From: George Madalin Giambasu (giambasu_at_gmail.com)
Date: Thu Jul 10 2008 - 17:00:50 CDT
from my experience with BG/L and BG/P machines you get better
performance when recompiling from the source.
also, we get better performance running in virtual mode.
our systems are ~40-60000 atoms and scaling goes very well up 512
processors in virtual mode (128 nodes).
I hope this helps,
On Thu, 2008-07-10 at 12:46 +0200, Andrew Emerson wrote:
> Dear All
> I want to use NAMD 2.6 on a Blue Gene/P machine. I have tried the
> downloadable executable but the performance and scaling are very poor,
> even with the twoAway options in the configuration file and setting
> BG_MAPPING=TXYZ (the system is also a bit small, 33000 atoms but I cant
> change that).
> Are there any other "simple" fixes which might help ?
> Assuming that the only solution is to compile it myself any suggestions
> before I dive in ?
> many thanks
> Andy Emerson
-- ________________________________________________________________________ George Madalin Giambasu PhD Student University of Minnesota Phone : (612) 625-6317 Department of Chemistry Fax : (612) 626-7541 207 Pleasant St. SE e-mail: Minneapolis, MN USA 55455-0431 GeorgeMGiambasu_at_umn.edu York Research Group giambasu_at_gmail.com http://theory.chem.umn.edu/ ________________________________________________________________________
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:49:38 CST