Last night I was toying a bit with the performance difference between managed and unmanaged code. Here I am refering to C# and C++. To put it simply I was pretty shocked with the difference, and wanted to share this experience and perhaps get enlightened if I am doing something wrong.
My testing application was calculating up to a million primenumbers. My testing machine was a Intel i-7 920, overclocked to 3.81Ghz.
Disclaimer: I didn't do any algorithm optimization of any sort. The algorithm is taken from a previous stackoverflow topic. The code is 100% identical, and I have not taken any parallel advantage on purpose, to make the test equal. Please taken in account I am not a C++ programmer, it is my first time playing with C++. I mainly only do C# and Java. I wrote the C++ part as a DLL file, which I will invoke in C# with Pinvoke.
C++:
#include <stdio.h>
extern "C"
{
__declspec(dllexport) long FindPrimeNumbers(int n)
{
int count = 0;
long a = 2;
while(count < n)
{
long b = 2;
int prime = 1;
while(b * b <= a)
{
if(a % b == 0)
{
prime = 0;
break;
}
b++;
}
if(prime > 0)
count++;
a++;
}
return (--a);
}
}
C#:
static long PrimeNumbers(int n)
{
int count = 0;
long a = 2;
while (count < n)
{
long b = 2;
int prime = 1;
while (b * b <= a)
{
if (a % b == 0)
{
prime = 0;
break;
}
b++;
}
if (prime > 0)
count++;
a++;
}
return (--a);
}
Please remember, I will compile the C++ method in C#.
[DllImport("myDLL.dll", CallingConvention = CallingConvention.Cdecl)]
public static extern long FindPrimeNumbers(int n);
static void Main(string[] args)
{
const int numbersToFind = 1000000;
Stopwatch sw = new Stopwatch();
sw.Start();
PrimeNumbers(numbersToFind); //C#
sw.Stop();
Console.WriteLine(sw.Elapsed);
sw.Restart();
FindPrimeNumbers(numbersToFind); //C++ Pinvoke
sw.Stop();
Console.WriteLine(sw.Elapsed);
Console.ReadLine();
}
There is a HUGE difference in the result, between these two codes:
C# = 00.00.32.6975885
C++ = 00.00.10.5427109
I know there can be made optimization for the algorithm such as using the parallel for loop provided by the .NET, but this is not the point. I just want to know, if this is a normal performance behavior between unmanaged and managed code? Or that I have maybe skipped something? I tried to search on Google yesterday, about the performance difference between C# and C++ and there was so many topics where people could swear by their graves, that the performance between these two languages are identical.
