VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Thu Sep 20 2012 - 17:37:50 CDT

Hi,
  There isn't a parallel version of psfgen presently. I think the longest
psfgen run I'd heard of previously was about a half-day for a 100 million
atom system, including solvent box generation. If yours took over 5 days,
then I'd have to say that it sounds like some of your structure building
steps are driving psfen in a much less optimal way than expected. If you
have a structure building script that you would be willing to give us to
use to test with, we can research improving the performance of whatever
psfgen step(s) are most responsible for the long build time you describe.

Your runtimes are so much slower than what we'd seen before, that I suspect
you're triggering some psfgen behavior that either hasn't been optimized,
or that has some particularly bad algorithmic performance. Before we would
even start thinking about parallelizing psfgen, I would first like to be
sure that all of the algorithmic components are no worse than linear time
in your particular case. I suspect you have found a combination of psfgen
build steps that are resulting in quadratic runtime of one or more of the
psfgen internal algorithms, and in a case we had not run into previously.
Let us know if you're willing to give us a copy of your files and build
scripts so we can reproduce the performance problem you describe.

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Thu, Sep 20, 2012 at 05:25:29PM -0500, Ganesh Kamath wrote:
> Hi John,
> It took about 5-6 days for this system to build. The primary reason was
> that I use patch to stitch the monomers of the polymer, the regenerate
> angle and dihedral is memory intensive. Finally the water box of around 2
> million molecules just took half a day.
> Is there a parallel version of psfgen?
> I am still making tests on the system.
> Cheers,
> Ganesh
>
> On Sep 20, 2012 5:08 PM, "John Stone" <johns_at_ks.uiuc.edu> wrote:
> Hi,
> I don't recall if anyone replied to this, and I was out sick for a
> few days after your email, but it is easy to write a .js file from
> a given PSF/PDB pair by making an "all" atom selection and doing:
> $sel writejs filename.js
>
> If your structure gets too large for psfgen to handle, you'll either
> need to use the standalone version of psfgen that can write js files
> directly, or I can make a special build of VMD for you that includes
> a psfgen plugin that knows how to work with the js files. Eventually
> this will be the standard version of psfgen shipped with VMD, but
> I can't switch over to it until we get a few issues resolved for the
> Windows version of VMD/psfgen.
>
> Cheers,
> John Stone
> vmd_at_ks.uiuc.edu
>
> On Fri, Sep 07, 2012 at 09:57:39AM -0500, Ganesh Kamath wrote:
> > Hi Folks,
> > I have a syste...
>
> --
> NIH Resource for Macromolecular Modeling and Bioinformatics
> Beckman Institute for Advanced Science and Technology
> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> http://www.ks.uiuc.edu/~johns/ Phone: 217-244-3349
> http://www.ks.uiuc.edu/Research/vmd/ Fax: 217-244-6078 

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
http://www.ks.uiuc.edu/~johns/           Phone: 217-244-3349
http://www.ks.uiuc.edu/Research/vmd/       Fax: 217-244-6078