Newsgroups: comp.parallel.mpi
From: jch@GS69.SP.CS.CMU.EDU (Jonathan Hardwick)
Reply-To: jch@cs.cmu.edu (Jonathan Hardwick)
Subject: Re: Query: chimp vs lam vs mpich MPI implementations
Organization: Carnegie-Mellon University, School of Computer Science
Date: 15 Feb 1995 16:34:20 GMT
Message-ID: <JCH.95Feb15113420@GS69.SP.CS.CMU.EDU>

In article <3hsji3$h5o@usenet.rpi.edu> flynnr2@alum01.its.rpi.edu
(Richard F. Flynn, Jr.) writes:
> Would anyone care to comment on or point me in the direction of any
> bodies of knowledge regarding the relative merits of the various
> MPI implementations?  I've been poking about lam and mpich mostly,
> have or am reading the docs, etc, but just wondered if a the
> collective voices of experience had any comments, sage wisdom, etc.
> 
> As an FYI, I'd be looking to use it on a bunch-o-workstations . . .

To my knowledge the only published comparison is "Performance Evaluation
of some MPI Implementations on Workstation Clusters", by Lionel Ni and
Natawut Nupairoj, presented at SPLC 94, and available on the WWW (or by
FTP) at ftp://ss36.cps.msu.edu:/pub/acstr/splc94b.ps.Z

Note that this reflects the state-of-play of MPICH, LAM, CHIMP and Unify
as of about 6 months ago, and each of the packages should have changed
(hopefully for the better!) since then.  

BTW, I'd like to issue a plea for authors of MPI-related papers to
submit BibTeX entries (and URLs where possible) to the MPI bibliography
at MSU (http://www.erc.msstate.edu/mpi/).  A centralized index makes it
much easier to track down papers.

Jonathan Hardwick, jch@cs.cmu.edu
http://www.cs.cmu.edu:8001/Web/People/jch

