Re: collective operations in the dynamic chapter
Eric Salo (salo@mrjones.engr.sgi.com)
Tue, 28 Jan 1997 23:59:55 -0800
Bill, thanks for the summary and analysis. I agree with just about all of it.
Regarding IACCEPT, I'm not entirely sure that we even need a non-blocking
version. As has been pointed out many times already, we aren't *really*
providing a true client/server interface, and it seems highly unlikely that
there will be many codes in which one "server" must be prepared to accept
multiple connection requests from different sets of MPI "clients". Instead, I
believe that the primary users of this stuff will be single MPI jobs which just
happen to consist of seperable parts, such as a compute module plus an optional
visualization module.
Clearly, a purely blocking ACCEPT would not be a good solution for a compute
module, since it is not guarenteed that there will ever be a matching CONNECT.
But what if we were to add an IPROBE-like version which didn't actually do
anything other than inform the calling process(es) whether there was an
outstanding CONNECT? It might or might not make sense to define a LISTEN-like
routine, since MPI could just reserve a buffer internally to hold that state.
--
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?"