[Mpi3-subsetting] MPI subsetting: charting the way forward atatelecon next week?

[Richard Treumann]

I am not completely happy with the name format I have suggested for
assertions. I would be glad to hear suggestions.  Here is how I would cast
each assertion.

MPI_NO_EAGER_THROTTLE:
    This program does not depend on libmpi to provide an EAGER_THROTTLE
semantic

MPI_NO_REDUCTION_DETERMINISM
    This program does not depend on libmpi to provide a
REDUCTION_DETERMINISM semantic

MPI_NO_DATATYPE_HETEROGENEITY
    This program does not depend on libmpi to provide support for
DATATYPE_HETEROGENEITY in communication

MPI_NO_SEND_CANCEL
     This program does not depend on libmpi to provide a SEND_CANCEL
semantic

MPI_NO_ANY_SOURCE
     This program does not depend on libmpi to provide an ANY_SOURCE
semantic

MPI_NO_MPI_IO
      This program does not depend on libmpi to provide MPI_IO support

etc

In each case we are talking about something the MPI standard says must be
provided by a compilant MPI implementation, just in case the application
needs it. For each case there are application sthat do not need it.

Does this help?

Dick Treumann  -  MPI Team/TCEM
IBM Systems & Technology Group
Dept 0lva / MS P963 -- 2455 South Road -- Poughkeepsie, NY 12601
Tele (845) 433-7846         Fax (845) 433-8363