[gmx-developers] Restarting with Parrinello-Rahman coupling

[Berk Hess]

Francesco Mercuri wrote:

>Hi all,
>I have a couple of questions about restarting runs in NpT simulations with Parrinello-Rahman coupling.
>1) Which are the variables actually read from the .edr file for restarting, concerning the Parrinello-Rahman barostat?
>2) Would it be possible to "plug" arbitrary values of these variables in a .edr file? This could be needed, for example, if you want to exactly restart a run from a previous one, but you need to change the barostat variables. If this option is not already implemented in the tools suite of gromacs, it would require some hack of parts of the code, I guess. Any help would be really appreciated.
>Thank you in advance and best regards,
>
>  
>
The variables for Parrinello-Rahman in the energy file are the box 
velocities.

Exact restarts are there so that several shorter runs will give binary 
identical
results to one long run. Or to exactly redo part of a run.
When you change a barostat variable the concept of exact restarts does 
not apply.

Berk.