Newsgroups: comp.parallel.mpi
From: salo@mrjones.engr.sgi.com (Eric Salo)
Subject: Re: MPI Non blocking Sends, send buffer not readable or unclear standard?
Organization: Silicon Graphics, Inc.  Mountain View, CA
Date: 29 May 1996 03:36:19 GMT
Message-ID: <4oggnj$3ia@murrow.corp.sgi.com>

> I can't agree with this.  Standards don't come out of the blue.  The 
> whole point of something like MPI is give users something useful and 
> something they can rely on.  If there is a gratuitous and annoying (and 
> even crippling) restriction that no implementors have even felt the need 
> to exploit, it should go.

I really need a vacation, 'cuz I just remembered an implementation of MPI
that does depend on this restriction: ours! The culprit is host-to-host
messages over HIPPI. The reasons are not particularly interesting (and
may even disappear at some point), but I think that if you try to reuse
a send buffer in this case before the send request has completed, the second
message sent might end up with garbage data.

This is rather embarrassing...

Eric Salo         Silicon Graphics Inc.             "Do you know what the
(415)933-2998     2011 N. Shoreline Blvd, 7L-802     last Xon said, just
salo@sgi.com      Mountain View, CA   94043-1389     before he died?"

