RE: [GD-Windows] Stopping a thread cleanly

[Brian S. <bs...@mi...>]

Function calls have nothing do with it.  "Assume aliasing across
function calls" is a separate setting.

This program has an aliasing problem, yet thread 2 has no function calls
in it.  See for yourself:

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

int gTrueCount, gFalseCount;

DWORD WINAPI ThreadProc(LPVOID param)
{
	bool * fooPtr =3D (bool *)param;
	while (1)
	{
		if (*fooPtr)
		{
            gTrueCount++;
		}
		else
		{
            gFalseCount++;
		}
	}
}

int main(int argc, char **argv)
{
    gTrueCount =3D 0;
    gFalseCount =3D 0;
    bool foo =3D true;
    CreateThread(NULL, 0, ThreadProc, &foo, 0, NULL);
    while (1)
    {
        foo =3D !foo;
        printf("True count: %d False count: %d\n", gTrueCount,
gFalseCount);
        ::Sleep(1000);
    }
    return 0;
}

> -----Original Message-----
> From: Jon Watte [mailto:hp...@mi...]
> Sent: Monday, February 18, 2002 10:56 AM
> To: Brian Sharon; bri...@py...; gamedevlists-
> wi...@li...
> Subject: RE: [GD-Windows] Stopping a thread cleanly
>=20
>=20
> It only needs to assume this if you're calling any other function
> from within that loop. Of course, Sleep(), or WaitForSingleObject(),
> or pretty much anything else you might want to do in the loop does
> count as a function call (as long as it doesn't get inlined) but
> I thought I'd mention it.
>=20
> volatile is your friend when you're using a shared-memory thread
> communication model.
>=20
> Cheers,
>=20
> 				/ h+
>=20
> > The compiler should always reload it unless you've specified "assume
no
> > aliasing" (which is not the default, even in optimized builds).
Since
> > it's not a local variable, it has to assume that pData->bKill can be
> > changed every time through the loop.
>=20