CUDA学习（三）之使用GPU进行两个数组相加

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

#include <iomanip>
#include <iostream>
#include <stdio.h>

using namespace std;

//检测GPU
bool CheckCUDA(void){
    int count = 0;
    int i = 0;

    cudaGetDeviceCount(&count);
    if (count == 0) {
        printf("找不到支持CUDA的设备!\n");
        return false;
    }
    cudaDeviceProp prop;
    for (i = 0; i < count; i++) {
        if (cudaGetDeviceProperties(&prop, i) == cudaSuccess) {
            if (prop.major >= 1) {
                break;
            }
        }
    }
    if (i == count) {
        printf("找不到支持CUDA的设备!\n");
        return false;
    }
    cudaGetDeviceProperties(&prop, 0);
    printf("GPU is: %s\n", prop.name);
    cudaSetDevice(0);
    printf("CUDA initialized success.\n");
    return true;
}//使用一维数组相加
__global__ void addForOneDim(double *a, double *b, double *c, int N);

//初始化一维数组
void InitOneDimArray(double *a, double b, int N);

int main(){
    //检测GPU
    if (!CheckCUDA()){
        cout << "No CUDA device.";
        return 0;
    }//****数组相加************************************************************************************************************************
    cout << "****************************************数组相加*********************************************************************" << endl;
    int N = 20;                  //定义数组大小
    double *h_a_one, *h_b_one, *h_c_one;      //声明在CPU上使用的指针
    double *d_a_one, *d_b_one, *d_c_one;             //声明在GPU上使用的指针
    //为数组分配大小
    h_a_one = new double[N];    
    h_b_one = new double[N];
    h_c_one = new double[N];

    cudaMalloc((void **)&d_a_one, sizeof(double)*N);    //在GPU上分配内存空间
    cudaMalloc((void **)&d_b_one, sizeof(double)*N);
    cudaMalloc((void **)&d_c_one, sizeof(double)*N);
    //为数组初始化
    InitOneDimArray(h_a_one, 1.1, N);
    InitOneDimArray(h_b_one, 2.2, N);
    
    //使用GPU中分配的指针指向CPU中的数组
    cudaMemcpy(d_a_one, h_a_one, sizeof(double)*N, cudaMemcpyHostToDevice);
    cudaMemcpy(d_b_one, h_b_one, sizeof(double)*N, cudaMemcpyHostToDevice);

    //调用核函数，使用1个线程块N个线程
    //addForOneDim<<<1, N>>>(h_a_one, h_b_one, d_c_one, N);    //不能使用h_a_one和h_b_one，只能使用GPU上定义的指针，不然结果如图一所示
    addForOneDim<<<1, N>>>(d_a_one, d_b_one, d_c_one, N);      //结果如图二所示
    //将GPU上计算好的结果返回到CPU上定义好的变量
    cudaMemcpy(h_c_one, d_c_one, sizeof(double)*N, cudaMemcpyDeviceToHost);

    //打印结果
    for (int i = 0; i < N; i++){
        cout <<  h_a_one[i] << " + " << h_b_one[i] << " = " << h_c_one[i] << endl;  
    }


    cout << endl << endl;
    system("pause");
    return 0;
}
//使用一维数组相加
__global__ void addForOneDim(double *a, double *b, double *c, int N){
    int tid = threadIdx.x;      //线程索引
    if (tid < N){
        c[tid] = a[tid] + b[tid];
    }
}

//初始化一维数组
void InitOneDimArray(double *a, double b, int N){
    for (int i = 0; i < N; i++){
        a[i] = (i+1) * b;
        //cout << a[i] << endl;
    }
}