CUDA C++ 实现矩阵加法运算，测试GPU性能

#include <iostream>
#include <cuda_runtime.h>
#include <time.h>

using namespace std;


// 在device上做矩阵加法运算
__global__ void sumArrays(int *a, int *b, int *res, const int size)
{
    int index = threadIdx.x + blockIdx.x * blockDim.x;
    if (index < size) {
        res[index] = a[index] + b[index];
    }
}

void random_ints(int* a, const int size)
{
    for (int i = 0; i < size; ++i)
        a[i] = rand();
}

int main()
{
    int dev = 0;
    cudaSetDevice(dev);

    cudaEvent_t cuda_start_time, cuda_stop_time;  // 定义两个CUDA事件类型的变量
    cudaEventCreate(&cuda_start_time);  // CUDA事件记时
    cudaEventCreate(&cuda_stop_time);   // 初始化定义的cudaEvent_t变量
    cudaEventRecord(cuda_start_time);   // 在需要记时的代码块之前记录代表时间开始的事件
    cudaEventQuery(cuda_stop_time);     // cudaEventQuery函数在TCC驱动模式的GPU下可省略，但在处于WDDM驱动模式的GPU必须保留

    clock_t c_start_time , c_end_time;
    c_start_time = clock();  // C语言函数返回记录开始的值

    

    int n = (2048 * 2048);
    int threads_per_block = 1024;
    int *a, *b, *c;
    int *d_a, *d_b, *d_c;

    int size = n * sizeof(int);
    cudaMalloc((void**)&d_a, size);
    cudaMalloc((void**)&d_b, size);
    cudaMalloc((void**)&d_c, size);

    a = (int*)malloc(size);
    random_ints(a, n);
    b = (int*)malloc(size);
    random_ints(b, n);
    c = (int*)malloc(size);
    cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice);
    cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice);

    sumArrays<<<(n + threads_per_block - 1)/threads_per_block, threads_per_block>>>(d_a, d_b, d_c, n);

    cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost);

    cudaFree(d_a);
    cudaFree(d_b);
    cudaFree(d_c);
    
    for (int i = 0; i < n; ++i) {
        cout << c[i] << "\t";
    }
    cout << endl;
    cout << cudaGetErrorString(cudaGetLastError()) << endl;

    free(a);
    free(b);
    free(c);

    cudaDeviceSynchronize();

    cudaEventRecord(cuda_stop_time);
    cudaEventSynchronize(cuda_stop_time);  // 等待事件被记录完毕

    float cuda_run_time;
    cudaEventElapsedTime(&cuda_run_time, cuda_start_time, cuda_stop_time);

    printf("\n(Cuda) The Running time is %f\n", cuda_run_time);

    cudaEventDestroy(cuda_start_time);
    cudaEventDestroy(cuda_stop_time);

    c_end_time = clock(); // 记录结束的值
 
    double c_run_time = ((double)(c_end_time - c_start_time)) / CLK_TCK;  /* 用结束时间减去开始时间，因为是毫秒单位，所以除以CLK_TCK来转化为秒 */
    printf("\n(C) The Running time is %f\n", c_run_time);


    return 0;
}

/*

Cuda程序运行流程如下：

准备待处理数据
在device上分配存储空间
把数据从host拷贝到device
执行device运算
把结果从device拷贝回host
释放device空间
释放host空间

*/