Edward d'Auvergne wrote:
I'll have a play around with the code and respond in more detail later
on. I can see three different ways of multi-CPU/machine etc. use:
Standard threading - a single machine with either multiple CPUs or a
multi-core CPU.
MPI - running on a cluster of identical machines.
Grid computing - running on multiple networked machines of different
speeds, operating systems, etc. (i.e. the most typical setup for an
NMR lab).
Hi Ed
This will all fit into the design because as you can see the startup
and the multiprocessor setup are modular. My thought was to impliment
the threads and mpi versions. I had a look
Each three of these methods could possibly use the same underlying
threading mechanism for the calculations (assuming that parent MPI
process will benefit from threading). I'm not saying you should
consider implementing these - just to take these into consideration as
possible future options when designing the MPI stuff.
Cheers,
Edward
P.S. For easier porting back to the 1.3 line it may be best to avoid
a large number of small, scattered changes to the current code base.
The more the new code is external to the current code, the easier the
job will be.
Of course I will try to do that but only upto the point where the
design stays good. Their will obviously need to be some changes to the
code that carries out calculations at the top level, but these should be
fairly isolated as well. The whole idea is to prototype anyway and then
if it looks good incorporate it I guess. If I have to recode some bit so
be it..
regards
gary
--
-------------------------------------------------------------------
Dr Gary Thompson
Astbury Centre for Structural Molecular Biology,
University of Leeds, Astbury Building,
Leeds, LS2 9JT, West-Yorkshire, UK Tel. +44-113-3433024
email: garyt@xxxxxxxxxxxxxxx Fax +44-113-2331407
-------------------------------------------------------------------
Re: r3208 - /branches/multi_processor/relax