From: Armand Sarkissian <sarkis@jonathan.ucsd.edu>
Newsgroups: comp.parallel.mpi
Subject: Implicit methods in MPI for time dependant PDE's.
Date: Fri, 18 Dec 1998 12:39:32 -0800
Organization: University of California, San Diego
Message-Id: <367ABD84.41C6@jonathan.ucsd.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


Hello Folks,


I would like to know what is the "state of the art" in solving time
dependant PDE's on a parallel architecture (using MPI). For example,
consider the diffusion equation in a square/cubic domain. There are
essentially two options: explicit or implicit time stepping. The
explicit (Euler) algorithm has the advantage of being trivial to 
parallelize but implies very small time steps for stability. On the
other hand, the implicit version is non-trivial to parallelize (IMHO)
but allows for larger time steps. Obvioulsy, there is a trade-off. I
would like to know what is the common wisdom on this subject. Namely, do
the parallelizing efforts pay off in the implicit scheme and where can I
find examples of parallel implicit codes.


Thanks,

Armand
--
Armand Sarkissian
INLS
UCSD, Mail Code 0402
9500 Gilman Drive
La Jolla CA 92093-0402
E-mail:  sarkis@physics.ucsd.edu