CUDA+Vector测试程序

/*
* Copyright 徐洪志(西北农林科技大学.信息工程学院).  All rights reserved.
* Data: 2012-4-15
*/
//
// 此程序是演示了vector型数据如何拷贝入、出显存
#include <cutil_inline.h>

#include <iostream>
#include <vector>
using namespace std;


///////////////////////////////////////////////////////////////////////////////////////////
//
// MAIN
//
///////////////////////////////////////////////////////////////////////////////////////////
int main(int argc, char** argv)
{
	CUT_DEVICE_INIT(argc, argv);  // 启动CUDA
	int row, col;
	/// Vector-->Device-->Host  1D
	cout << "Vector-->Device-->Host 1D" << endl;
	vector<int> vec;  // Host端vector
	int *gpu_data;    // Device端data
	int *cpu_data;    // Host端data
	int dataWd = 20;
	cpu_data = (int*)calloc(dataWd, sizeof(int));  // 申请内存空间
	cutilSafeCall( cudaMalloc((void**) &gpu_data, sizeof(int) * dataWd));  // 申请显存空间
	cutilSafeCall( cudaMemset(gpu_data, 0, sizeof(float) * dataWd));
	if((cpu_data == NULL)||(gpu_data == NULL))   // 判断空间是否申请成功
	{	
		cout << "Alloc Memery Error" << endl;
		return -1;
	}
	for(row = 0; row < dataWd; ++row)   // 给Host端的vector初始化
		vec.push_back(row);
	cutilSafeCall( cudaMemcpy(gpu_data, &vec[0] , sizeof(int) * dataWd, cudaMemcpyHostToDevice));  // 将Host端vector拷贝入Device端data
	cutilSafeCall( cudaMemcpy(cpu_data, gpu_data, sizeof(int) * dataWd, cudaMemcpyDeviceToHost));  // 将Device端data拷贝入Host端data

	for(row = 0; row < dataWd; ++row)   // 打印Host端data
		cout << cpu_data[row] << " ";
	cout << endl;
	
	cutilSafeCall( cudaFree(gpu_data));  // 释放显存空间
	free(cpu_data);                      // 释放内存空间

	/// vector-->Device-->Host  2D
	cout << "Vector-->Device-->Host 2D" << endl;
	vector< vector<int> > vec2D;  // Host端vector
	int *cpu_data2D;              // Host端data
	int *gpu_data2D;              // Device端data
	size_t pitch;                 // 字节对齐
	int Wd = 10;               // 宽度
	int Ht = 5;                // 高度

	cutilSafeCall( cudaMallocPitch((void**) &gpu_data2D, &pitch, sizeof(int) * Wd, Ht));  // 申请显存空间
	cutilSafeCall( cudaMemset2D(gpu_data2D, pitch, 0, sizeof(int)*Wd, Ht));               // 显存空间初始化
	cpu_data2D = (int*)calloc(Wd * Ht, sizeof(int));                                      // 申请内存空间
	if((cpu_data2D == NULL)||(gpu_data2D == NULL))   // 判断空间是否申请成功
	{
		cout << "Alloc Memery Error" << endl;
		return -1;
	}
	for(row = 0; row < Ht; ++row)   // 初始化Vector
	{
		vector<int> temp;
		for(col = 0; col < Wd; ++col)
		{
			temp.push_back(row+col);
		}
		vec2D.push_back(temp);
		temp.clear();
	}
	cout << "Vetor2D" << endl;
	for(row = 0; row < Ht; ++row)
	{
		for(col = 0; col < Wd; ++col)
			cout << vec2D[row][col] << " ";
		cout << endl;
	}
	// 将vector中的数据拷贝到Device端data
	for(row = 0; row < Ht; ++row)
	{
		cutilSafeCall( cudaMemcpy(&gpu_data2D[row*(pitch/sizeof(int))], &vec2D[row][0], sizeof(int)*Wd, cudaMemcpyHostToDevice));
	}
	// 将Device端data拷贝到Host端data
	cutilSafeCall( cudaMemcpy2D( cpu_data2D, sizeof(int) * Wd, gpu_data2D, pitch, sizeof(int) * Wd, Ht, cudaMemcpyDeviceToHost));   
	cout << "cpu_data2D" << endl;   // 打印Host端data
	for(row = 0; row < Ht; ++row)
	{
		for(col = 0; col < Wd; ++col)
		{
			cout << cpu_data2D[row*Wd + col] << " ";
		}
		cout << endl;
	}
	cutilSafeCall( cudaFree(gpu_data2D));     // 释放显存空间
	free(cpu_data2D);                         // 释放内存空间
	CUT_EXIT(argc, argv);   // 退出CUDA
};