[MPI3 Fortran] Agenda for MPI3 Fortran Working group next week
longb at cray.com
Thu Jun 4 12:22:28 CDT 2009
Jeff Squyres wrote:
> On Jun 4, 2009, at 12:59 PM, N.M. Maclaren wrote:
>> >One of the decisions the MPI group needs to make is whether using an
>> >module is required for the new interface. If the answer is no, then
>> >the space of options is reduced quite a bit. I don't get a vote,
>> but if
>> >I did, I would favor a requirement that any program unit that uses MPI
>> >facilities has access to the MPI module.
>> I don't, either, and I quite agree.
> I'd guess that we're fine with a module (can't speak for everyone, of
> course, but the temperature had been fairly favorable to a module in
> prior meetings).
> We already have "use mpi"; I think that it's an open question as to
> whether this will be "use mpi3" (or some other name) for the new
> interface. I'm thinking that it'll have to be something new (e.g.,
> "mpi3") because we want to change some of the existing interfaces.
> I think this is something we need to talk about in Menlo Park next week:
> - what to do with mpif.h (I think we should deprecate it -- MPI
> implementations can choose to keep supporting it for legacy
> apps/customers, of course)
Or, perhaps, specify an mpif.h for MPI 3 that is empty except for some
Comment lines that remind the reader about accessing the MPI3 module.
> - what to do with "use mpi" (deprecate?)
> - what to do with this new interface ("use mpi3"?)
Bill Long longb at cray.com
Fortran Technical Support & voice: 651-605-9024
Bioinformatics Software Development fax: 651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120
More information about the mpiwg-fortran