sum_vector_block.cu

#include <stdlib.h>
#include <iostream>
#define N 1024
using namespace std;

__global__ void vectorAdd(int *a, int *b, int *c) {
    int i = blockIdx.x;
    c[i] = a[i] + b[i];
}

// create random numbers and put on the vector
void randomInts(int *vector) {
    int i = 0;
    time_t t;
    srand((unsigned) time(&t));
    for (; i < N; i++)
        vector[i] = rand() % 1000 + 1;
}

int main() {
    int *a, *b, *c, i = 0;
    int *d_a, *d_b, *d_c;
    int size = N * sizeof(int);

    a = (int *) malloc(size); randomInts(a);
    b = (int *) malloc(size); randomInts(b);
    c = (int *) malloc(size);

    cudaMalloc(&d_a, size);
    cudaMalloc(&d_b, size);
    cudaMalloc(&d_c, size);

    cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice);
    cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice);

    vectorAdd<<<N,1>>>(d_a, d_b, d_c);

    cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost);

    // printing the first ten number of result vector
    for (; i < 10; ++i)
        cout << "c[" << i << "] = " << c[i] << endl;

    free(a); free(b); free(c);
    cudaFree(d_a); cudaFree(d_b); cudaFree(d_c);

    return 0;
}