向量相加，請以「nvcc vecAdd01.cu -Xcompiler "/wd 4819"」指令編譯； 以後之範例除了更換檔名以外，原則上其餘皆同：

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

#define SIZE 100000

__global__ static void vecAdd(int *a, int *b, int *c, int len){
	int idx = blockIdx.x * blockDim.x + threadIdx.x;
	if(idx<len){
		c[idx] = a[idx] + b[idx];
	}
}

int a[SIZE], b[SIZE], c1[SIZE], c2[SIZE];

int main()
{
	int *gpu_a, *gpu_b, *gpu_c;
	int i, err = 0;
	
	srand(time(NULL));
	for(i=0;i<SIZE;i++){
		a[i] = rand() % 100;
		b[i] = rand() % 100;
		c1[i] = a[i] + b[i];
	}

	cudaMalloc((void**)&gpu_a, sizeof(int) * SIZE);
	cudaMalloc((void**)&gpu_b, sizeof(int) * SIZE);
	cudaMalloc((void**)&gpu_c, sizeof(int) * SIZE);
	
	cudaMemcpy(gpu_a, a, sizeof(int) * SIZE, cudaMemcpyHostToDevice);
	cudaMemcpy(gpu_b, b, sizeof(int) * SIZE, cudaMemcpyHostToDevice);
	
	vecAdd<<<100, 1000>>>(gpu_a, gpu_b, gpu_c, SIZE);
	cudaThreadSynchronize();
	
	cudaMemcpy(c2, gpu_c, sizeof(int) * SIZE, cudaMemcpyDeviceToHost);
	
	for(i=0;i<SIZE;i++){
		err += abs(c1[i]-c2[i]);
	}
	printf("Err: %d", err);
	
	return 0;
}