#include<stdio.h>
#include<stdlib.h>
#include<stdlib.h>
#define n 3

/*__global__ void kernel(int **support[],int size)
{

    int i=threadIdx.x;
    int j=threadIdx.y;
    if(i<size)
        support[i*size+j]=10;
}
*/
int main()
{
    long int **matrix=NULL;
    long int **dev_matrix=NULL;
    long int **support=NULL;
    int i,j,er=1000;
    long int **result=NULL;
   int size;
   matrix=(long int **)malloc(sizeof(long int *)*n);
    for(i=0;i<n;i++)
    {
        matrix[i]=(long int *)calloc(n,sizeof(long int));
    }

    result=(long int **)malloc(sizeof(long int *)*n);
    for(i=0;i<n;i++)
    {
        result[i]=(long int *)calloc(n,sizeof(long int));
    }
    for(i=0;i<n;i++)
    {
        for(j=0;j<n;j++)
            {
                result[i][j]=100;
            }
        printf("\n");
    }
    printf("Result before copy:\n");
    for(i=0;i<n;i++)
    {
        for(j=0;j<n;j++)
            {
                printf("%ld\t",result[i][j]);
            }
        printf("\n");
    }


    //pointer array to store list of elements
    er=cudaMalloc((void **) &dev_matrix,n*sizeof(long int *));
    if(er) printf("%d Error in allocation",er);
    //For copying the pointers to GPU this support helps
    support = (long int **) malloc(sizeof(long int*)*n);

    //Copied the support vector of the pointers
    er=cudaMemcpy(support,dev_matrix,n*sizeof(long int*),cudaMemcpyDeviceToHost);
    if(er) printf("%dError in copying the pointers",er);
    for (i=0;i<n;++i)
    {
        //printf("hello");
        er=cudaMalloc((void**)&support[i],sizeof(long int)*n);
        if(er) printf("%dError in allocation",er);
    }

    for(i=0;i<n;i++)
    {

        for(j=0;j<n;j++)
        {
            //printf("I am coming from device to host:\n");
            cudaMemcpy(support[i*n+j],matrix[i*n+j],sizeof(long int),cudaMemcpyHostToDevice);
        }
    }
    size=n;
    /*for(i=0;i<n;i++)
    {
    kernel<<<1,n>>>(dev_matrix[i],size);
    }*/
    for(i=0;i<n;i++)
    {

        for(j=0;j<n;j++)
        {
            //printf("I am coming from device to host:\n");
            cudaMemcpy(result[i*n+j],support[i*n+j],sizeof(long int),cudaMemcpyDeviceToHost);
        }
    }
    printf("Result after copy:\n");
    for(i=0;i<n;i++)
    {
        for(j=0;j<n;j++)
            {
                printf("Element[%d][%d]=%ld\t",i,j,result[i][j]);
            }
        printf("\n");
    }


return 0;
}



/*
The output that get is
Result before copy:
100 100 100
100 100 100
100 100 100

Result after copy:
Element[0][0]=0 Element[0][1]=100   Element[0][2]=100
Element[1][0]=0 Element[1][1]=100   Element[1][2]=100
Element[2][0]=0 Element[2][1]=100   Element[2][2]=100



*/