cuda 入门--向量相加

 文件名为main.cu

#include<stdio.h>
#include<iostream>
// 两个向量加法kernel，grid和block均为一维
__global__ void add(float* x, float * y, float* z, int n)
{
    // 获取全局索引
    int index = threadIdx.x + blockIdx.x * blockDim.x;
    // 步长
    int stride = blockDim.x * gridDim.x;
    for (int i = index; i < n; i += stride)
    {
        z[i] = x[i] + y[i];
        
    }
}
int main()
{
    int N = 1 << 10;
    int nBytes = N * sizeof(float);
    // 申请host内存
    float *x, *y, *z;
    x = (float*)malloc(nBytes);
    y = (float*)malloc(nBytes);
    z = (float*)malloc(nBytes);

    // 初始化数据
    for (int i = 0; i < N; ++i)
    {
        x[i] = 10.0;
        y[i] = 20.0;
    }

    // 申请device内存
    float *d_x, *d_y, *d_z;
    cudaMalloc((void**)&d_x, nBytes);
    cudaMalloc((void**)&d_y, nBytes);
    cudaMalloc((void**)&d_z, nBytes);

    // 将host数据拷贝到device
    cudaMemcpy((void*)d_x, (void*)x, nBytes, cudaMemcpyHostToDevice);
    cudaMemcpy((void*)d_y, (void*)y, nBytes, cudaMemcpyHostToDevice);
    // 定义kernel的执行配置
    dim3 blockSize(256);
    dim3 gridSize((N + blockSize.x - 1) / blockSize.x);
    // 执行kernel
    add << < gridSize, blockSize >> >(d_x, d_y, d_z, N);

    // 将device得到的结果拷贝到host
    cudaMemcpy((void*)z, (void*)d_z, nBytes, cudaMemcpyDeviceToHost);
    cudaDeviceSynchronize();
    // 检查执行结果
    float maxError = 0.0;
    for (int i = 0; i < N; i++)

    {maxError = fmax(maxError, fabs(z[i] - 30.0));
    
    }
    std::cout << "最大误差: " << maxError << std::endl;

    // 释放device内存
    cudaFree(d_x);
    cudaFree(d_y);
    cudaFree(d_z);
    // 释放host内存
    free(x);
    free(y);
    free(z);

    return 0;
}

CMakeLists.txt：

cmake_minimum_required(VERSION 3.14)
project(cuda_matrixmulti)
FIND_PACKAGE(CUDA REQUIRED)
INCLUDE_DIRECTORIES(
        ./
        ${CUDA_INCLUDE_DIRS}
)
SET(CMAKE_C_COMPILER g++)
add_compile_options(-std=c++11)
add_compile_options(-g)
set(SRC_LIST main.cu)
CUDA_ADD_EXECUTABLE(cuda_matrixmulti ${SRC_LIST})
target_link_libraries(cuda_matrixmulti ${CUDA_LIBRARIES})

 

环境说明：ubuntu18 clion cuda9

参考博客：https://blog.csdn.net/xiaohu2022/article/details/79599947#cuda%E7%BC%96%E7%A8%8B%E6%A8%A1%E5%9E%8B%E5%9F%BA%E7%A1%80