[gmx-developers] possible interplay between intermolecular restraints and parallelization?

[David Mobley]

Hi, Berk, Eric,

By "special" bonds, angles, and dihedrals, I just mean simple harmonic
bonds, angles, and dihedrals, as already implemented in GROMACS.

Yes, by "stability problems" I mean "unstable".

We are not running with velocity verlet here. We also do not have inter
charge-group constraints, if I remember correctly.

We have successfully verified in this system that simulations run stably to
completion when we run in serial, but we reliably get crashes in parallel.

I will submit a Redmine. I'm getting input files from my student so it may
take a little bit. Please note that you will likely have to run a while to
get the crashes -- these are not immediate; we're always able to minimize
successfully and get some distance into actual dynamics. As I mentioned
it's a stability problem.

David



On Tue, Mar 18, 2014 at 12:58 AM, Berk Hess <hess at kth.se> wrote:

>  Hi,
>
> What do you mean with "special" bonds, angles and dihedrals? Are these
> potentials you coded yourself?
> I assume that with "stability problems" you mean that things are unstable,
> not that the simulation stops with a fatal error because it can't find
> atoms in interactions.
>
> If you are using the velocity verlet integrator with inter charge-group
> constraints, that can give issues in parallel. I just uploaded a fix to
> gerrit for release-4-6.
>
> Cheers,
>
> Berk
>
>
> On 03/18/2014 07:04 AM, Erik Lindahl wrote:
>
> Hi David,
>
>  It's virtually impossible to say without a specific example, but of
> course things should work fine in parallel too.
>
>  Create a redline and upload a specific example where you get different
> results or a stability issue and describe the _exact_ conditions so it can
> be reproduced.
>
>  Cheers,
>
>  Erik
>
> --
> Erik Lindahl <erik.lindahl at scilifelab.se>
> Professor of Biophysics
> Science for Life Laboratory
> Stockholm University & KTH
> Office (SciLifeLab): +46 8 524 81567
> Cell (Sweden): +46 73 4618050
> Cell (US): 1 267 307 8746
>
>
> On Mar 17, 2014, at 6:51 PM, David Mobley <dmobley at gmail.com> wrote:
>
>     Dear Devs,
>
>  For many years we have been doing binding free energy calculations with
> restraints to hold the ligand in the binding site. Lately, however, as we
> move to bigger systems, we simply can't get these simulations to run
> stably. Does anyone have experience with using restraints between
> relatively distant atoms in parallel simulations? Any encounters with
> similar problems or suggestions?
>
>  Specifically, we normally use a set of six restraints between the ligand
> and the protein to restrain the six rigid-body degrees of freedom. These
> use three reference atoms in the protein and three in the ligand, as
> described in the reference below. Normally these atoms are relatively
> distant from one another, by design.
>
>  We have in the past mostly used distance_restraints, angle_restraints
> and dihedral_restraints for implementing these. However, these seem to lead
> to stability problems when we run on a few cores (perhaps 8) as we move to
> larger systems, and especially so when using constraints on bonds involving
> hydrogen. Additionally, since these restraints cannot be applied between
> 'molecules' in the topology, this means there are technical reasons we
> prefer another form for the restraints.
>
> Lately, we have been testing using special bonds, angles, and dihedrals
> for restraints between these atoms. However, we keep encountering stability
> problems there as well, at least when running in parallel (we are testing
> whether running on a single core will alleviate these). We DO already know
> that if we take a topology file which will run stably in serial with the
> old form of restraints, convert the restraints to the new form (bonds,
> angles, dihedrals) and attempt to run it in parallel we tend to run into
> problems.
>
>  In any case, does anyone have experience with using restraints between
> relatively distant atoms, and suggestions on how to get our systems to run
> stably? For what it's worth, these are systems which run fine in parallel
> without the restraints, and the restraints are not unusually strong. (We've
> run with stronger restraints many times in the past without issues).
>
> Thanks!
>
>  Reference:
> http://dx.doi.org/10.1021/jp0217839
>
>
> --
> David Mobley
> dmobley at gmail.com
> 949-385-2436
>
>  --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_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-developersor send a mail to
> gmx-developers-request at gromacs.org.
>
>
>
>
>
> --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_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-developersor send a mail to
> gmx-developers-request at gromacs.org.
>



-- 
David Mobley
dmobley at gmail.com
949-385-2436
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20140321/6a69607e/attachment.html>