[gmx-users] Exact continuation run

[Berk Hess]

>From: "Mark Abraham" <mark.abraham at anu.edu.au>
>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] Exact continuation run
>Date: Sun, 16 Sep 2007 12:33:52 +1000 (EST)
>
> >
> > Hi
> > I have been running exact continuation runs using tpbconv along with the
> > .edr file.  I lost the .edr file to the last run
> > but I do have the .log file with energy parameters from the same run.  
>Is
> > there some way to get an exact
> > exact continuation run with data from the .log file?  I am using 
>pressure
> > coupling.
>
>No - the .log file output is at very different precision. It won't be so
>different that you enter a different ensemble and need to re-equilibrate,
>but you will not generate the same trajectory that a real exact restart
>would do. More pertinent is that there is no method to take those values
>and put them in an .edr file, or to use anything but an .edr file. So your
>only option is to restart from the last point you have all the information
>for. Take care of your data!
>
>Mark

Indeed.

But note that exact continuation is completely irrelevant in most cases.
Without an exact restart your results will not be incorrect.
The main use of it for a normal user is if you are looking for rare events
in very long simulations. In that case you might not want to generate
frequent output, but in case something has happend you want to redo
part of the run (exactly) with more frequent output.
(Or in case you lost some data in the middle of a long series of simulations
and you want to rerun it to regenerate the data.)

Berk.

_________________________________________________________________
Play online games with your friends with Messenger 
http://www.join.msn.com/messenger/overview