Newsgroups: comp.parallel.mpi,comp.parallel
From: spimchon@cs.anu.edu.au (M. Spinks)
Subject: MPI_IPROBE
Organization: Australian National University
Date: Wed, 31 Aug 1994 04:42:25 GMT
Message-ID: <3411nhINNd8g@dubhe.anu.edu.au>

In the MPI Standard (p. 49, 4-10) the following is stated about MPI_IPROBE:

'MPI_IPROBE...returns flag=true if there is a message that can be receive and that
matches the pattern specified by the arguments source, tag and comm.   The call
matches the same message that would have been received by a call to MPI_RECV...
executed at the same point in the program...'

Does anyone know if this means that MPI_IPROBE matches *only* messages that would be
received by MPI_RECV?    The above makes no mention of the applicability of 
MPI_IPROBE to MPI_RECV; is this because it is illustrative, or definitive?   Can MPI_IPROBE be applied to MPI_IRECV?   I can see that it might not, owing to implementation and/or safety issues.   On the other hand, though, I can see that it
might, since probing for a message fits in with an asynchronous paradigm.   The 
standard also mentions (p. 35, 28) that nonblocking sends can be matched with 
blocking receives, and vice-versa, which would seem to indicate that MPI_IPROBE can
indeed be applied to MPI_IRECV...

Once again, please respond via e-mail.

Thanks,
       - Matthew.

--
Matthew Spinks                 | 'But all in all, what a man does,
Computer Science Honours       |  is what he must, to be true to himself,
Department of Computer Science |  if he be worthy of the name of man.'
ANU ACT 0200 Australia         | 
spimchon@mehta.anu.edu.au      | - Laurence Chisholm Young.





