测试cuda程序
#include <stdio.h>
#include <cuda.h>
__global__ void add(int* a, int* b, int* c) {
int idx = blockIdx.x * blockDim.x + threadIdx.x;
c[idx] = a[idx] + b[idx];
}
void cuda_test(int n) {
int *a, *b, *c;
int size = n * sizeof(int);
// Allocate memory on the host
a = (int*)malloc(size);
b = (int*)malloc(size);
c = (int*)malloc(size);
// Initialize arrays
for (int i = 0; i < n; i++) {
a[i] = i;
b[i] = i;
}
// Allocate memory on the device
int *d_a, *d_b, *d_c;
cudaMalloc(&d_a, size);
cudaMalloc(&d_b, size);
cudaMalloc(&d_c, size);
// Copy input data from host to device
cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice);
cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice);
// Launch kernel on the device
int threadsPerBlock = 256;
int blocksPerGrid = (n + threadsPerBlock - 1) / threadsPerBlock;
add<<<blocksPerGrid, threadsPerBlock>>>(d_a, d_b, d_c);
// Copy output data from device to host
cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost);
// Print results
for (int i = 0; i < n; i++) {
printf("%d + %d = %d\n", a[i], b[i], c[i]);
}
// Cleanup
free(a);
free(b);
free(c);
cudaFree(d_a);
cudaFree(d_b);
cudaFree(d_c);
}
int main(){
cuda_test(1024);
return 0;
}
打开VS2013 x64 兼容工具命令提示，切换到程序所在目录，使用nvcc编译
nvcc cudatest.cu -o a.exe
得到可执行的a.exe文件，运行的效果
动态库
源码
#include <stdio.h>
#include <cuda.h>
__global__ void add(int* a, int* b, int* c) {
int idx = blockIdx.x * blockDim.x + threadIdx.x;
c[idx] = a[idx] + b[idx];
}
extern "C"
void __declspec(dllexport) cuda_test(int n) {
int *a, *b, *c;
int size = n * sizeof(int);
// Allocate memory on the host
a = (int*)malloc(size);
b = (int*)malloc(size);
c = (int*)malloc(size);
// Initialize arrays
for (int i = 0; i < n; i++) {
a[i] = i;
b[i] = i;
}
// Allocate memory on the device
int *d_a, *d_b, *d_c;
cudaMalloc(&d_a, size);
cudaMalloc(&d_b, size);
cudaMalloc(&d_c, size);
// Copy input data from host to device
cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice);
cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice);
// Launch kernel on the device
int threadsPerBlock = 256;
int blocksPerGrid = (n + threadsPerBlock - 1) / threadsPerBlock;
add<<<blocksPerGrid, threadsPerBlock>>>(d_a, d_b, d_c);
// Copy output data from device to host
cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost);
// Print results
for (int i = 0; i < n; i++) {
printf("%d + %d = %d\n", a[i], b[i], c[i]);
}
// Cleanup
free(a); free(b); free(c);
cudaFree(d_a); cudaFree(d_b); cudaFree(d_c);
}
说明
extern "C" 支持c语言形式导出名
__declspec(dllexport) 动态库符号导出
__declspec(dllimport) 动态库符号导入
编译生成动态库
使用nvcc命令生成dllcudatest.dll动态库文件
nvcc -m32 --shared cudatestdyn.cu -o dllcudatest.dll
cudatestdyn.cu
正在创建库 dllcudatest.lib 和对象 dllcudatest.exp
-m32 生成的动态库位数，如果是需要64位动态库则是-m64
生成了dllcudatest.dll、dllcudatest.exp、dllcudatest.lib三个文件
使用动态库文件
1.新建win32命令窗口项目，将动态库导出的lib文件放到项目目录下
2.添加一个源文件
#include <Windows.h>
#pragma comment(lib,"dllcudatest.lib")//加载动态库导出文件
extern "C"
void __declspec(dllimport) cuda_test(int);//动态库导入函数声明
int main(){
cuda_test(1024);//使用库函数功能
system("pause");
return 0;
}
3.编译连接，运行，会提示找不到动态库文件，将生成的动态库文件放入到程序的运行目录下，所在项目的Debug或者Release文件夹。
运行成功
静态库
源码
#include <stdio.h>
#include <cuda.h>
__global__ void add(int* a, int* b, int* c) {
int idx = blockIdx.x * blockDim.x + threadIdx.x;
c[idx] = a[idx] + b[idx];
}
extern "C"
void cuda_test(int n) {
int *a, *b, *c;
int size = n * sizeof(int);
// Allocate memory on the host
a = (int*)malloc(size);
b = (int*)malloc(size);
c = (int*)malloc(size);
// Initialize arrays
for (int i = 0; i < n; i++) {
a[i] = i;
b[i] = i;
}
// Allocate memory on the device
int *d_a, *d_b, *d_c;
cudaMalloc(&d_a, size);
cudaMalloc(&d_b, size);
cudaMalloc(&d_c, size);
// Copy input data from host to device
cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice);
cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice);
// Launch kernel on the device
int threadsPerBlock = 256;
int blocksPerGrid = (n + threadsPerBlock - 1) / threadsPerBlock;
add<<<blocksPerGrid, threadsPerBlock>>>(d_a, d_b, d_c);
// Copy output data from device to host
cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost);
// Print results
for (int i = 0; i < n; i++) {
printf("%d + %d = %d\n", a[i], b[i], c[i]);
}
// Cleanup
free(a);
free(b);
free(c);
cudaFree(d_a);
cudaFree(d_b);
cudaFree(d_c);
}
编译静态库
使用nvcc编译
nvcc -m32 --lib cudateststa.cu -o libcudatest.lib
生成一个lib文件
使用静态库文件
1.新建一个win32命令窗口项目，将静态库lib文件放入到项目目录
2.静态库编译链接时需要cuda静态库支持。在cuda的库目录使用下面的命令得到所有lib文件名称，并且加入编译项
dir /B *.lib
3.新建源代码
#include <Windows.h>
#pragma comment(lib,"libcudatest.lib")//导入静态库
#pragma comment(lib,"D:\\BuildChanTools\\NVDIA GPU Computing Toolkit\\lib\\Win32\\cuda.lib")
#pragma comment(lib,"D:\\BuildChanTools\\NVDIA GPU Computing Toolkit\\lib\\Win32\\cudadevrt.lib")
#pragma comment(lib,"D:\\BuildChanTools\\NVDIA GPU Computing Toolkit\\lib\\Win32\\cudart.lib")
#pragma comment(lib,"D:\\BuildChanTools\\NVDIA GPU Computing Toolkit\\lib\\Win32\\cudart_static.lib")
#pragma comment(lib,"D:\\BuildChanTools\\NVDIA GPU Computing Toolkit\\lib\\Win32\\nvcuvid.lib")
#pragma comment(lib,"D:\\BuildChanTools\\NVDIA GPU Computing Toolkit\\lib\\Win32\\OpenCL.lib")
extern "C"
void cuda_test(int);//函数声明
int main(){
cuda_test(1024);
system("pause");
return 0;
}
4.运行
