CUDA编程入门——实现简单二维矩阵加法，并记录运算时间

念霜学者

已于 2023-04-12 22:08:38 修改

阅读量169

点赞数

文章标签：矩阵算法 c++

于 2023-04-12 22:06:09 首次发布

本文链接：https://blog.csdn.net/weixin_44326290/article/details/130118012

版权

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

#define M 4
#define N 4

__global__ void add(int a[M][N], int b[M][N], int c[M][N])
{
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;

   if (i < M && j < N)
   {
       c[i][j] = a[i][j] + b[i][j];
   }
}

int main()
{
clock_t start, end;

   int(*a_h)[N] = new int [M][N];
   int(*b_h)[N] = new int [M][N];
   int(*c_h)[N] = new int [M][N];

   int(*a_d)[N];
   int(*b_d)[N];
   int(*c_d)[N];

   cudaMalloc((void**)&a_d, sizeof(int) * M * N);
   cudaMalloc((void**)&b_d, sizeof(int) * M * N);
   cudaMalloc((void**)&c_d, sizeof(int) * M * N);

   for (int i = 0; i < M; i++)
   {
       for (int j = 0; j < N; j++)
       {
           a_h[i][j] = 2;
       }
   }
   for (int i = 0; i < M; i++)
   {
       for (int j = 0; j < N; j++)
       {
           b_h[i][j] = 3;
       }
   }

start = clock();

cudaMemcpy(a_d, a_h, sizeof(int) * M * N, cudaMemcpyHostToDevice);
cudaMemcpy(b_d, b_h, sizeof(int) * M * N, cudaMemcpyHostToDevice);

dim3 DimGrid(1, 1);
dim3 DimBlock(4, 4);

add << <DimGrid, DimBlock >> > (a_d, b_d, c_d);

cudaMemcpy(c_h, c_d, sizeof(int) * M * N, cudaMemcpyDeviceToHost);

end = clock();

   for (int i = 0; i < M; i++)
   {
       for (int j = 0; j < N; j++)
       {
           printf("%d ", c_h[i][j]);
       }
       printf("\n");
   }

float timel = (float)(end - start) / CLOCKS_PER_SEC;
printf("执行时间为：%f\n", timel);

   free(a_h);
   free(b_h);
   free(c_h);

   cudaFree(a_d);
   cudaFree(b_d);
   cudaFree(c_d);

return 0;
}