[gmx-users] What's the practical effect of 'continuation = yes/no'?

[Mark Abraham]

Hi,

On Tue, Sep 22, 2015 at 8:22 AM Trayder Thomas <trayder.thomas at monash.edu>
wrote:

> What exactly does 'continuation = yes' do?
>

Tells mdrun to behave as though the simulation didn't stop :-) We need a
way to let people stop the simulation, perhaps change something, and
continue on with the minimum disruption necessary.

For better or worse, mdrun defaults to doing a call to the constraint code
before entering the MD loop, which perhaps helps with some cases of
starting new simulations from states that don't belong in the ensemble.
Thus this .mdp option.

My understanding thus far is that it will move particles in the first step
> (using the input velocities) before applying the constraints (LINCS bond
> lengths etc.). When continuing a simulation with 'continuation=no' you
> would apply constraints twice in a row.
>

Yes - per Fig 3.3 of the reference manual, except that constraints happen
either before the loop enters, or after the update.

So when using pull code, 'continuation=no' would give you a double-pull,
> which would be bad.
>
> But if you were dealing with, for example, only LINCS/SHAKE constraints,
> would 'continuation=no' make a practical difference?
>

No difference in typical practice. Obviously you remove yourself even
further from the possibility of a binary reproducible trajectory.

The manual states "The relative constraint deviation [for LINCS after one
> iteration] will be less than 0.0001 for every constraint". Which suggests
> to me that the error in applying a second constraint step would be less
> than continuing from a .gro file.
>
> Is my understanding correct?
>

Seems good.

Mark


> Thanks,
> -Trayder
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>