cuda,内存变量定义

#include <stdio.h>

#include "cuda_runtime.h"

#include "device_launch_parameters.h"

#include <iostream>

#include <stdlib.h>

#include <conio.h>

using namespace std;

__global__ void func1(int* data1, int* data2, int* data3);

__device__ void func2(int *data2, int i);

#define ARRAY_SIZE 10

#define ARRAY_SIZE_IN_BYTES (sizeof(int)*(ARRAY_SIZE)) //定义了一个int型长度128的数组

int cpu_data1[ARRAY_SIZE];

int cpu_data2[ARRAY_SIZE];

int cpu_data3[ARRAY_SIZE];

int main()

{

const int num_blocks = 1;

const int num_threads = ARRAY_SIZE / num_blocks;

int * gpu_data1;

int * gpu_data2;

int * gpu_data3;

cudaMalloc((void **)&gpu_data1, ARRAY_SIZE_IN_BYTES);

cudaMalloc((void **)&gpu_data2, ARRAY_SIZE_IN_BYTES);

cudaMalloc((void **)&gpu_data3, ARRAY_SIZE_IN_BYTES);

func1 << <num_blocks, num_threads >> >(gpu_data1, gpu_data2, gpu_data3);

cudaMemcpy(cpu_data1, gpu_data1, ARRAY_SIZE_IN_BYTES, cudaMemcpyDeviceToHost);

cudaMemcpy(cpu_data2, gpu_data2, ARRAY_SIZE_IN_BYTES, cudaMemcpyDeviceToHost);

cudaMemcpy(cpu_data3, gpu_data3, ARRAY_SIZE_IN_BYTES, cudaMemcpyDeviceToHost);

cudaFree(gpu_data1);

cudaFree(gpu_data2);

cudaFree(gpu_data3);

for (int i = 0; i < ARRAY_SIZE; i++)

{

cout << cpu_data1[i] << " " << cpu_data2[i] << " " << cpu_data3[i] << endl;

}

cin.get();

return 0;

}

__device__ int A = 10;

__constant__ int static C=10;

__shared__ unsigned int B;

__global__ void func1(int* data1, int* data2, int* data3)

{

int i = (blockIdx.x*blockDim.x) + threadIdx.x;

if (i % 2)

{

A = A + 1;

data1[i] = A;

B = B + 10;

data2[i] = B;

data3[i] &= C;

}

}

__device__ void func2(int *data2, int i)

{

}