[Mpi3-bwcompat] Phone conf schedule?

[Quincey Koziol]

Hi Fab,

On May 17, 2011, at 1:16 PM, Fab Tillier wrote:

> Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15
> 
>> On May 17, 2011, at 1:01 PM, Fab Tillier wrote:
>> 
>>> The meeting request I had expired 4/15/2011.  Do we want to resume it?
>>> 
>>> Couple topics of interest, actually on topic for our working group:
>>> - Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI
>>> 3.0.  They are still available via MPI 2.2 support, and an implementation can
>>> easily provide concurrent support for MPI 2.2 and MPI 3.0, should that be
>>> required.
>>> - Defining the C++ bindings as optional in MPI 3.0.  This allows
>>> implementations that do not ship C++ bindings to be standard compliant.
>>> - Removing C++ bindings from MPI 3.0.  This one goes counter to the
>>> previous one, but follows the same logic as for the other deprecated
>>> functions.  If we don't provide C++ bindings for new MPI 3.0
>>> functionality, we should remove them from MPI 3.0.  They can still be
>>> supported via MPI 2.2 compliance.
>> 
>> 	Anything to discuss about const buffers?
> 
> We passed the first reading, so we're going for 1st vote at the next meeting.  There was one minor ticket0 wording change for the IN description, but nothing major.  I don't think there's much to discuss at this point.
> 
>> 	I've got someone working on extensions to the datatype
>> construction routines, creating a use case/justification for why we should add
>> those routines to 3.0.  Adam Moody was going to create a ticket for those,
>> once I give him this use case document.  [Dunno if we need to talk about
>> anything yet on this though]
> 
> That would be nice to see.  Jeff and I are also working on the MPI Timer request stuff, don't know if we want to discuss that in this working group, I'd be up for it, I like this crowd...

        Awww, we like you too, Fab. :-) :-)

        I've got two items I'd like to bounce around with folks:

- How best to return an exit code from an MPI application?

- What about adding a "quiet" flag to mpiexec to request that mpiexec (not the application) doesn't make any output?

        Quincey