[gmx-developers] Monte Carlo Simulations

[Jared Thompson]

On Wed, Feb 1, 2012 at 11:19 AM, Shirts, Michael (mrs5pt) <
mrs5pt at eservices.virginia.edu> wrote:

> Hi, Andre-
>
> > As Michael said, the main problem would be making a consistent framework
> for
> > several MC move types and compatibility with the MD code. Furthermore,
> for
> > straight MC simulations, significant changes would need to be made to
> > energy/force evaluation routines and possibly parallelization routines
> too.
>
> There are already inner loops that are energy only, so that's not an issue
> -- that level of planning has for MC has been around for a while.  I don't
> think the parallelization issues of MC have been thought about as much, but
> the current scheme should work fine for a first pass.
>
>   I agree that a straightforward MC would be a very, almost embarrassingly
parallel problem, but it's not something we would need to sweat over at
first. I strongly agree with Michael that 5.0 needs to be cleaned and
streamlined, with a more flexible framework, when we put in a MC
integrator.


> > As to the MC moves, as I said
> > earlier, the CRA moves and simple moves for rigid body motion and
> internal
> > coordinate displacement have been implemented. I am also implementing the
> > CRRUBAR moves of Betancourt (JCP 2005, JCP 2011). There is oubviously a
> large
> > number of other MC moves for various types of molecules that could be
> > implemented, and if we are able to group of number of people working on
> this
> > it would be great.
>
> Once we get the framework in, I think people can definitely go crazy with
> the moves.  One other aspect of this is being able to validate move sets
> producing the correct ensemble. I have some code working (hopefully paper
> sent for review soon) that will be able to do this.
>
>    Agreed.  Really the move sets and the basic integrator are separate
issues. As long as an efficient description of the molecular structure is
available, then folks can easily implement any number of different moves
they want, in any order they please.





> Best,
> ~~~~~~~~~~~~
> Michael Shirts
> Assistant Professor
> Department of Chemical Engineering
> University of Virginia
> michael.shirts at virginia.edu
> (434)-243-1821
>
>
> > As to the MC moves, as I said
> > earlier, the CRA moves and simple moves for rigid body motion and
> internal
> > coordinate displacement have been implemented. I am also implementing the
> > CRRUBAR moves of Betancourt (JCP 2005, JCP 2011). There is oubviously a
> large
> > number of other MC moves for various types of molecules that could be
> > implemented, and if we are able to group of number of people working on
> this
> > it would be great.
> >
> > Cheers,
> > Andre.
> >
> >
> > ________________________________
> >  De: Jared Thompson <thompson.jared.79 at gmail.com>
> > Para: André Assunção S. T. Ribeiro <aastr at yahoo.com.br>; Discussion
> list for
> > GROMACS development <gmx-developers at gromacs.org>
> > Enviadas: Quarta-feira, 1 de Fevereiro de 2012 13:41
> > Assunto: Re: [gmx-developers] Monte Carlo Simulations
> >
> >
> > Andre,
> > MC/MD is good, but I was thinking more of a robust, state-of-the-art MC
> > simulation platform for GROMACS. I would like to have something like
> this for
> > my Ph.D. thesis.
> >
> > ~j
> >
> >
> > 2012/1/31 André Assunção S. T. Ribeiro <aastr at yahoo.com.br>
> >
> > Hi,
> >>
> >>
> >> Some time ago I sent an email saying that the CRA moves of Ulmschneider
> and
> >> Jorgensen had been implemented in 4.5.3. Since then we have used this
> code in
> >> combination with regular MD and the results seem promising. A paper
> reporting
> >> our earlier results has been recently published
> (DOI 10.1002/jcc.22925). The
> >> code is still based on 4.5.3, but I could put this into the latest
> >> development version, if that interests anyone.
> >>
> >>
> >> Cheers,
> >> Andre.
> >>
> >>
> >>
> >> ________________________________
> >> De: "Shirts, Michael (mrs5pt)" <mrs5pt at eservices.virginia.edu>
> >> Para: Discussion list for GROMACS development <
> gmx-developers at gromacs.org>
> >> Enviadas: Terça-feira, 31 de Janeiro de 2012 22:31
> >> Assunto: Re: [gmx-developers] Monte Carlo Simulations
> >>
> >>
> >> Hi, all-
> >>
> >> MC is definitely an interest of mine, and I think I've already
> volunteered
> >> to help organize more MC functionality on the core developer side.
> >>
> >> Any MC functionality will be part of 5.0 going forward, and will not be
> in
> >> any 4.0 version.  The most likely scenario (there's no settled plan
> yet) is
> >> to have MC be the parent integrator, and have MD as essentially a
> >> rejectionless MC variant; this will allow things like hybrid MD/MC, etc.
> >> The integrator code is too complicated right now, and this will likely
> be
> >> part of a larger effort to make the integrators simpler in 5.0.
> >>
> >> Right now all
> >  the developers are concentrating very hard on getting 4.6 out
> >> the door, so there is not much time to think about it new 5.0
> features.  But
> >> definitely ping me about this as soon as you hear about 4.6 being
> released,
> >> and we'll start diving in.
> >>
> >> The hard part is not WRITING the MC code, it's fitting it into the rest
> of
> >> the code in a smooth way.
> >>
> >> Best,
> >> ~~~~~~~~~~~~
> >> Michael Shirts
> >> Assistant Professor
> >> Department of Chemical Engineering
> >> University of Virginia
> >> michael.shirts at virginia.edu
> >> (434)-243-1821
> >>
> >>
> >>> From: Jared Thompson <thompson.jared.79 at gmail.com>
> >>> Reply-To: Discussion list for GROMACS development
> >>> <gmx-developers at gromacs.org>
> >>> Date: Tue, 31 Jan 2012 19:00:32 -0500
> >>> To: <gmx-developers at gromacs.org>
> >>> Subject: [gmx-developers] Monte Carlo Simulations
> >>>
> >>> Hi All,
> >>>
> >>> There has been some discussion in the past with regards to developing a
> >>> Monte Carlo integrator for the gromacs platform. I recently downloaded
> and
> >>> poked through 4.5.5 and I can't find anything in it that looks like a
> MC
> >>> integrator function (I'm thinking of do_md). Does anyone have an
> update on
> >>> that or what? Does anyone feel like helping me develop a nice MC
> scheme in
> >>> Gromacs?
> >>>
> >>> ~j
> >>> --
> >>> gmx-developers mailing list
> >>> gmx-developers at gromacs.org
> >>> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> >>> Please don't post (un)subscribe requests to the list. Use the
> >>> www interface or send it to gmx-developers-request at gromacs.org.
> >>
> >> --
> >> gmx-developers mailing list
> >> gmx-developers at gromacs.org
> >> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> >> Please don't post (un)subscribe requests to the list. Use the
> >> www interface or send it to gmx-developers-request at gromacs.org.
> >>
> >>
> >>
> >> --
> >> gmx-developers mailing list
> >> gmx-developers at gromacs.org
> >> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> >> Please don't post (un)subscribe requests to the list. Use the
> >> www interface or send it to gmx-developers-request at gromacs.org.
> >>
> > --
> > gmx-developers mailing list
> > gmx-developers at gromacs.org
> > http://lists.gromacs.org/mailman/listinfo/gmx-developers
> > Please don't post (un)subscribe requests to the list. Use the
> > www interface or send it to gmx-developers-request at gromacs.org.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20120201/95b87376/attachment.html>