[gmx-users] particle decomposition with distance restraints

[Mark Abraham]

That does indeed sound like a bug now. The combination of PD, distance
restraints and lincs has likely never been tested :-( Sadly, there are far
too many combinations of algorithms to ever test. Please file an issue at
redmine.gromacs.org, and please include .tpr files with and without lincs
(and preferably with and without distance restraints). Also the .top + .gro
+ .mdp in case we need to rebuild the .tpr for some reason.

Mark

On Fri, Dec 14, 2012 at 3:25 PM, Da-Wei Li <lidawei at gmail.com> wrote:

> Dear Mark
>
> Here it is some information that may be useful. I checked all my -pd, with
> distance restraints MD runs. It seemed that crash usually happened when
> protons and the carbon they attached were in different processors.
>
> best,
>
> dawei
>
>
> On Thu, Dec 13, 2012 at 4:12 PM, Da-Wei Li <lidawei at gmail.com> wrote:
>
> > Dear Mark
> >
> > Thanks for the quick response. This has bothered me for some time. I feel
> > something is wrong because
> >
> > 1. If I turn off lincs (dt will be 0.001 instead of 0.002 ), I can run
> > the  same simulation for very long (>500 ns) without any problem. By
> saying
> > "same simulation", I mean everything is identical except dt (=0.001) and
> > constraints (= none). Once I turned on lincs  (even with dt is still
> 0.001)
> > by set constraints to hbonds, the md run will crash from time to time.
> >
> > 2. With the help of cpt file (it is a numerical exact re-start because I
> > don't change anything and I use the same number of processors), I usually
> > can pass the last time crash point without error.
> >
> > best,
> >
> > dawei
> >
> >
> >
> > On Thu, Dec 13, 2012 at 12:17 PM, Mark Abraham <mark.j.abraham at gmail.com
> >wrote:
> >
> >> 1) AFAIK there are indeed conditions under which domain decomposition
> does
> >> not work with distance restraints
> >> 2) Your symptoms could just be that (some of) your initial conditions,
> or
> >> them in combination with the restraints are problematic. Without more
> >> description about why and how you're using the restraints, and that your
> >> systems are stable in the absence of restraints, it's hard to say much.
> >> 3) Generally, turning off the restraints is not a proper solution (and
> >> particularly not without adjusting the time step accordingly) - see
> >> http://www.gromacs.org/Documentation/Terminology/Blowing_Up
> >> 4) We're a long way from supposing there's a code bug :-)
> >>
> >> Mark
> >>
> >> On Thu, Dec 13, 2012 at 4:40 PM, Da-Wei Li <lidawei at gmail.com> wrote:
> >>
> >> > Dear Gromacs users
> >> >
> >> > I recently tried distance restraints simulation and  I have been
> forced
> >> to
> >> > use -PD (particle decomposition) because domain decomposition doesn't
> >> work
> >> > with distance restraints. However, my MD run will crash randomly for
> >> some
> >> > of my proteins but works well for others. Once it is crashes, I try to
> >> > rescue it by restart it using CPT file. Some times, the simulation can
> >> pass
> >> > the crash point and continue for a long time and some times it just
> keep
> >> > crashing. The only work-around I found is to turn off lincs. Does this
> >> mean
> >> > a bug in the code? I observed this for 4.5 and 4.6.
> >> >
> >> > best,
> >> >
> >> > dawei
> >> > --
> >> > gmx-users mailing list    gmx-users at gromacs.org
> >> > http://lists.gromacs.org/mailman/listinfo/gmx-users
> >> > * Please search the archive at
> >> > http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> >> > * Please don't post (un)subscribe requests to the list. Use the
> >> > www interface or send it to gmx-users-request at gromacs.org.
> >> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >> >
> >> --
> >> gmx-users mailing list    gmx-users at gromacs.org
> >> http://lists.gromacs.org/mailman/listinfo/gmx-users
> >> * Please search the archive at
> >> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> >> * Please don't post (un)subscribe requests to the list. Use the
> >> www interface or send it to gmx-users-request at gromacs.org.
> >> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >>
> >
> >
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>