From: "Anthony Skjellum" <skjellum@netdoor.com>
Newsgroups: comp.parallel.mpi
References: <36879DE9.3A0C1CE7@LL.MIT.EDU>
Subject: Re: use of MPI_WAIT in non-blocking mode
Message-Id: <AeNh2.294$fK.1825@axe.netdoor.com>
Date: Mon, 28 Dec 1998 09:12:04 -0600
Organization: Internet Doorway, Inc. -- http://www.netdoor.com/


A non-blocking receive, followed by a wait has the same
semantic behavior as a blocking receive.  Your example
is probably faulty, because a blocking receive should not
have a request, just a status in it (but that is aside your
question).

Because implementations may treat IRECV+WAIT differently
from blocking RECV, performance will vary, but you are
correct that they accomplish the same effect.

-Tony

Michael Starkie wrote in message <36879DE9.3A0C1CE7@LL.MIT.EDU>...
>I am having trouble understanding the difference between the following 2
>examples.
>
>example 1:
>    ...
>    MPI_IRECV(inval, 1, MPI_REAL, 1, 0, req, ierr)
>    MPI_WAIT(req, status, ierr)
>     ...
>
>example 2:
>    ...
>    MPI_RECV(inval, 1, MPI_REAL, 1, 0, req, ierr)
>    ...
>
>Example 1 is used in the Interface Standard Document.  In example 1, a
>non-blocking receive is followed by a wait. Isn't this the same behavior
>as a blocking receive?
>
>--
>Cordially,
>Michael Starkie
>
>
>

