Re: [mpiwg-tools] Telecon this week?
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi all, I hope I'll be able to join as well (though I'm technically on vacation). I think the proposed changes for ticket 383 doesn't reflect what we had in mind (or at least what I had in mind ;-)), and we should discuss this very carefully - ideally with Martin as this affects all tools which will support MPI_T. My idea was, and maybe that wasn't clear from the discussions we had, that on heterogeneous systems variables with the same name have the same semantics, not only the processes in the same MPI_COMM_WORLD. An example is the DEEP system (www.deep-project.eu), where we'll have a cluster of Xeons and an attached cluster of Xeon Phis (a.k.a. booster) , both running MPI and processes on the cluster should offload kernels to the booster via MPI_COMM_SPAWN. Thus, the processes on the cluster and those on the booster are connected. Now I would like to ensure that variables with the same name are required to have the same semantics across all processes, be it cluster or booster, e.g. the eager limit or total network packages sent (although there should be counters for the different network interfaces). With the proposed changes to the ticket, this is no longer guaranteed. I think classifying the events in the library by MPI library versions prevents that entirely, because in the case above, events with the same name and semantics exist on the Xeons and the Xeon Phis, which are classified differently. If we go with the proposed addition of checking the MPI library version string, my concern is that MPI implementations would opt for the easy way and simply provide different version strings for the libraries, which in the end would result in the same mess. My understanding is that MPI versions from different vendors can't easily connect as there is no MPI ABI, so that the issue with two independent libraries shouldn't be a major problem. - -Michael On 09/30/2014 11:01 AM, Schulz, Martin wrote:
Yes, the notes are on the Wiki, but feel free to add if I forgot something.
Sorry, I can’t make it on Thursday,
Martin
On 9/29/14, 6:34 PM, "Kathryn Mohror" <[email protected]> wrote:
Hi Marc-Andre,
Great! I’m glad you can join too.
Martin (thank you Martin!) put notes up on the wiki: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Tools/notes-2014-09-
1
5
Kathryn _________________________________________________________________
Kathryn Mohror, [email protected], http://scalability.llnl.gov/
Scalability Team @ Lawrence Livermore National Laboratory, Livermore, CA, USA
On 9/29/14, 9:29 AM, "Marc-Andre Hermanns" <[email protected]> wrote:
Hi Kathryn,
I should be able to make it as well.
@Martin: Do I recall correctly that you made some notes on the comments on the tool tickets? It would be great if you could mail them (to the list or Kathryn) beforehand. I would think along those notes we can reconstruct the discussion ;-)
Cheers, Marc-Andre
On 29.09.14 18:04, Kathryn Mohror wrote:
Yes, the usual time, 11am US Eastern, 8am US Pacific.
Talk to you then! Kathryn _________________________________________________________________
Kathryn Mohror, [email protected], http://scalability.llnl.gov/
Scalability Team @ Lawrence Livermore National Laboratory, Livermore, CA, USA
On 9/29/14, 8:51 AM, "Jeff Squyres (jsquyres)" <[email protected]> wrote:
That's Thursday at 11am US Eastern, right?
If so, I should be able to be there.
On Sep 29, 2014, at 11:05 AM, Mohror, Kathryn <[email protected]> wrote:
Hi all,
I¹m hoping to have our regular telecon this week to get the details on what happened at the F2F in Japan. I know Martin is on travel and won¹t be able to call in. Will anyone who attended the forum meeting be able to join in?
Happy Monday! Kathryn _________________________________________________________________
Kathryn Mohror, [email protected], http://scalability.llnl.gov/
Scalability Team @ Lawrence Livermore National Laboratory, Livermore, CA, USA
_______________________________________________ mpiwg-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-tools
Jeff Squyres [email protected] For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
_______________________________________________ mpiwg-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-tools
mpiwg-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-tools
-- Marc-Andre Hermanns Jülich Aachen Research Alliance, High Performance Computing (JARA-HPC) German Research School for Simulation Sciences GmbH
Schinkelstrasse 2 52062 Aachen Germany
Phone: +49 241 80 99753 Fax: +49 241 80 6 99753 www.grs-sim.de/parallel email: [email protected]
_______________________________________________ mpiwg-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-tools
_______________________________________________ mpiwg-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-tools
- -- Michael Knobloch Institute for Advanced Simulation (IAS) Jülich Supercomputing Centre (JSC) Telefon: +49 2461 61-3546 Telefax: +49 2461 61-6656 E-Mail: [email protected] Internet: http://www.fz-juelich.de/jsc -----BEGIN PGP SIGNATURE----- Version: GnuPG v1 iQIcBAEBAgAGBQJUKqZjAAoJENAucJeJjnBJOhYP/1HjzkjumyR7CFQxWbqG1roq 1qHpoUupQ/AntwvnKykCtd4jY4O1PdF5J5NeEsSbAhvOHn6ivg7n5YTLkGVue49z 7+Tx/BbhkbVPVElQ91LuoqN+IAtgeQRLxDjhxOYAi/vbC80AJ4w0ZGprack6NefF Ix4hP575WEgjYR7c5x9YiDfsgj22vqimE4tMABFNYUrIGNil++Dx0NmbRfrIQQnb r6hx1EpLlfE+jpe14536bKbUGF1s4QiB5oH9lNRT0/UrlzFvekGJN/0zbHiJs22K 6sRGgYT7zkvi5d4cdjOiMo8Xxi8sUrYtnZkytXkubVWrM+e7sGtc8JdccJiUNYgC agglll+RDxC7yR+H2JXx7GnklDI1QYXmqEEwZbgK7Wejh0/n3FNFjgdFELUJKM1I LRJCXHlh9R+Mitzv+N1hJ/H5ZIYWutEc3CYlmI5Tj/cwk7bArIEvxrvOOnPeeiFO Jc8jdIJWr5vJPEMGeCe8En/AvH227dUabUmmBajFnWsU5pdS3qs69ozEs7yfjT9F EPcvjqAOajRkKes7RL6NYLlULtA8Hv1rQcrTdtk32mrv+xstjWtm7UbwPujb1EV/ MZpnHa2nfJTqaw5NVL9mH9h/9XVC01W8TZeq84tSRp8KUuoeJXnqci6a6moc/3S4 SShTdUxC/HhbNVivbYmX =VOx0 -----END PGP SIGNATURE----- ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------ Forschungszentrum Juelich GmbH 52425 Juelich Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. Schmidt ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------
participants (1)
-
Michael Knobloch