cuda编程实现计算两个向量之间的距离

代码解释：
实现的多个向量和一个向量之间的距离的计算。让kernal函数去执行开设一定的线程去做。代码详细解释如下：

#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <iostream>
#include <fstream>
#include <iomanip>
#include <stdio.h>
const int N=33*1024;//修改向量的个数
const int D=256;
//const int MAX=10;
#include<sys/time.h>
#include<math.h>
#include<time.h>
#include<iostream>
using namespace std;
__global__ void MatrixMultiply(float *a,float *b,float*c,int N){
	int tx=threadIdx.x+blockIdx.x*blockDim.x;//因为线程块里面是多个快的线程并发执行的。
		while(tx < N) {//N 表示的是总的向量个数，每个kernal都算的是（线程块*线程数）个向量的距离
		float sum = 0;
		for (int k = 0; k < D; ++k) {
			sum += ((a[tx * D + k] - b[k])*(a[tx * D + k] - b[k]));
		}
		c[tx] = sum;
		tx+=blockDim.x*gridDim.x;//每次增加的是一个grid的数量，也就是一个kernal执行的线程数量
	}
}

cudaError_t matrixMultiplyWithCuda(float *a, float *b, float *c, size_t size);

int main()
{
		//添加计算时间的函数
	struct timeval start,end;
	gettimeofday(&start,NULL);
	float *a=new float[N*D];
	float *b = new float[D];
	float *c = new float[N];

	// read 
	ifstream fin("../data/features_256.txt");//打开文件
//读入数据
	for(int i=0;i<N;i++)
       	     for(int j=0;j<D;j++) 
	            fin>>a[i*D+j];
    fin.close();
    //输出刚刚读入的数据
    for(int i=0;i<N;i++,cout<<endl)
        for(int j=0;j<D;j++) 
            //cout<<"the num of a is"<<a[i*D+j];
     //cout<<a[1*D+2];
	
//printf("this num is %d\n",a[0]);		


        for( int j=0;j<D;j++ )
	{
	  	b[j] = a[1*D+j];
                //printf("this b is %f\n",b[j]);
	}
	//printf("this b is %f\n",b[0]);	
	cudaError_t cudaStatus = matrixMultiplyWithCuda(a, b, c, N);

	
	cudaStatus = cudaThreadExit();

	// host free 
	   //释放申请的空间
    //for(int i=0;i<N;i++) delete []a[i];
    delete []a;

	delete[] b;
	delete[] c;
	gettimeofday(&end,NULL);
       	int timeuse=1000000*(end.tv_sec-start.tv_sec)+end.tv_usec-start.tv_usec;
	printf("total time is %d ms\n",timeuse/1000);
	return 0;
}
cudaError_t matrixMultiplyWithCuda(float *a,float *b,float *c,size_t N){
	float *dev_a = 0;
	float *dev_b = 0;
	float *dev_c = 0;
	cudaError_t cudaStatus;
	cudaStatus = cudaMalloc((void**)&dev_a, N * D * sizeof(float));
	cudaStatus = cudaMalloc((void**)&dev_b, D * sizeof(float));
	cudaStatus = cudaMalloc((void**)&dev_c, N * sizeof(float));
	cudaStatus = cudaMemcpy(dev_a, a, N * D * sizeof(float), cudaMemcpyHostToDevice);
	cudaStatus = cudaMemcpy(dev_b, b, D * sizeof(float), cudaMemcpyHostToDevice);
	if (cudaStatus != cudaSuccess) {
		printf("Something wrong\n");
			}
	// kernal invocation 
	//dim3 threadPerBlock(128, 1, 1);
	//dim3 numBlocks(N / threadPerBlock.x+1, 1, 1);
	MatrixMultiply<<<256, 512>>>(dev_a, dev_b, dev_c, N);//每个kernal分配256个块，每个块都有512个线程
	if (cudaStatus != cudaSuccess) {
		printf( "Calculate wrong\n");
		goto Error;
	}
	cudaStatus = cudaMemcpy(c, dev_c, N * sizeof(float), cudaMemcpyDeviceToHost);
       for(int i=0;i<N;i++)
       {
       	printf("the %d ge distance is %f\n",i,c[i]);
       }
      // printf("the %f and %f is %f",a[0],b[0],c[0]);
Error:
	cudaFree(dev_a);
	cudaFree(dev_b);
	cudaFree(dev_c);
	return cudaStatus;
}