cmake编译cuda

cuda程序在教程中多是用nvcc来进行编译，但是实际项目中cuda程序往往是和c++混在一起的，所以用cmake编译会更方便

cmake目前可以较好的支持cuda编译，本文只列出其中一种方法，仅供参考

 

 

CMakeLists.txt

cmake_minimum_required(VERSION 3.15)
project(test)
find_package(CUDA REQUIRED)

cuda_add_executable(test test.cu)
target_link_libraries(test)

 

test.cu

#include <iostream>

void __global__ add(const int* a, const int* b, int* c)
{
    int index = threadIdx.x;
    c[index] = a[index] + b[index];
}

int main(void)
{
    const int N = 10;
    int *h_a = new int[N];
    int *h_b = new int[N];
    int *h_c = new int[N];
    for (int i = 0; i < N; ++i)
    {
        h_a[i] = 1;
        h_b[i] = 2;
    }

    int *d_a, *d_b, *d_c;
    cudaMalloc((void**)&d_a, sizeof(int) * N);
    cudaMalloc((void**)&d_b, sizeof(int) * N);
    cudaMalloc((void**)&d_c, sizeof(int) * N);

    cudaMemcpy(d_a, h_a, sizeof(int) * N, cudaMemcpyHostToDevice);
    cudaMemcpy(d_b, h_b, sizeof(int) * N, cudaMemcpyHostToDevice);
    add<<<1, N>>>(d_a, d_b, d_c);
    cudaDeviceSynchronize();

    cudaMemcpy(h_c, d_c, sizeof(int) * N, cudaMemcpyDeviceToHost);
    for (int i = 0; i < N; ++i)
    {
        std::cout << h_c[i] << std::endl;
    }

    delete [] h_a;
    delete [] h_b;
    delete [] h_c;
    cudaFree(d_a);
    cudaFree(d_b);
    cudaFree(d_c);

    return 0;
}