(OpenMP, C++ME, System.Threading.Thread) -> Resource Leak.

[PimpDaddy]

Here is the source code of my C++/ME app. When I run it I can use "Windows Task Manager" to see that thread HANDLEs are leaking.

#include "stdafx.h"
#include <omp.h>

public __gc class TestClass
{
public:
	void ThreadProc()
	{
		int sum = 0;
	
#pragma omp parallel for reduction(+:sum)
		for ( int i = 0; i < 10000; i++ )
		{
			sum += i;
		}

		System::Console::WriteLine( sum.ToString() );
	}
};

int main()
{
	for (;;)
	{
		TestClass __gc * o = __gc new TestClass();
		System::Threading::Thread __gc * t = __gc new System::Threading::Thread( 
                    __gc new System::Threading::ThreadStart( o, &TestClass::ThreadProc ) );

		t->Start();

		System::Threading::Thread::get_CurrentThread()->Sleep( 100 );
	}
}

If I put something that is not using Open MP into the ThreadProc no HANDLEs will be leaked.
I realize that what happens is: Omp creates unmanaged threads inside a managed thread (i.e. System::Threadi ng::Thread). I just have very little idea about what could be wrong with it. If somebody will try to reproduce this, I am currently using VS2005.
Thanks for all your comments and suggestions.

[gnanapoongothai]

I think i have got ur problem. there are unmanaged threads when using openmp.is it so?

openmp architecture is like when u specify the openmp pragma, it creates a new thread for ur loop and once finished joins the main thread.this might be ur problem. I am not versatile in openmp. But this is a new concept only less ppl use it.so keep in touch to discuss about openmp

[PimpDaddy]

We recently discovered that the thread leak has nothing to do with .NET. The same leak occurs even if we run OpenMP code inside a Win32 thread. So, here is the new example. Suggestions\ide as, please?

#include <windows.h>
#include <stdio.h>
#include <omp.h>

int calculateArithmeticalProgressionSumOMP( int a1, int d, int n )
{
	int sum = 0;
	omp_set_num_threads(3);
#pragma omp parallel 
	{
#pragma omp for reduction (+: sum)
		for( int i = 0; i < n; i++)
		{
			int a = a1 + i * d;
			sum += a;                                                                                                                                                                                      
		}
	}
	return sum;
}

DWORD threadFunc(LPVOID pParams)
{
	int sum = calculateArithmeticalProgressionSumOMP(0, 1, 10000);
	printf("%d\n", sum);
	return 0;
}

int main()
{
	//create some threads
	for(int i = 0; i < 100; i++)
	{
		HANDLE handle;
		handle = CreateThread(NULL, 0, (LPTHREAD_START_ROUTINE )threadFunc, NULL, 0, NULL);
		WaitForSingleObject(handle, INFINITE);
		Sleep(100); //imitating some other work
	}

	return 0;
}