Hi Jim,

Another thought that occurred to me was getting rid of get_info - in favour of using the MPI_T interface to figure out what the MPI library has actually done. There is a well-defined semantic for MPI_T variables w.r.t. when can the value change, what variables exist, what do particular values of particular variables mean, etc.

We could propose to standardise some MPI_T variables, such as MPI_ANY_TAG_is_supported or MPI_ANY_TAG_is_permitted, to replace the get_info feedback mechanism.

Cheers,
Dan.

On 17 Oct 2016, at 16:49, Jim Dinan <james.dinan@gmail.com> wrote:

Ahh crud, sorry I missed this meeting.  I've got the next one on my calendar and should be there.

I was recently telling a colleague about this get_info stuff and he made a suggestion that initially sounded nuts, but is becoming more appealing.  Can we just remove set_info?  Many of the challenges we have are caused by this function, and yet its value is questionable.  What do you think?  Worth looking into?

 ~Jim.

On Fri, Oct 7, 2016 at 4:24 AM, Dan Holmes <dholmes@epcc.ed.ac.uk> wrote:
Hi all,

The next point-to-point teleconference will be on Monday 10th October 2016 at 11am Central US via webex.

Connection details are on the p2p wiki:
https://github.com/mpiwg-p2p/p2p-issues/wiki

Agenda:

1) Deprecation of send cancel (issue #26): at the Sept 2016 meeting, Rolf correctly pointed out multiple additional administrative changes that are needed for this. I have a list of these changes and I am working through them. Once that is done, we will need 4 reviewers (Rolf + 3).

2) Communicator assertions (issue #11): at the Sept 2016 meeting, Pavan et al raised the valid objection that fixing the train-wreck of INFO_GET/SET may have knock-on effects on this and so we should delay the vote on it until we have the full picture.

3) Fixing INFO_GET/SET (issue #new): Jim and I have discussed a plan to get this moving.

Slides from Sept 2016 meeting with notes on discussion and way forward:

In addition to our next meeting (10th Oct), we will have 24th Oct and 7th Nov before 21st Nov, which is the 2-week deadline for the next face-to-face meeting.

Plan:
Withdraw #11 from vote on 23rd Sept. [DONE]
Create #new and produce a first draft of the text for the Standard. [IN PROGRESS]
Discuss at the meeting on 10th Oct.
Adjust text for the Standard and produce a first draft of the slides for presenting at the next face-to-face.
Discuss at the meeting on 24th Oct.
Find and fix all the references and stuff that's needed for a proper proposal (Rolf-proofing).
Discuss at the meeting on 7th Nov.
Final adjustments for #new, text, and slides.
Announce #11 (1st vote) and #new (reading) to mpi-forum list by 21st Nov.

If we discover during the work on #new that #11 needs to be adjusted then we can re-start its formal procedure with a reading in December.



Cheers,
Dan.

-- 
Dan Holmes
Applications Consultant in HPC Research
EPCC, The University of Edinburgh
James Clerk Maxwell Building
The Kings Buildings
Peter Guthrie Tait Road
Edinburgh
EH9 3FD
T: +44(0)131 651 3465
E: dholmes@epcc.ed.ac.uk

*Please consider the environment before printing this email.*

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

_______________________________________________
mpiwg-p2p mailing list
mpiwg-p2p@lists.mpi-forum.org
https://lists.mpi-forum.org/mailman/listinfo/mpiwg-p2p

_______________________________________________
mpiwg-p2p mailing list
mpiwg-p2p@lists.mpi-forum.org
https://lists.mpi-forum.org/mailman/listinfo/mpiwg-p2p