本文共 1857 字，大约阅读时间需要 6 分钟。

.c与.cpp是C与C++源代码，.cu是cuda源代码，核函数是GPU计算用的函数，用__global__标记

.c或.cpp是不可以使用直接调用“add <<<1,3 >>> (a_gpu, b_gpu, c_gpu, num);”类似这样来调用__global__核函数，如果想调用，可以在 .cu 文件中封装一层，就是在 .cu 文件中写一个传同样参数的核函数调用函数addKernel(int *a, int *b, int *c, int size)，即在addKernel(int *a, int *b, int *c, int size)函数中调用核函数add<<<2, 128>>>(dev_a, dev_b, dev_c,size)，然后在外边main.cpp文件中用调用普通函数的方法调用addKernel(int *a, int *b, int *c, int size)函数

#include "cuda_runtime.h"#include "addfunc.cuh"#include "device_launch_parameters.h"#include <stdio.h>#include <time.h>int main(void){	// init data	int num = 10;	int a[10], b[10], c[10];	int* a_gpu, * b_gpu, * c_gpu;	for (int i = 0; i < num; i++)	{		a[i] = i;		b[i] = i * i;	}	cudaMalloc((void**)&a_gpu, num * sizeof(int));	cudaMalloc((void**)&b_gpu, num * sizeof(int));	cudaMalloc((void**)&c_gpu, num * sizeof(int));	// copy data	cudaMemcpy(a_gpu, a, num * sizeof(int), cudaMemcpyHostToDevice);	cudaMemcpy(b_gpu, b, num * sizeof(int), cudaMemcpyHostToDevice);	addKernel(a_gpu, b_gpu, c_gpu, num);	//printf("%d + %d = %d\n", a_gpu[0], b_gpu[0], c_gpu[0]);	// get data	cudaMemcpy(c, c_gpu, num * sizeof(int), cudaMemcpyDeviceToHost);	// visualization	for (int i = 0; i < num; i++)	{		printf("%d + %d = %d\n", a[i], b[i], c[i]);	}	return 0;}//int main()//{//	getThreadNum();//}

#include "cuda_runtime.h"#include "addfunc.cuh"#include "device_launch_parameters.h"#include <stdio.h>#include <time.h>#define CUDA_KERNEL_LOOP(i, n) \  for (int i = blockIdx.x * blockDim.x + threadIdx.x; \       i < (n); \       i += blockDim.x * gridDim.x)__global__ void add(int* a, int* b, int* c, int num){	int i = threadIdx.x;	CUDA_KERNEL_LOOP(i, num) {		if (i < num)		{			c[i] = a[i] + b[i];		}	}}//中间过渡函数void addKernel(int* a_gpu, int* b_gpu, int* c_gpu, int num){		add <<<1,3 >>> (a_gpu, b_gpu, c_gpu, num);}//int main()//{//	getThreadNum();//}

上述是C++项目，如果是C语言项目，可以参照如下文章