#include <iostream>
#include <array>
#include <amp.h>
#include <stdio.h>
#include <tchar.h>
using namespace concurrency;
using namespace std;

int _tmain(int argc, _TCHAR* argv[])
{
    LARGE_INTEGER timeStart;
    QueryPerformanceCounter(&timeStart);

    std::array<int, 10500> primes;
    array_view<int, 1> primesAV(10500, primes);
    int numPrimes = 0;
    array_view<int> numPrimesAV(1, &numPrimes);
    concurrency::extent<1> numKernels(150000);

    parallel_for_each(
        numKernels,
        [=](index<1> idx) mutable restrict(amp)
    {
        int curIdx = idx[0] + 2;

        bool prime=true;
        for (int j=2; j*j <= curIdx; j++)
        {
            if (curIdx % j == 0)
            {
                prime=false;
                break;
            }
        }

        if(prime)
        {
            int curNumPrimes = atomic_fetch_inc(&numPrimesAV(0));
            if(curNumPrimes <= 10500)
                primesAV[curNumPrimes] = curIdx;
        }
    }
    );

    primesAV.synchronize();
    numPrimesAV.synchronize();

    std::sort(begin(primes), end(primes));

    LARGE_INTEGER timeEnd;
    QueryPerformanceCounter(&timeEnd);
    float timeDiff = timeEnd.LowPart - timeStart.LowPart;

    LARGE_INTEGER freqPerSecond;
    QueryPerformanceFrequency(&freqPerSecond);
    float time = timeDiff/(float)freqPerSecond.LowPart;

    for(int i = 0; i < 10000; i++)
    {
        cout << primes[i] << " ";
    }

    cout<<"Time elapsed: "<<time;

    int a;
    cin>>a;

    return 0;
}