VS+CUDA环境配置

1. 安装CUDA驱动

2.安装CUDA TOOLKIT

3.在NVIDIA Corporation文件夹下下载CUDA Samples

4.系统环境变量中添加

CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.6
CUDA_SDK_PATH=C:\ProgramData\NVIDIA Corporation\CUDA Samples\v11.6
 
CUDA_BIN_PATH=%CUDA_PAT%\bin
CUDA_LIB_PATH=%CUDA_PATH%\lib\x64
CUDA_SDK_BIN_PATH=%CUDA_SDK_PATH%\bin\win64
CUDA_SDK_LIB_PATH=%CUDA_SDK_PATH%\common\lib\x64
CUDA_SKD_PATH=C:\ProgramData\NVIDIA Corporation\CUDA Sample\v11.3

5.在系统环境变量里添加

%CUDA_BIN_PATH%
%CUDA_LIB_PATH%
%CUDA_SDK_BIN_PATH%
%CUDA_SDK_LIB_path%

6.在cude的文件夹下： C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\extras\visual_studio_integration\MSBuildExtensions 将所有文件复制到D:\Program Files (x86)\Microsoft Visual Studio\2019\Professional\MSBuild\Microsoft\VC\v160\BuildCustomizations 文件夹下；

7.VS中新建项目，右键项目，选择生成依赖项，选择生成自定义，在生成自定义中勾选CUDA。

8.将新建的源文件右击，选择属性->常规->项类型，将项类型设置为CUDA C/C++

9.打开项目属性，项目->“属性”->“配置属性”->“VC++目录”->"包含目录，添加包含目录:$(CUDA_PATH)\include

10.打开项目属性，“VC++目录”->“库目录”，添加库目录:$(CUDA_PATH)\lib\x64

11.“配置属性”->“链接器”->“输入”->“附加依赖项”

cublas.lib
cuda.lib
cudadevrt.lib
cudart.lib
cudart_static.lib
OpenCL.lib

12.测试代码

#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include<iostream>
#include <stdio.h>
using namespace std;
constexpr size_t MAXSIZE = 20;
 
__global__ void addKernel(int* const c, const int* const b, const int* const a)
{
	int i = threadIdx.x;
	c[i] = a[i] + b[i];
}
 
int main()
{
	constexpr size_t length = 6;
	int host_a[length] = { 1,2,3,4,5,6 };
	int host_b[length] = { 10,20,30,40,50,60 };
	int host_c[length];
	//为三个向量在GPU上分配显存
	int* dev_a, *dev_b, *dev_c;
	cudaMalloc((void**)&dev_c, length * sizeof(int));
	cudaMalloc((void**)&dev_a, length * sizeof(int));
	cudaMalloc((void**)&dev_b, length * sizeof(int));
	//将主机端的数据拷贝到设备端
	cudaMemcpy(dev_a, host_a, length * sizeof(int), cudaMemcpyHostToDevice);
	cudaMemcpy(dev_b, host_b, length * sizeof(int), cudaMemcpyHostToDevice);
	cudaMemcpy(dev_c, host_c, length * sizeof(int), cudaMemcpyHostToDevice);
	//在GPU上运行核函数，每个线程进行一个元素的计算
	addKernel << <1, length >> > (dev_c, dev_b, dev_a);
	//将设备端的运算结果拷贝回主机端
	cudaMemcpy(host_c, dev_c, length * sizeof(int), cudaMemcpyDeviceToHost);
	//释放显存
	cudaFree(dev_a);
	cudaFree(dev_b);
	cudaFree(dev_c);
	for (int i = 0; i < length; ++i)
		cout << host_c[i] << " ";
	cout << endl;
	getchar();
	system("pause");
	return 0;
}