From: Chris Chipot (chipot_at_ks.uiuc.edu)
Date: Wed May 13 2009 - 18:16:54 CDT
Because the dual-topology paradigm is implemented pretty deep in
the code, at the level of the non-bonded forces, with multiple
PME grids for incoming-outgoing particles, this feature will not
work in the Cuda-version. The latter can, however, be used as an
MD engine to generate configurations at λ, to which perturbation
theory can be applied by means of external scripting.
Chris
Axel Kohlmeyer wrote:
> On Wed, 2009-05-13 at 18:00 -0400, Roman Petrenko wrote:
>> By the way, is fep implemented in cuda-version of namd?
>
> check out.
>
> http://gladiator.ncsa.uiuc.edu/PDFs/accelerators/day2/session4/phillips.pdf
>
> NAMD currently puts only the most time consuming part of the
> calculation on the GPU (non-bonded forces) so the rest should
> just work as is without any special requirements.
>
> cheers,
> axel.
_______________________________________________________________________
Chris Chipot, Ph.D.
on leave from Nancy Université, CNRS
Theoretical and Computational Biophysics Group
Beckman Institute
University of Illinois at Urbana-Champaign
405 North Mathews Phone: (217) 244-5711
Urbana, Illinois 61801 Fax: (217) 244-6078
E-mail: chipot_at_ks.uiuc.edu
Christophe.Chipot_at_edam.uhp-nancy.fr
Web: http://www.ks.uiuc.edu
http://www.edam.uhp-nancy.fr
I disapprove of what you say, but I will defend to the death your right
to say it
Evelyn Beatrice Hall
_______________________________________________________________________
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:52:48 CST