[Mpi-forum] Clarification on MPI communicator semantics

Scott Pakin pakin at lanl.gov
Fri Oct 11 20:17:53 CDT 2013

On 10/11/2013 07:00 PM, Pavan Balaji wrote:
> It is valid to do such a comparison.  This is mentioned on page 12, lines 38-39.
> "MPI procedures that operate on opaque objects are passed handle arguments to access these objects. In addition to their use by MPI calls for object access, handles can participate in assignments and comparisons."

Ah, I missed that (obviously).  However, I should point out that that
text didn't have any of the keywords I was searching for -- except
"communicator", which appears too often in the document to be a
meaningful search term.  Do you therefore think it might make sense
for the MPI_Comm_compare() description to refer back to that section?
Perhaps it could distinguish an "==" that returns true from an
MPI_Comm_compare() that returns MPI_IDENT.  (I'm sure there's some way
in MPI to produce two different handles to the same communicator
object, but I can't think of what that might be.  MPI_Comm_f2c()

Thanks for the quick reponse!

-- Scott

More information about the mpi-forum mailing list