Hi Marc-Andre I’m afraid I have a conflicting meeting at the time of this one, and so I cannot attend. However, I would like to bring forward something that Martin proposed at SC a few weeks ago. We were discussing PMIx and the capabilities it offers, and Martin suggested perhaps exposing those to users via the MPI_T definitions (as opposed to users directly calling PMIx) as a way of offering a little more abstraction. Might be worth some initial discussion, if time permits. I believe you will have some folks there who are at least somewhat familiar with PMIx. Ralph

On Dec 6, 2016, at 6:26 AM, Marc-Andre Hermanns <[email protected]> wrote:

Hi all,

For everyone wanting to join via WebEx, here is the agenda page with the link:

http://mpi-forum.org/meetings/2016/12/agenda

On the second day of the forum face-to-face meeting, we are going to focus on the performance tools.

The fall has been rough on our progress here, as we only had a single call since Edinburgh to recap on MPI_T events.

I hope Kathryn can lead the call on site (as it's always a little tricky to track the conversation being the only one on Webex ;-))

From the MPI_T events side of things, I'd like to discuss:

- In Edinburgh, Martin actually wanted to allow to read events as a whole. How do we communicate the padding to the user?

- How are we going to allow tools to attach and detach additional event buffers to the system?

- Regarding timestamps, what structure should MPI_T_time have. Single 64-bit value, with implementation dependent internals? Should we fix the internals?

Finally, as I assume Jeff and Martin to be there, it would also be good, to revisit PMPI2 as well.

Cheers, Marc-Andre

-- -- Marc-Andre Hermanns Jülich Aachen Research Alliance, High Performance Computing (JARA-HPC) Jülich Supercomputing Centre (JSC)

Wilhelm-Johnen-Str. 52425 Jülich Germany

Phone: +49 2461 61 2509 | +49 241 80 24381 Fax: +49 2461 80 6 99753 www.jara.org/jara-hpc email: [email protected]

_______________________________________________ mpiwg-tools mailing list [email protected] https://lists.mpi-forum.org/mailman/listinfo/mpiwg-tools

_______________________________________________ mpiwg-tools mailing list [email protected] https://lists.mpi-forum.org/mailman/listinfo/mpiwg-tools