C#/C++/CLI运行效率测试之二: Native C++调用Native C++ DLL

 本文分为两方面

一.  Native C++调用Native C++ DLL 并测试运行时间

二. 完全由C# 实现运算代码 并测试运行时间

详细过程:

一:

至于如何获取EfficiencyNativeCPPDLL.dll 和 EfficiencyNativeCPPDLL.lib 以及生成项目EfficiencyNativeCPPDLL请参见前面的文章 C#通过CLR/C++调用Native CPP 类

本文讲述在VS2010中本地C++调用C++类库以进行效率测试

1. 常见Win32控制台应用程序 ConsoleCPPInvokeDLL

  

2.  选择下一步点击完成就可以了

3  对EfficiencyNativeCPPDLL.lib的调用

在项目ConsoleCPPInvokeDLL中的资源文件上\右击添加现有项可选择 EfficiencyNativeCPPDLL.lib

  

4  书写测试主体文件 

CPPInvokeDLL.cpp#include "stdafx.h"#include "EfficiencyNativeCppDll.h"#include <time.h>#define NITER 201int _tmain(int argc, _TCHAR* argv[]){int i;clock_t start, stop;EfficiencyNativeCppDll* effPro = new EfficiencyNativeCppDll();effPro->InitPositions();effPro->UpdatePositions();start=clock();for( i=0; i<NITER; i++ ) {effPro->Pot = 0.0;/*//低效率模式effPro->ComputePot();if (i%10 == 0) printf("%5d: Potential: %10.3f\n", i, effPro->Pot);*///高效率模式 ??不晓得为什么这样的速度会快那么多 快了近20倍if (i%10 == 0) printf("%5d: Potential: %10.3f\n", i, effPro->ComputePot());effPro->UpdatePositions();}stop=clock();printf ("Seconds = %10.9f\n",(double)(stop-start)/ CLOCKS_PER_SEC);delete effPro;getchar();}

测试结果:

二: C#代码实现及测试结果:

1. 

EffCompute.csusing System;using System.Collections.Generic;using System.Linq;using System.Text;namespace ConsoleEfficiencyCS{    public class EffCompute    {        public double Pot{get;set;}        private int DIMS = 3;        private int NPARTS = 1000;        private double[,] _r ;        private Random _rand;        public EffCompute()        {            _r = new double[DIMS, NPARTS];            _rand = new Random();        }               public void InitPositions()        {            for(int i = 0; i < DIMS; i++)        {        for (int j = 0; j < NPARTS; j++)        {        _r[i, j] = 0.5 + _rand.NextDouble();        }        }        }        public void UpdatePositions()        {            for (int i = 0; i < DIMS; i++)            {                for (int j = 0; j < NPARTS; j++)                {                    _r[i, j] -= 0.5 + _rand.NextDouble();                }            }        }        public double ComputePot()        {            double distx, disty, distz, dist;            double pot;            distx = 0;            disty = 0;            distz = 0;            pot = 0;            for (int i = 0; i < NPARTS; i++)            {                for (int j = 0; j < i - 1; j++)                {                    distx = Math.Pow((_r[0, j] - _r[0, i]), 2);                    disty = Math.Pow((_r[1, j] - _r[1, i]), 2);                    distz = Math.Pow((_r[2, j] - _r[2, i]), 2);                    dist = Math.Sqrt(distx + disty + distz);                    pot += 1.0 / dist;                }            }            this.Pot = pot;            return pot;        }    }    }

2.

using System;using System.Collections.Generic;using System.Linq;using System.Text;namespace ConsoleEfficiencyCS{    class Program    {        static void Main(string[] args)        {            EffCompute provider = new EffCompute();            const int NITER = 201;            provider.InitPositions();            provider.UpdatePositions();            int start = Environment.TickCount;            for (int i = 0; i < NITER; i++)            {                provider.Pot = 0.0;                //低效模式               /* provider.ComputePot();                if (i % 10 == 0)                    Console.WriteLine("{0}: Potential: \t {1}", i, provider.Pot);                 */                //高效模式                                if (i % 10 == 0)                    Console.WriteLine("{0}: Potential: \t {1}", i, provider.ComputePot());                provider.UpdatePositions();            }            int stop = Environment.TickCount;            Console.WriteLine("Seconds = {0,10}", (double)(stop - start) / 1000);            Console.ReadKey();        }    }}

3 运算结果:

  

综合以上两篇文章可见 在均是Release情况下分别对

C#调用CLI/C++转而调用 Native C++ Dll ;,本地C++调用 和 CS的.Net实现

可以看出 

本地C++调用效率最高

C#调用CLI/C++转而调用 Native C++ Dll的效率其次 可以说跟本地C++调用差不多

而CS的实现效率大约相差20倍.