[gmx-users] continuing a MD simulation on different architectures

[Berk Hess]

Hi,

I added a section on Exact vs binary identical continuation on the extending simulations wiki page.

Berk

From: gmx3 at hotmail.com
To: gmx-users at gromacs.org
Subject: RE: [gmx-users] continuing a MD simulation on different architectures
Date: Thu, 2 Apr 2009 15:20:03 +0200








Hi,

If you had a computer with unlimited precision, or if you integrated the time discretized
equations of motion by hand, exact continuation would lead to identical results.
But since computer have limited precision (single or double precision) and MD is chaotic,
trajectories will diverge very rapidly if one bit (the least significant one) is rounded differently.
Such trajectories will all be equally valid, but very different.
In most cases you don't care that a trajectory does not run exactly identical when you stop
it and continue it, as opposed to keeping it running in a single run.
The most common case where you do care is when you have a crash at e.g. step 567894 and
you want to reproduce it to track down if it is a problem with your system or a bug.

A discussion like this should be added to the wiki.
I just noticed that currently there are two pages (doing restarts and extending simjulations)
that should be merged.

Berk

> Date: Thu, 2 Apr 2009 15:04:13 +0200
> From: spitaleri.andrea at hsr.it
> To: gmx-users at gromacs.org
> Subject: Re: [gmx-users] continuing a MD simulation on different architectures
> 
> Hi,
> thanks for the answer. One thing: what do you mean for "but not binary identical"?
> >From the manual mdrun says:
> "The only disadvantage of dynamic load balancing is that runs are no longer binary reproducible, but
> in most cases this is not important"
> 
> this sentence is also not very clear to me.
> 
> thanks in advance
> 
> Regards
> 
> andrea
> 
> 
> Berk Hess wrote:
> > Hi,
> > 
> > As mdrun will tell you in stderr and md.log when you try continuing,
> > you will have exact continuation, but not binary identical (which you would
> > not have anyhow if you use dynamic load balancing).
> > Exact continuation mean perfect physical and chemical reliability.
> > 
> > Berk
> > 
> > 
> >> Date: Thu, 2 Apr 2009 13:28:21 +0200
> >> From: spitaleri.andrea at hsr.it
> >> To: gmx-users at gromacs.org
> >> Subject: [gmx-users] continuing a MD simulation on different architectures
> >>
> >> Dear all,
> >> is there any problem to continue (extend) a simulation using a different architecture hardware?
> >> I have a set of simulations coming from xt4 and xt5 architecture (eepc and csc clusters) and I would
> >> like to extend the simulation step on my cluster (intel) using the previous cpt file and generating
> >> the new tpr on my cluster. I did it and it runs, however I was wondering about the physical and
> >> chemical reliability.
> >>
> >> thanks in advance
> >>
> >> Regards
> >>
> >> andrea
> >>
> >> --
> >> -------------------------------
> >> Andrea Spitaleri PhD
> >> Dulbecco Telethon Institute
> >> c/o DIBIT Scientific Institute
> >> Biomolecular NMR, 1B4
> >> Via Olgettina 58
> >> 20132 Milano (Italy)
> >> http://biomolecularnmr.ihsr.dom/
> >> Tel: 0039-0226434348/5622/3497/4922
> >> Fax: 0039-0226434153
> >> -------------------------------
> >>
> >> -----------------------------------------------------------------
> >>
> >> La tua mano puo' lasciare un segno importante.
> >> Dona il tuo 5 per mille al San Raffaele di Milano.
> >>
> >> E' SEMPLICE E NON COSTA NULLA.
> >> Basta indicare nell'apposito riquadro della dichiarazione dei redditi "Finanziamento della ricerca sanitaria"
> >> il codice fiscale della Fondazione Centro S. Raffaele del Monte Tabor:
> >> 03 06 42 80 153 e ricordarsi di firmare.
> >> Per saperne di piu': 5permille at hsr.it o vai sul sito http://www.5xmille.org.
> >> _______________________________________________
> >> gmx-users mailing list gmx-users at gromacs.org
> >> http://www.gromacs.org/mailman/listinfo/gmx-users
> >> Please search the archive at http://www.gromacs.org/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/mailing_lists/users.php
> > 
> > ________________________________
> > Express yourself instantly with MSN Messenger! MSN Messenger<http://clk.atdmt.com/AVE/go/onm00200471ave/direct/01/>
> > 
> 
> -- 
> -------------------------------
> Andrea Spitaleri PhD
> Dulbecco Telethon Institute
> c/o DIBIT Scientific Institute
> Biomolecular NMR, 1B4
> Via Olgettina 58
> 20132 Milano (Italy)
> http://biomolecularnmr.ihsr.dom/
> Tel: 0039-0226434348/5622/3497/4922
> Fax: 0039-0226434153
> -------------------------------
> 
> -----------------------------------------------------------------
> 
> La tua mano puo' lasciare un segno importante.
> Dona il tuo 5 per mille al San Raffaele di Milano.
> 
> E' SEMPLICE E NON COSTA NULLA.
> Basta indicare nell'apposito riquadro della dichiarazione dei redditi "Finanziamento della ricerca sanitaria"
> il codice fiscale della Fondazione Centro S. Raffaele del Monte Tabor:
> 03 06 42 80 153 e ricordarsi di firmare.
> Per saperne di piu':  5permille at hsr.it o vai sul sito  http://www.5xmille.org.
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/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/mailing_lists/users.php

What can you do with the new Windows Live? Find out
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20090402/91fdc547/attachment.html>