<pre> # include <cmath> // This library enable the use of sqrt.  # include <iostream>  # include <ctime>  #include<iomanip>  #include<cstdlib>  # include <cuda_runtime.h>  //#include <times.h>    using namespace std;    inline clock_t getMilliSecs() {  return clock() / (CLOCKS_PER_SEC / 1000);  }    __global__ void primegen(bool prime, int number2,int x,int *primes_d,int index)  {  int c = 0;  int idx = blockIdx.x * blockDim.x + threadIdx.x;  // number2 =(int) floor (sqrt (x));  // cout<< "value of idx" << idx <<endl;  /* if( !(x!=idx && x%idx == 0 ) )  {  primes_d[c] = x;  c++;    } */    primes_d[idx] = 666;  c++;        }       /*for (int i = 1; i <= x; i++)  {  for ( int j = 2; j <= number2; j++)  {  if ( i!=j && i % j == 0 )  {  prime = false;  break;  }  }  if (prime)  {  primes_d[c]=i;  c += 1;  }  prime = true;     } */        void primenum(long double); // Prototype...        int main()  {  long double x = 0;  cout<<"\n This program will generate all prime numbers up to the"<<"\n number you have entered below...\n";  cout<<"\n Please enter a numberImage: ";  cin>> x;  cout<<"\n Here are all the prime numbers up to "<<x<<".\n";  primenum(x); //function invocation...  //cout<<endl<<"\nThere are "<<c  //<<" prime numbers less than or equal to "<<x<<".\n\n";  return 0;  }    // This function will determine the primenumbers up to num.  void primenum(long double x)  {  int n = x;  int d;  bool prime = true;  //struct tms start_time, stop_time;  int number2;  // number2 =(int) floor (sqrt (x));  clock_t start = getMilliSecs();    cudaDeviceProp prop;  cudaGetDevice(&d);  cudaGetDeviceProperties(&prop, d);  int nThreads = prop.maxThreadsDim[0];  int n_max = nThreads * propgpuA3error.maxGridSize[0];  if ( n> n_max) {  n = n_max;  cout << "n reduced to " << n << endl;  }     //Array to hold generated primes on host  int *primes_h = new int[(int)x];    //Device array to hold the primes on the device  int *primes_d = new int[(int)x];    //allocate device memory and initialize device memory  cudaMalloc((void**)&primes_d, (int)x * sizeof(int));    // cudaMalloc((void**)&c_d, sizeof(int));  cudaMemset(&primes_d,0,x * sizeof(int));    primes_h[0] = 666;  //error checking  cudaError_t error ;  for(int i=0; i<(int)x ; i++){  //if(primes_h[i]>=2 && primes_h[i]<=(int)x){    cout<<"BEFORE:"<<primes_h[i]<<endl;  //}  }  //Kernal goes here  //for ( int i=2; i<= x; i++)  //{  int i=10;  number2 =(int) floor (sqrt ((long double)i));  // primegen<<<1, x>>>(prime,number2,(int)i,primes_d,666);  // primegen<<<1, x>>>(prime,number2,(int)i,primes_d,666);  cudaDeviceSynchronize();  //}  // extract error code from the kernel's execution    error = cudaGetLastError();  if (error != cudaSuccess) {  cout << cudaGetErrorString(error) << endl;  }    //copy the array holding primes from device to host    error =cudaMemcpy(primes_h, primes_d, ((int)x) * sizeof(int), cudaMemcpyDeviceToHost);    if (error != cudaSuccess) {  cout << cudaGetErrorString(error) << endl;  }  // cudaMemcpy(c_h, c_d, sizeof(int), cudaMemcpyDeviceToHost);  //display the primes  for(int i=0; i<(int)x ; i++){  //if(primes_h[i]>=2 && primes_h[i]<=(int)x){  cout<<primes_h[ipng|thumb|widthpx| ]<<endl;  //}  }  cout << "Elapsed time: " << (getMilliSecs() - start) << "ms" << endl;  // cout<< "time: "<< (stop_s-start_s)/double(CLOCKS_PER_SEC)<<endl;  //free allocated memory    delete [] primes_h;  cudaFree(primes_d);    getchar();  } </pre>