Newsgroups: comp.parallel
From: a283206@sunday.lrz-muenchen.de (Matthias Brehm)
Subject: How arrange blocking sends and receives
Organization: Leibniz-Rechenzentrum, Muenchen (Germany)
Date: 5 Sep 1995 14:42:50 GMT
Message-ID: <42hnla$7hi@usenet.srv.cis.pitt.edu>

How arrange blocking sends and receives ?

Given P(i=0,nprocs-1) processes.
Given a arbitrary number of messages each of length L(i,j) which are exchanged
between process P(i) and P(j). Typically the matrix L(i,j) is sparse, i.e.
each process sends only to a few other processes.

How can one arrange the message passing using blocking sends and receives
in a way that the communication time is minimized (e.g. no artificial
sequentiallisation due to dependancies happens) ?

+------------------------------------------------------------------------+
| Matthias Brehm                                 Phone: +(0)89-2105-8773 |
| Leibniz-Rechenzentrum der                      Fax:   +(0)89-2809460   |
| Bayerischen Akademie der Wissenschaften                                |
| Barer Strasse 21                                                       |
| D-80333 Muenchen                                                       |
| Germany                                                                |
| email (RFC-822): brehm@lrz-muenchen.de                                 |
| LRZ Home Page  : http://www.lrz-muenchen.de/Lrz/homepage_ge.html       |
+------------------------------------------------------------------------+

