// DPC++ material used in the code sample:
// •	A one dimensional array of data.
// •	A device queue, buffer, accessor, and kernel.
#include <CL/sycl.hpp>
#include <array>
#include <iostream>
#include <windows.h>

using namespace sycl;

cpu_selector d_selector;
queue q(d_selector);

unsigned long long  DeltaFPS, CountFPS;

typedef std::array<float4, array_size> IntArray;
constexpr size_t array_size = 100 * 100 * 100;

IntArray a, sum_parallel;

void    VectorAdd(queue &q, const IntArray &a_array, IntArray &sum_parallel)
{
    range<1> num_items{a_array.size()};

    buffer a_buf(a_array);
    buffer sum_buf(sum_parallel.data(), num_items);

    q.submit([&](handler &h)
    {
        accessor a(a_buf, h, read_only);
        accessor sum(sum_buf, h, write_only);

        h.parallel_for(num_items, [=](auto i)
        {
            sum[i] = a[i] + 10;
        });
    });
}

void    ShowFPS()
{
    CountFPS++;
    if (GetTickCount64() > DeltaFPS + 1000)
    {
        printf("FPS = %lld\n", CountFPS);

        CountFPS = 0;
        DeltaFPS = GetTickCount64();
    }
}
int     main()
{
    std::cout << "Running on device: " << q.get_device().get_info<info::device::name>() << "\n";

    while (true)
    {
        VectorAdd(q, a, sum_parallel);
        ShowFPS();
    }

    return 0;
}