Open mpiforumbot opened 8 years ago
Originally by jdinan on 2015-10-16 14:33:31 -0500
Sept. 2015 meeting -- name suggestion of mpi_assert_exact_length. Comment was that no_underflow suggests that overflow could be allowed (even though MPI forbids it).
Originally by jdinan on 2015-10-16 14:44:30 -0500
Sept. 2015 meeting -- Comment: should this be an info key or a new receive operation?
Originally by jdinan on 2015-10-16 14:45:25 -0500
We may want to restrict this to point-to-point communication to avoid impacting collectives.
Originally by jdinan on 2015-09-24 03:29:27 -0500
Motivation
This info key allows the programmer to assert that the length of messages received on the given communicator will be equal to the length of their corresponding receive buffers.
Without this assertion, a receiver does not know how much data will be sent. For large messages, the receiver is not able to determine ahead of time whether the sender will use an eager or rendezvous protocol.
Proposed Info Keys
The following text would be added to page 249, line 12 of MPI 3.0:
Related Tickets
381 -- Communicator info keys
461 -- No wildcards