Newsgroups: comp.parallel.pvm
From: sits@cs.anu.edu.au (David Sitsky)
Subject: MPI_Bcast problems?
Keywords: MPI
Organization: Australian National University
Date: 14 Jun 1994 07:13:45 GMT
Message-ID: <2tjlb9INNhgv@dubhe.anu.edu.au>

I know this may not be the appriopiate newsgroup, but since comp.parallel.mpi
doesn't exist yet, here goes.

I'm aware of an application that is going to be built in MPI, but some 
problems have been raised.  In the application, a process blocks until
a message arrives, where it services the request and then blocks waiting
for another message to arrive.

Message are passed to this process using ordinary point to point calls,
however it is sometimes necessary for another processes to broadcast a "request"
to all the other processors in a group.

In MPI, messages can only be broadcasted using MPI_Bcast, where all processes in
the group have to call it.  This doesn't really fit the application, where the
process doesn't know in advance when a broadcasted message is going to be sent.

Although a non blocking MPI_Bcast operation may solve this problem, I'm just
curious why the MPI standard doesn't include a function MPI_Mcast which
has the same semantics as pvm_mcast (ie non-collective call but message appears
as an ordinary message to the destination processes).

This seems to me like an important function that isn't present in MPI.  Is there
some reason why it wasn't included?  Are there any workarounds?

Cheers,
David Sitsky
Australian National University.




