray = static_cast<int *>(malloc_device(42 * sizeof(int), q));

  for (int i = 0; i < 42; i++) {
    hostArray[i] = i;
  }

  q.memcpy(deviceArray, hostArray, 42 * sizeof(int));
  q.wait();

  q.submit([&](handler &h) {
    h.parallel_for(range<1>(42), [=](auto ID) { deviceArray[ID]++; });
  });

  q.wait();

  q.memcpy(hostArray, deviceArray, 42 * sizeof(int));
  q.wait();

  for (int i = 0; i < 42; i++) {
    std::cout << hostArray[i] << " ";
  }

  std::cout << "\n";
}