From: Hubert Ertl Newsgroups: comp.parallel.mpi Subject: Re: superlinear speedup in MPI? Date: Tue, 13 Oct 1998 18:43:40 +0200 Organization: Genias Software GmbH Message-Id: <3623833C.91A5785A@genias.de> References: <6vdc09$t65$1@ftp.curtin.edu.au> <6vgifo$437$1@cerberus.ee.duke.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Other application areas like domain decomposition methods not only uses the effect that the amount of local data decreases with higher parallelism, but also some additional re-partitioning or re-meshing of data is done resulting in a different (and some- times more efficient) scheme of calculations to be done. As an example you can have a look at: http://www.genias.de/projects/parasol/newsletter/issue-2/newsletter_2.html Or directly at: http://www.genias.de/projects/parasol/newsletter/issue-2/newsletter_2.html#FETI_Solver_shows If you call this effect super-linear speedup or not is up to you ;-) Hubert Ertl -- - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - Dr. Hubert Ertl, GENIAS Software GmbH, Phone: +49 9401 9200-50 *** PaTENT MPI 4.0 http://www.genias.de/products/patent/ ***