I have placed some of my recent thoughts about multisander on the
amber-wiki:
http://amber.scripps.edu/cgi-bin/wiki.pl/Latest_Development_News
....thanks!....dac
Hi All,
I looked carefully at the proposal for expanding the role of
multisander
and how it would affect nudged elastic band (NEB).
It is hard to argue with the fact that it would simplify NEB, in
particular the way that igb.f was modified so that the seperate
images/replicates along the path do not interact. Also, using multisander
for NEB would allow NEB to work outside of using the generalized Born
framework.
But the big drawback to making the change is that multisander is
embedded
firmly within mpi. In the computing environment I am in, I often run many
more images along a path (~30) than there are processors available for the
calculation. In general, this is likely to be the way that many users
will
do NEB calculations. In that case, multisander would be a bad
infrastructure because multiple concurrent processes would be run on the
same processor.
One way to reconcile this might be to have a multisander
implementation
that could adapt to run each of the images that is on the same processor
in
serial. Carlos, do you have any thoughts about whether this is feasible?
Thanks,
Dave
___________________________________________________
David Mathews, MD, PhD
Assistant Professor of Biochemistry & Biophysics
Center for Human Genetics and Molecular Pediatric Disease
Aab Institute of Biomedical Sciences
University of Rochester Medical Center
601 Elmwood Avenue, Box 703
Rochester, New York 14642
Received on Wed Apr 05 2006 - 23:49:59 PDT
