[gmx-users] continuation options

[Daniel Larsson]

On Nov 11, 2008, at 15:39 , Berk Hess wrote:

> > With Gromacs 4.0 I get this message:
> >
> > NOTE: The simulation uses pressure coupling and/or stochastic  
> dynamics.
> > tpbconv can not provide binary identical continuation.
> > If you want that, supply a checkpoint file to mdrun
> >
> > Is it possible to extend a trajectory with mdrun, a checkpoint file
> > and a new tpr-file? E.g:
> >
> > tpbconv -s original.tpr -f original.trr -e original.edr -o
> > extension.tpr -extend 10000
> > mdrun -s extension.tpr -cpi original.cpt
>
> Simply do not supply the trr and edr to tpbconv.
> mdrun does not read the state from the tpr file, but from the  
> checkpoint file
> when you give the -cpi option to mdrun.

But my example above would also be valid, or? That is, mdrun would  
take the state from the checkpoint file and the ".mdp- 
parameters" (tinit, nsteps,  etc.) from the .tpr-file?

Is this the best way to perform a continuation/extension of a  
simulation?

> > Wouldn't it be nice if one could feed tpbconv with a checkpoint file
> > to generate a tpr file?
>
> Yes, but then the tpr file should contain the whole state.
> You now never need to create a new tpr file,
> unless you want to make the run longer than you orginally thought.

Perhaps it would be more intuative if tpbconv could take a checkpoint  
file as input and generate a new .cpi-file as starting point for  
mdrun? Or extend the tpr file format so that it can also include  
system stat variables.

Also, why does mdrun override the naming flags for files when using a  
checkpoint file to start? It always adds ".part000X" before the file  
extension.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20081111/23db0134/attachment.html>