[Mpi-forum] Wording in MPI standard
pakin at lanl.gov
Mon Nov 26 13:16:13 CST 2012
On 11/26/2012 11:44 AM, Bronis R. de Supinski wrote:
> The problem with your suggestion is that completion
> has a well defined meaning in MPI that requires the
> completion call. If the receive completed without
> any call then a subsequent attempt to complete it
> would be an error. I believe that is reason the term
> "satisfied" was used.
The current text says that "the receive *should* complete even if no
call is executed by the sender to complete the send." If a subsequent
attempt to complete such a send would be an error, then the standard
would have to say, "the receive *must not* complete even if no call is
executed by the sender to complete the send," which is in disagreement
with what a typical (all?) MPI implementation does.
I'm merely taking issue with the use of "should," which seems to let
an MPI implementation decide whether to complete the receive but not
allow applications to rely on this happening. "Must" or "is required
to" would let applications reason better about progress.
More information about the mpi-forum