[gmx-users] how to continue simulation?

Berk Hess gmx3 at hotmail.com
Thu Jul 8 15:45:04 CEST 2004




>From: Erik Lindahl <lindahl at csb.stanford.edu>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: Re: [gmx-users] how to continue simulation?
>Date: Thu, 8 Jul 2004 14:46:34 +0200
>
>Hi,
>>
>>Do also check the queue system for automatic continuation of jobs. With
>>PBS you can let a job start as soon as the previous one has finished.
>>
>
>Just to add to what David said: If the job manager sends a SIGHUP (kill -1) 
>30 seconds before it really kills the job, Gromacs will save the current 
>frame in the trajectory and finish in a clean way. If this works you don't 
>have to make sure the simulation is shorter than 24 hours.
>
>You can then add a tpbconv command in your job script and just have Gromacs 
>do automatic continuations indefinitely.

The only drawback here is that it does not necessarily terminate on
a step number which is a multiple of nstenergy, nstxout and nstxtcout.
With the current tpbconv the new run will write at the correct intervals,
but when concatenating files you would need to remove an energy
and a trajectory frame to get correct results with some of the analysis
tools.

Berk.

_________________________________________________________________
Play online games with your friends with MSN Messenger 
http://messenger.msn.nl/




More information about the gromacs.org_gmx-users mailing list