[gmx-users] number of DD cells

Justin A. Lemkul jalemkul at vt.edu
Fri Mar 4 18:02:39 CET 2011



Moeed wrote:
> Hello Justin,
> 
> Thanks.
> 
> 1- I changed the setting below because of the note I used to get:
> 
> NOTE 1 [file aminoacids.dat, line 1]:
>   The optimal PME mesh load for parallel simulations is below 0.5
>   and for highly parallel simulations between 0.25 and 0.33,
>   for higher performance, increase the cut-off and the PME grid spacing
> 

I'm guessing this pertains to your "expanded" system, before you've obtained an 
optimum density?  If so, you're only going to be losing performance for some 
simulations, but not all.  It also depends on just how bad the load imbalance 
is.  If you can set it such that you can easily specify -npme to match the 
PP:PME ratio, then you shouldn't be hurt too badly.

> Also, earlier on I read one of your messages about the trade off between 
> accuracy and speed for this setting and that no systematic study has 
> been done how exactly accuracy is changing. I actually need to work with 
> systems with 40 000 atoms or even more and electrostatics slows down my 
> runs to a great extent. Can you please give an idea on to what extent 
> the accuracy affects the result say for fourierspacing      =  0.2 . I 
> mean it affects second decimal or...? I need to go for a setting which 

You may want to look into g_tune_pme.  It should provide this sort of 
information.  There was also some discussion about this just a few weeks ago, 
but I don't have the time or desire to go rooting through the list archive at 
the moment.

> speeds up calculations as much as possible while being reasonable. How 
> would you deal with this? What value would you pick?
> 

Ideally, one that is well-tested and that you can defend to a critical audience. 
  If 0.12 is standard, and I have no desire to experiment with accuracy, I tend 
to stick with that :)

-Justin

> Thank you,
> moeed
> 
> 
>         fourierspacing      =  0.3                
> 
> 
>     You're setting yourself up for inaccurate PME calculations with
>     this.  Use 0.1 - 0.12 unless you've got some compelling reason to
>     decrease your accuracy.
> 
>     -Justin
> 
>     -- 
>     ========================================
> 
>     Justin A. Lemkul
>     Ph.D. Candidate
>     ICTAS Doctoral Scholar
>     MILES-IGERT Trainee
>     Department of Biochemistry
>     Virginia Tech
>     Blacksburg, VA
>     jalemkul[at]vt.edu <http://vt.edu> | (540) 231-9080
>     http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
> 
>     ========================================
>     -- 
>     gmx-users mailing list    gmx-users at gromacs.org
>     <mailto: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
>     <mailto:gmx-users-request at gromacs.org>.
>     Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> 
> 
> 
> 
> 
> -- 
> Moeed Shahamat
> Graduate Student (Materials Modeling Research Group)
> McGill University- Department of Chemical Engineering
> Montreal, Quebec H3A 2B2, Canada
> Web:http://mmrg.chemeng.mcgill.ca/pages/current-group-members/moeed-shahamat.php
> Web:http://mmrg.chemeng.mcgill.ca/
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================



More information about the gromacs.org_gmx-users mailing list