[gmx-users] how to modify the code to generate more controllable checkpoints to restart?

Tue Jan 26 15:52:32 CET 2016

The phase space state is defined by momenta and coordinates of all objects
in a dynamical system.

Regarding what people call exact continuation, you are welcome to compare
both solutions using small runs. I doubt that the difference in any
property of interest will be observable.

https://en.wikipedia.org/wiki/Phase_space
In mathematics <https://en.wikipedia.org/wiki/Mathematics> and physics
<https://en.wikipedia.org/wiki/Physics>, a *phase space* of a dynamical
system <https://en.wikipedia.org/wiki/Dynamical_system> is a space
<https://en.wikipedia.org/wiki/Space_%28mathematics%29> in which all
possible states of a system <https://en.wikipedia.org/wiki/System> are
represented, with each possible state corresponding to one unique point in
the phase space. For mechanical systems
<https://en.wikipedia.org/wiki/Classical_mechanics>, the phase space
usually consists of all possible values of position
<https://en.wikipedia.org/wiki/Position_%28vector%29> and momentum
<https://en.wikipedia.org/wiki/Momentum> variables.

On Tue, Jan 26, 2016 at 12:39 PM, Mark Abraham <mark.j.abraham at gmail.com>
wrote:

> Hi,
>
> Vitaly's suggestion is incomplete, inasmuch as there are things like
> thermodynamic coupling algorithms that need to preserve information across
> restarts that are not x/v/f, so don't get written to the .trr file. In the
> old days, those were written to the .edr file, and the combination of .edr
> and .trr was enough for a binary-identical continuation. That's still
> possible, but the .cpt is much more convenient for the purpose. Note that
> you will generally not be able to reproduce the same trajectory, even
> though the restart is binary identical. See
> http://www.gromacs.org/Documentation/Terminology/Reproducibility. But
> anyway you were wanting to change parameters, so grompp -p new.top -t
> old.cpt is your friend.
>
> Mark
>
> On Tue, Jan 26, 2016 at 2:04 PM Yong Wang <wyongciac at gmail.com> wrote:
>
> > Hi Mark, thank you very much. It is really a good way to do it without
> > touching the code. Although as you said, it is not a perfect solution,
> but
> > I think it is sufficient to do what I want.
> >
> > Also thank you Vitaly for your comment. But I am not quite clear about
> the
> > difference between the .cpt file and the .trr file. So can I really get
> an
> > exact restart simulation from .trr file too? (here "exact" I mean the
> > restarting trajectory will be exactly the same as the original
> trajectory).
> > If so, why do we need to write the .cpt file?
> >
> > 2016-01-26 13:14 GMT+01:00 Mark Abraham <mark.j.abraham at gmail.com>:
> >
> > > Hi,
> > >
> > > There's no need to modify the code to do this. You can schedule your
> own
> > > halt e.g. 10ns from the end by choosing nsteps, save that checkpoint
> > file,
> > > and then continue. That's slightly wasteful of computer time, doing
> extra
> > > shut-down and start-up work, but you can have it working in five
> minutes
> > > ;-)
> > >
> > > Mark
> > >
> > > On Tue, 26 Jan 2016 12:22 VITALY V. CHABAN <vvchaban at gmail.com> wrote:
> > >
> > > > I suppose the TRR file, which you can write down at given frequency,
> > > > contains the same information with the same precision as it is
> written
> > > down
> > > > to the CPT file.
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > On Tue, Jan 26, 2016 at 7:50 AM, Yong Wang <wyongciac at gmail.com>
> > wrote:
> > > >
> > > > > Hi,
> > > > >
> > > > > I'd like to restart the MD simulation from some time points (e.g.
> > 10ns
> > > > > before the end) but with minorly adjusted parameters. But it seems
> > the
> > > > > current Gromacs code can't do this right now.
> > > > >
> > > > > I am trying to understand how the checkpoints in gromacs work. But
> I
> > > only
> > > > > found very limited document about it:
> > > > > "The checkpoint file is by default written to disc every 15 minutes
> > > (and
> > > > > always at the very end of a simulation), and it contains the latest
> > > > > full-precision velocities, forces, and energies that are required
> to
> > > > > continue a simulation, or restart a crashed-simulation. "
> > > > >
> > > > > Is it possible that we can control the output of checkpoint files
> as
> > > > > flexibly as other files (e.g. .xtc or .edr) by introducing some mdp
> > > > > options? More concretely, like using nstcptout=1000000 to write the
> > > > > checkpoint files every 2ns with a suffix of md.cpt_${step}. I
> guess I
> > > > have
> > > > > to do some coding, but don't know if there will be too much work.
> > Could
> > > > > anyone show some guidance on this? Thanks.
> > > > >
> > > > > Yong
> > > > >
> > > > > --
> > > > >
> > > > > ==========================
> > > > > Yong Wang
> > > > > PhD Student, Structural Biology and NMR Laboratory (SBiNLab)
> > > > > Department of Biology, University of Copenhagen, Denmark
> > > > > Ole Maaloes Vej 5, DK-2200 Copenhagen N
> > > > > https://www.researchgate.net/profile/Yong_Wang15
> > > > > --
> > > > > 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.
> > > > >
> > > > --
> > > > 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.
> > > >
> > > --
> > > 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.
> > >
> >
> >
> >
> > --
> >
> > ==========================
> > Yong Wang
> > PhD Student, Structural Biology and NMR Laboratory (SBiNLab)
> > Department of Biology, University of Copenhagen, Denmark
> > Ole Maaloes Vej 5, DK-2200 Copenhagen N
> > https://www.researchgate.net/profile/Yong_Wang15
> > --
> > 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.
> >
> --
> 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.
>