Re: NAMD CUDA

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Thu Mar 11 2010 - 03:50:25 CST

On Thu, Mar 11, 2010 at 4:11 AM, bahare bamdad <baharebamdad_at_gmail.com> wrote:
> Dear all,
>
> I am running a job using NAMD CUDA, but this stops in a special step,
> although the program still works (when I use top command I can see namd2). I
> use the following command to run namd:
> ./charmrun ++local +p8 ./namd2 +idlepoll +devices 0,2,3 <configfile>
> Can I ask you what happend?

you can ask, but without reporting the output text *where* it stops
and - even better - providing the necessary inputs to reproduce
this scenario, it is close to impossible to give any advice.
the fact that you see namd2 executables in top doesn't mean anything.
they are just polling to wait for a message, they are likely not doing
anything.

what seems odd is that you start 8 copies of NAMD
while you have only 3 GPUs. i would first test with 3
instances of NAMD. depending on the size of your
problem, you may get some additional speedup by
going to 6 instances, but more is likely to slow you
down. do you have by any chance configured your
gpus to run in compute-exclusive mode?

cheers,
   axel.
>
> Thanks in advance,
>
> B. Bamdad
>

-- 
Dr. Axel Kohlmeyer    akohlmey_at_gmail.com
http://sites.google.com/site/akohlmey/
Institute for Computational Molecular Science
Temple University, Philadelphia PA, USA.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:53 CST