#include <stdio.h>
#include <time.h>
#include <cutil.h>

#define MAX 1048576
#define N   1024

int mat_1[MAX];
int mat_2[MAX];
int result[MAX];

__global__ void kernel(int* A, int* B, int* C)
{
	int x = blockIdx.x * blockDim.x + threadIdx.x;
	if(x < MAX)
	{
		C[x] = A[x] + B[x];
	}
}

int main(void)
{
	int i = 0;
    unsigned int hTimer;
	int *A, *B, *C;

	for(i = 0; i < MAX; i++)
	{
		mat_1[i] = i;
		mat_2[i] = MAX-i;
	}	

    cutCreateTimer(&hTimer);
	cudaThreadSynchronize();
	cutResetTimer(hTimer);
	cutStartTimer(hTimer);

	unsigned int blockSize = 512;
	unsigned int nBlocks = (MAX+1)/blockSize;

	cudaMalloc((void **) &A, MAX * sizeof(int));
	cudaMalloc((void **) &B, MAX * sizeof(int));
	cudaMalloc((void **) &C, MAX * sizeof(int));

	cudaMemcpy(A, mat_1, MAX * sizeof(int), cudaMemcpyHostToDevice);
	cudaMemcpy(B, mat_2, MAX * sizeof(int), cudaMemcpyHostToDevice);
	//cudaMemcpy(C, result, MAX * sizeof(int),cudaMemcpyHostToDevice);

	kernel<<<nBlocks, blockSize>>>(A, B, C);
	
	cudaMemcpy(result, C, MAX * sizeof(int), cudaMemcpyDeviceToHost);
	//for(i = 0; i < MAX; i++)
	//{
	//	printf("C[%d] = %f\n", i, result[i]);
	//}
    cudaThreadSynchronize();
    cutStopTimer(hTimer);
    printf("GPU time: %f msecs.\n", cutGetTimerValue(hTimer));

    cutCreateTimer(&hTimer);
	cudaThreadSynchronize();
	cutResetTimer(hTimer);
	cutStartTimer(hTimer);

	for(i = 0; i < MAX; i++)
	{
		result[i] = mat_1[i] + mat_2[i];
	}
    cudaThreadSynchronize();
    cutStopTimer(hTimer);
    printf("GPU time: %f msecs.\n", cutGetTimerValue(hTimer));
	return 0;
}