From: Robert Brunner (rbrunner_at_uiuc.edu)
Date: Tue Jan 18 2005 - 11:05:00 CST
Hello Bora,
If your job always stops in _exactly_ the same place, its unlikely that
its a hardware or low-level communications problem. Usually those sorts
of things are triggered by background activity, so tend to not occur at
exactly the same point every time.
A 14K atom simulation should run on a 16 CPU system, it just won't run
very efficiently. Changing the number of processors might cause the
crash to disappear, but that would just be hiding the problem, not
fixing it.
I don't have an immediate solution for you, but Brian's advice about
trying ++debug and/or memory_paranoid sounds like a good start.
Robert
On Jan 16, 2005, at 3:59 AM, bora erdemli wrote:
> Hi Brian;
>
> thank you for your consideration once again...
>
> my system did not create any core when simulation was
> stopped.
>
> I use "qsub" in order to submit my jobs.
>
> and my jobs are always stopped at the same spot.
>
> I have never run my jobs with ++debug switch and/or
> memory_paranoid.
>
> As far as I uderstood from your e mails, there is an
> hardware problem . I tried to it with lower number of
> cps and nodes.
>
> Thank you for your suggestions.
> If you have further ones, I would be appreciated..
>
> best Regards
>
> Bora
>
>
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:39:07 CST