CUDA:二维矩阵运算的应用

本文旨在分享通过cuda进行的一种二维矩阵运算方法

具体的二维矩阵内存分配，以及参数传递，通过以下的一个二维矩阵加法例子进行展示。

 运行环境：cuda8.0; visual studio 2015

 源码如下所示：

#include <stdio.h>
#include <stdlib.h>
#include"cuda_runtime.h"
#include"device_launch_parameters.h"
#include"iostream"
using namespace std;
/*   Example:2维矩阵加法  */

const int N = 10;
const int M = 5;
__global__ void add_2D(float A[][M], float B[][M], float C[][M]);
int main()
{
	/*cpu上开空间*/
	float(*h_a)[M] = new float[N][M];
	float(*h_b)[M] = new float[N][M];
	float(*h_c)[M] = new float[N][M];
	cout << "h_a:" << endl;
	for (int i = 0; i < N; i++) {
		for (int j = 0; j < M; j++) {
			h_a[i][j] = i + j;
			h_b[i][j] = i + j;
			cout << h_a[i][j] << " ";
		}
		cout << endl;
	}
	/*gpu上开空间*/
	float(*d_a)[M], (*d_b)[M], (*d_c)[M];
	cudaMalloc((void **)&d_a, sizeof(float)*N*M);
	cudaMalloc((void **)&d_b, sizeof(float)*N*M);
	cudaMalloc((void **)&d_c, sizeof(float)*N*M);

	/*cpu 到 gpu*/
	cudaMemcpy(d_a, h_a, sizeof(float)*N*M, cudaMemcpyHostToDevice);
	cudaMemcpy(d_b, h_b, sizeof(float)*N*M, cudaMemcpyHostToDevice);
	cudaMemcpy(d_c, h_c, sizeof(float)*N*M, cudaMemcpyHostToDevice);

	/*核函数*/
	dim3 blockNum(1);
	dim3 threadNum(N, M);
	add_2D << <blockNum, threadNum >> >(d_a, d_b, d_c);

	/*gpu 到 cpu*/
	cudaMemcpy(h_c, d_c, sizeof(float)*N*M, cudaMemcpyDeviceToHost);
	cout << "h_c:" << endl;
	for (int i = 0; i < N; i++) {
		for (int j = 0; j < M; j++) {
			cout << h_c[i][j] << " ";
		}
		cout << endl;
	}
	return 0;
}

__global__ void add_2D(float A[][M], float B[][M], float C[][M])
{
	unsigned int ix = threadIdx.x + blockDim.x*blockIdx.x;
	unsigned int iy = threadIdx.y + blockDim.y*blockIdx.y;
	if (ix < N && iy < M) {
		C[ix][iy] = A[ix][iy] + B[ix][iy];
	}
}
	

运行结果：