Re: TclForces and efficient PBC corrections

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Tue Nov 12 2013 - 01:32:28 CST

On Tue, Nov 12, 2013 at 8:19 AM, Norman Geist
<norman.geist_at_uni-greifswald.de> wrote:
> 1. I donít think so.
>
>
>
> 2. To get the clostest periodic distance between two atoms I usually
> do (pseudo code):
>
>
>
> $xd = pmodulo($x1-$x2+$boxX/2,$boxX) - $boxX/2
>
> $yd = pmodulo($y1-$y2+$boxY/2,$boxY) - $boxY/2
>
> $zd = pmodulo($z1-$z2+$boxZ/2,$boxZ) - $boxZ/2
>
> $dist = sqrt(pow($xd,2) + pow($yd,2) + pow($zd,2))

this last line is a *very* bad idea. pow() will be doing a call to
exp() and log(), both of which are very expensive. $x*$x however is
fast (like > 100x faster).

axel.
>
>
>
> Make sure to use the pmodulo ďpositive arithmetic remainderĒ, standard
> module canít do negative numbers.
>
>
>
> 3. IMHO tcl donít mind. Check if you can improve parts of your script.
>
>
>
> Norman Geist.
>
>
>
> Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im Auftrag
> von Morgan, Brittany
> Gesendet: Montag, 11. November 2013 19:03
> An: namd-l_at_ks.uiuc.edu
> Betreff: namd-l: TclForces and efficient PBC corrections
>
>
>
> Hi All,
>
>
>
> I use TclForces and I need to correct for periodic boundary conditions. I
> have procedures which correct for PBC brute force, but I'm having serious
> performance issues when I turn them on (the time required more than
> doubles).
>
>
>
> Can anyone tell me if:
>
> 1. For the scalar distance, does "getbond <coor1> <coor2>" return a distance
> corrected for periodic boundary conditions?
>
>
>
> 2. Is there a built-in way to obtain the vector distance between two
> positions, corrected for periodic boundary conditions? Or does anyone know
> of a resource for doing more efficient calculations in Tcl?
>
>
>
> 3. Is the IBVERBS version of charmrun faster when using TclForces?
>
> I didn't have serious performance issues with my PBC corrections previously,
> but I've recently started working on a new cluster and this problem has
> appeared. Previously, I was using NAMD 2.9b2, but IBVERBS is not currently
> working on the new cluster. I've been using NAMD 2.9-multicore instead.
>
>
>
>
>
> Any suggestions would be much appreciated.
>
>
>
> Brittany Morgan
>
>

-- 
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com  http://goo.gl/1wk0
International Centre for Theoretical Physics, Trieste. Italy.

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:58 CST