Newsgroups: comp.parallel.mpi
From: salo@mrjones.engr.sgi.com (Eric Salo)
Subject: Re: MPI_TEST
Organization: Silicon Graphics, Inc.  Mountain View, CA
Date: 24 Dec 1996 21:30:02 GMT
Message-ID: <59pi0q$q6n@murrow.corp.sgi.com>

Thanks to everyone who sent in a reply. I've received some very
interesting examples, so I thought I would summarize them here. It
looks like there are at least two classes of codes for which MPI_TEST
is truly useful:

1) Codes that involve non-MPI asynchronous events, such as I/O or X11.
For codes such as this, it is not acceptable to block in any one place
while waiting for an interesting event to occur.

2) Codes with dynaminc load-balancing. Each process has some big chunk
of work to do which it breaks up into smaller chunks. When a process
runs out of work to do, it asks one of its neighbors to donate a small
chunk. These requests for more work may or may not occur, so it is not
acceptable to block while waiting for them.

Thanks again for the education.

Eric Salo         Silicon Graphics Inc.             "Do you know what the
(415)933-2998     2011 N. Shoreline Blvd, 8U-802     last Xon said, just
salo@sgi.com      Mountain View, CA   94043-1389     before he died?"

