[gmx-developers] rerun thermostated

[Berk Hess]

Michel Cuendet wrote:

> Hi
>
> I just found out that when you rerun a trajectory with NH thermostats 
> on, they actually rescale the velocities found in the trr file. This 
> means that the temperature and pressure measured on reruns are 
> incorrect. I used mdrun 3.3b (and grompp 3.2.1).
>
The rerun should be thermostatted, but of course in the proper way.
The 3.3 version uses a reversible NH integrator, so all the state 
information
should be present in the trr file and the scaling could work correctly.
I also thought it would be, but maybe it is not.
Are both the run and the rerun done with 3.3?

> It is probably an easy fix to disable all thermostats and barostats by 
> default with -rerun. Also, the averages are weird. nstenergy should be 
> automatically set to 1, and the number of steps taken from the 
> trajectory and not from the tpr file.

I agree.
But fixing the averages is not trivial as you don't know beforehand what 
the time
steps between frames are (they are also not necessarily constant).

Berk.