[Interest] Program runs only on four core processors....

[Etienne Sandré-Chardonnal]

>> Hi Guido
>>
>> It's very hard to make a guess with so little information.
>
> I know. But currently I don't have more and since there are quite a few experienced programmers here, I hoped someone would have experienced something similar. :-)
>
>> If the number of cores appears to affect the outcome of the execution, it
>> usually means a threading problem (where "threading" here can be across
>> processes too). If you increase the number of cores and it stops working, it's
>> usually a race condition.
>
> *sigh* This will be hell to find. Especially since I don't have the slightest idea where a race condition could be. And even worse, even if all communication of the main program with the other three programs stops, it should not freeze the whole process. More or less I send a few images via tcp to a couple of services, which process them and send them back. If the images never come back, the visible data in the main program isn't updated correctly, but thats all I'd expect.
>
> Any chance that the non-blocking Qt network classes block? Never seen or heard of it, but this would explain the behaviour.
>
>> Since decreasing the number of cores makes it work,
>
> Nope, increasing makes it work. But apparently not always. Or not for all machines. Not sure, which of the two scenarios you described I should prefer now. Even if I had the choice. ;-)
>
> Thanks,
> Guido


Changing the number of cores should not freeze the program because of
blocking classes! Even on 1 core it's possible to run many threads
together.


There is one thing that may change with the number of core : the
result of QThread::idealThreadCount() if it's used somewhere. Also,
poorly using QThreadPool or QtConcurrent may be affected by the number
of cores, at it uses it in QThreadPool::maxThreadCount()
Try doing QThreadPool::globalInstance()->setMaxThreadCount(1) at the
beginning of your app to check if it reproduces the issue even on 4
cores or more