Re: [AMBER] Problem in running parallel

From: David A Case <case.biomaps.rutgers.edu>
Date: Wed, 31 Oct 2012 10:46:17 -0400

On Wed, Oct 31, 2012, hari krishna wrote:
>
> I have successfully installed serial and parallel version of AMBER12 in the
> server (in my user user only) using gnu compiled openmpi. Also the test was
> successfully completed. I used ./configure_openmpi in AmberTools/src for
> openmpi.
>
> while using PBS script to run the program in parallel, in the queue, the
> program is running and processors are allocated and the program is shown
> like running in the server. But no output files are created in the given
> folder even after long time. Please look at the following PBS script. Any
> suggestions.

Error messages can get lost in queuing systems. Is you test case
an extremely short run (say 10 steps)? Can you test the exact same
(short!) run without going through the queuing system? Suppose you use
fewer threads (e.g. use 1 node with 8 ppn (and -np 8): does that make a
difference?

...etc...etc...this is debugging!

...dac

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 31 2012 - 08:00:03 PDT
Custom Search