> Do you know of anyone who is creating a Parallel Virtual Machine
> version of NEC, for use on a Beowulf cluster of PC's?
...
> We at BCL expect to commission our first Beowulf in a month or so,
> with a faster one later. (i.e. new PCs instead of the pile of
> 486DX4/100 machines which I have forestalled being thrown out!)
> Initial application will be our propagation tools.
>
> I would love a faster NEC !!!
Especially on parallel machines in combination with large problems, I
would urge you have a look into the matrix solver. Whereas the EM part
of NEC is too well programmed to play around with, you can gain quite
a bit of speed by replacing the matrix solving part in NEC with an
optimized for your machine (for any UNIX machine use the vendor
subroutines dxml, scilib, essl, sunperf or what their name is, on
INTELS use ASCI RED routines...) I think for large problems that's
where you can gain the most speed.
For parallel NECs, there was an article in ACES Journal vol 13 no 2
1999, pp 99 -- 106, "Running SuperNEC on the 22 procs SP2 @
Southampton Uni" by Nitch, Fourie and Reeve email
nitch_at_odie.ee.wits.ac.za
cheers
juergen
Received on Fri Apr 23 1999 - 04:33:53 EDT
This archive was generated by hypermail 2.2.0 : Sat Oct 02 2010 - 00:10:39 EDT