[gmx-users] multiple time step

[Berk Hess]

Hi,

Our philosophy up till now is that with bond constraints and hydrogens
replaced by virtual sites one can reach a time step of 4 or 5
femtoseconds. This is roughly equal to the largest time step, used for
the PME mesh part, in multiple time stepping. But we then do less
work on the non-bonded and bonded interactions. Therefore we think
that out aproach is somewhat more efficient.
But we are thinking about implementing multiple time stepping as well.
I already implemented proper reversible multiple time stepping for
the typical Gromos twin-range non-bonded force setup for the 4.5
release.

Berk

> Date: Wed, 30 Jun 2010 01:41:01 -0400
> From: chris.neale at utoronto.ca
> To: gmx-users at gromacs.org
> Subject: [gmx-users] multiple time step
> 
> I recall seeing something online about how gromacs developers have  
> decided to focus on increasing the overall speed and allowing  
> generally large timesteps (via e.g. angle constraints) vs.  
> implementing multiple timestepping (no mailing list ref. sorry). I  
> agree that this is not a logically exclusive decision -- PME takes  
>  >20% of the time so if they were doing it only 1/2 as often then  
> there is the possibility of a real speedup. Nevertheless, it would  
> need to be tested and perhaps the developers have indeed run these  
> tests -- I've simply not seen the results. One would think, though,  
> that PME could nearly always be done less often than, for example,  
> bonded interactions.
> 
> I agree with Erik that the work would be large, but I disagree that  
> the benefits would be small -- although I suspect that neither of us  
> has any data to support such claims :).
> 
> Unfortunately, the bottom line is that this is free software and if  
> you want it then you can code it. Or at the very least, you should  
> benchmark the benefits on NAMD and then submit an "enhancement"  
> bugzilla.
> 
> Chris.
> 
> 
> -- original message --
> 
> The core developers have the answer for this one, but I can make an
> educated guess:
> 
> Implementing it would mean a LOT of work and the rewards are small. The
> latter because most particles will have rougly the same oscillation
> period if one uses all-atom forcefields and turn on virtual sites and
> constrains the bonds, hence the point of having multiple time steps is lost.
> 
> Vitaly Chaban skrev:
> > Chris,
> >
> > An interesting question...
> >
> > BTW, is there any philosophy of gromacs developers to avoid this
> > algorithm in the MD engine?
> >
> > Vitaly
> >
> >
> >
> >
> >> multiple timesteps are not possible as far as gromacs 4.0.7. NAMD  
> >> can do this.
> >>
> >>
> >> -- original message --
> >>
> >> Is it possible to carry out  multiple time step  molecular
> >> dynamics simulations
> >> in Gromacs
> >> 4.0. versions ? Could you
> >> please give me some information about this issue ?
> >>
> >> Thank you very much for your attention.
> >>
> 
> 
> -- 
> 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/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! 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/20100630/14eba27b/attachment.html>