Newsgroups: comp.parallel
From: Joerg Eisenreich <je@ziam.rmi.de>
Subject: Q: Domain Decomposition for parallel FD Codes
Organization: RMI Net - EUnet EUregio POP Aachen
Date: Thu, 12 Jan 1995 18:38:05 GMT
Message-ID: <3f3mji$509@infoac.rmi.de>

I am currently working on my master thesis about the
parallelisation of Finite Difference codes for the simulation
of electromagnetic waves (Maxwell equations). A well-known
approach is domain decomposition, i.e. to divide the problem
space grid into cubes and distribute them onto the processors.

My implementation base is (probably) PVM or some PD MPI
implementation, I'm still not sure which to take. Target hardware
will be some ethernet-connected SUNs and Parsytec PowerXplorers
or PowerGC, both running with Motorola PowerPC 601 chips.
Both Parsytec machines have a 2D grid, virtual links can be
established using this grid. The routing is done by 2 Inmos T805
per PPC 601.

It would be nice to have some pointer or hints from The Net.
Please MAIL it to me (je@ziam.rmi.de), I will make a summary
and post it to this news group.

Thanks in advance and best regards,

Joerg Eisenreich



