cuda 1000个数据 开256 线程

#include "cuda_runtime.h"
#include "device_launch_parameters.h"

#include <stdio.h>
#include <iostream>
using namespace std;
#define data_size 1000
#define thread_num 256

__global__ void kernel(int *d_idata,int * d_odata)
{
	const int tid=threadIdx.x;
	int sum=0;
	for (int i=tid;i<data_size;i+=thread_num)
	{
		sum+=d_idata[i]*d_idata[i];
	}
	d_odata[tid]=sum;
}

int main()
{
	int h_idata[data_size];
	for (int i=0;i<data_size;i++)
	{
		h_idata[i]=i;
	}
	int * d_idata;
	int * d_odata;
	cudaMalloc(&d_idata,sizeof(int)*data_size);
	cudaMalloc(&d_odata,sizeof(int)*thread_num);

	cudaMemcpy(d_idata,h_idata,sizeof(int)*data_size,cudaMemcpyHostToDevice);

	kernel<<<1,thread_num,0>>>(d_idata,d_odata);

	int gpu_sum[thread_num];
	cudaMemcpy(&gpu_sum,d_odata,sizeof(int)*thread_num,cudaMemcpyDeviceToHost);

	cudaFree(d_idata);
	cudaFree(d_odata);


	int final_gpu_sum=0;
	for (int i=0;i<thread_num;i++)
	{
		final_gpu_sum+=gpu_sum[i];
	
		printf("gpu_sum[i]%d\n",gpu_sum[i]);
	}
	printf("final_gpu_sum=%d\n",final_gpu_sum);





	int cpu_sum = 0; 
	for(int i = 0; i < data_size; i++)
	{ 
		cpu_sum+= h_idata[i] * h_idata[i]; 
	} 
	printf("cpu_sum: %d\n", cpu_sum);

	cin.get();


   
}