February 2014 - mpiwg-hybridpm - mailman3.cels.anl.gov

Reminder: Telecon tomorrow
by Balaji, Pavan 13 Feb '14

13 Feb '14

All, Reminder that we’ll have the special hybrid telecon tomorrow at 11am central as decided on the last call. The plan is to go through the tickets text once more before the Forum reading. I’ve made changes to ticket-357 based on Dan’s comments and uploaded a new version. There were no comments on ticket-411. Thanks, — Pavan

1 0

Re: [mpiwg-hybridpm] Input needed for the Agenda for MPI Forum Meeting in March
by Balaji, Pavan 12 Feb '14

12 Feb '14

Hi Martin, For the hybrid/communicators WG, we want to do readings for #357, #411 and #381. Some of this might need discussion time too, so I think a 2-hour slot will be good (plenary + reading). We don’t need WG time. Cc’ed the WG in case someone feels otherwise. Thanks, — Pavan On 2/12/14, 12:30 AM, "Schulz, Martin" <schulzm(a)llnl.gov> wrote: >Dear MPI Forum WG leads, > >I am in the process of putting together the agenda for the next forum >meeting in the next few weeks and will be putting up a first draft in the >next few days. To help me with this, please send me the following >information for your WG(s): > >- Any tickets for readings or votes that the WG wants to bring up? >- For readings: how much time would you like reserved for this? >- Any additional plenary times for WG topics? >- How much WG time would you like? >- For the WG time: any major conflicts that I should be aware of? > >Thanks! > >Martin > > >________________________________________________________________________ >Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm >CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > >

1 0

Re: [mpiwg-hybridpm] Reminder: Hybrid WG telecon tomorrow
by Balaji, Pavan 10 Feb '14

10 Feb '14

All, Here are the meeting minutes for today¹s call: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2014-02- 10 We decided to have another telecon this Friday at 11am central to wrap up the text for the Forum reading. Here are the tickets being considered: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/357 https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/381 ‹ Pavan On 2/9/14, 10:51 PM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: >All, > >This is a reminder for the Hybrid WG telecon tomorrow at 11am central >time. Webex information is here: > >https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid > > >The agenda is to go over the drafts for the formal readings of various >tickets, as discussed on the last telecon. > >I¹ve uploaded the drafts for tickets 357 and 411 as promised (though I¹m a >bit late on that, sorry): > >https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/357 >https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 > >Jim: what¹s the status of the draft for tickets 380 and 381? We are only >a week away from the final deadline for sending out the drafts to the >Forum. > >Thanks, > > ‹ Pavan >

1 0

tickets 411 and 357
by Balaji, Pavan 10 Feb '14

10 Feb '14

Folks, I¹ve updated the ticket 411 ticket description and the writeup as per the discussion today. Please take a look at it and let me know if I missed something: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 The changes we made during the call for ticket 357 have been uploaded as well: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/357 Dan: can you please make a pass on the chapter to make sure I didn¹t miss anything? Thanks, ‹ Pavan

1 0

381 ticket has WG summary - please read and modify if necessary
by Jeff Hammond 10 Feb '14

10 Feb '14

Please make sure the ticket comment captures today's discussion. Jeff ---------- Forwarded message ---------- From: MPI Forum <mpi-forum(a)lists.mpi-forum.org> Date: Mon, Feb 10, 2014 at 12:33 PM Subject: Re: [MPI Forum] #381: Communicator info keys To: #381: Communicator info keys -------------------------------------+------------------------------------- Reporter: jdinan | Owner: Type: | jdinan Enhancements to standard | Status: Priority: Not | assigned ready / author rework | Milestone: Version: MPI | 2013/03/11 Chicago, USA <next> | Resolution: Keywords: hybrid | Implementation status: Author: Bill Gropp: 0 | Unnecessary Author: Adam Moody: 0 | Author: Rich Graham: 0 Author: Dick Treumann: 0 | Author: Torsten Hoefler: 0 Author: George Bosilca: 0 | Author: Jesper Larsson Traeff: 0 Author: Bronis de Supinski: 0 | Author: David Solt: 0 Author: Jeff Squyres: 0 | Author: Rajeev Thakur: 0 Author: Rolf Rabenseifner: 0 | Author: Alexander Supalov: 0 -------------------------------------+------------------------------------- Comment (by jhammond): The WG discussion today led to the following conclusions: - It will be hard to implement this in a Forum-convincing way right now, and certainly for the MPI-3.1 deadline. - Without experience and existing implementations, it is hard to know if we are doing this exactly right. - A good way to move things forward is to write a paper about this topic and tries to convince pre-standard implementations. (WG members should feel free to revise this as appropriate.) -- Ticket URL: <https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/381#comment:31> MPI Forum <https://svn.mpi-forum.org/> MPI Forum -- Jeff Hammond jeff.science(a)gmail.com

1 0

Re: [mpiwg-hybridpm] Reminder: Hybrid WG telecon tomorrow
by Jeff Hammond 10 Feb '14

10 Feb '14

> For discussion today: How to deal with MPI_Comm_compare(ep_comm[0], > ep_comm[1], &result). Where ep_comm is an array of handles to different > ranks in the same endpoints comm. Currently, MPI defines: IDENT, CONGRUENT, > SIMILAR, and UNEQUAL. None are a good match, seems like we need a new one > for the case where they are the same communicator and same group (IDENT), > but represent a different rank in the comm. I suggest MPI_ALIASED to describe this situation. Jeff -- Jeff Hammond jeff.science(a)gmail.com

1 0

Re: [mpiwg-hybridpm] Reminder: Hybrid WG telecon tomorrow
by Jim Dinan 10 Feb '14

10 Feb '14

Please see attached and look for instances of "ticket380". For discussion today: How to deal with MPI_Comm_compare(ep_comm[0], ep_comm[1], &result). Where ep_comm is an array of handles to different ranks in the same endpoints comm. Currently, MPI defines: IDENT, CONGRUENT, SIMILAR, and UNEQUAL. None are a good match, seems like we need a new one for the case where they are the same communicator and same group (IDENT), but represent a different rank in the comm. ~Jim. On Mon, Feb 10, 2014 at 11:01 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > I'm creating the official PDF draft for #380. Will have it in time for > today's meeting. > > ~Jim. > > > On Sun, Feb 9, 2014 at 11:51 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > >> All, >> >> This is a reminder for the Hybrid WG telecon tomorrow at 11am central >> time. Webex information is here: >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid >> >> >> The agenda is to go over the drafts for the formal readings of various >> tickets, as discussed on the last telecon. >> >> I've uploaded the drafts for tickets 357 and 411 as promised (though I'm a >> bit late on that, sorry): >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/357 >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 >> >> Jim: what's the status of the draft for tickets 380 and 381? We are only >> a week away from the final deadline for sending out the drafts to the >> Forum. >> >> Thanks, >> >> -- Pavan >> >> >> _______________________________________________ >> mpiwg-hybridpm mailing list >> mpiwg-hybridpm(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm >> > >

1 0

Re: [mpiwg-hybridpm] Reminder: Hybrid WG telecon tomorrow
by Jim Dinan 10 Feb '14

10 Feb '14

I'm creating the official PDF draft for #380. Will have it in time for today's meeting. ~Jim. On Sun, Feb 9, 2014 at 11:51 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > All, > > This is a reminder for the Hybrid WG telecon tomorrow at 11am central > time. Webex information is here: > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid > > > The agenda is to go over the drafts for the formal readings of various > tickets, as discussed on the last telecon. > > I've uploaded the drafts for tickets 357 and 411 as promised (though I'm a > bit late on that, sorry): > > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/357 > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 > > Jim: what's the status of the draft for tickets 380 and 381? We are only > a week away from the final deadline for sending out the drafts to the > Forum. > > Thanks, > > -- Pavan > > > _______________________________________________ > mpiwg-hybridpm mailing list > mpiwg-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm >

1 0

Re: [mpiwg-hybridpm] Reminder: Hybrid WG telecon tomorrow
by Balaji, Pavan 10 Feb '14

10 Feb '14

We already talked about this at the Forum and decided that the chapter committee can fix it. We can do it for the errata. — Pavan On 2/10/14, 8:38 AM, "Jeff Hammond" <jeff.science(a)gmail.com> wrote: >Can we approve the first (and easier) half of >https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/352 as an erratum? > It's a word correction, but an important one in my opinion. > >Jeff > >On Sun, Feb 9, 2014 at 10:51 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: >> All, >> >> This is a reminder for the Hybrid WG telecon tomorrow at 11am central >> time. Webex information is here: >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid >> >> >> The agenda is to go over the drafts for the formal readings of various >> tickets, as discussed on the last telecon. >> >> I’ve uploaded the drafts for tickets 357 and 411 as promised (though >>I’m a >> bit late on that, sorry): >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/357 >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 >> >> Jim: what’s the status of the draft for tickets 380 and 381? We are >>only >> a week away from the final deadline for sending out the drafts to the >> Forum. >> >> Thanks, >> >> — Pavan >> >> >> _______________________________________________ >> mpiwg-hybridpm mailing list >> mpiwg-hybridpm(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm > > > >-- >Jeff Hammond >jeff.science(a)gmail.com >_______________________________________________ >mpiwg-hybridpm mailing list >mpiwg-hybridpm(a)lists.mpi-forum.org >http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm

1 0

Fwd: [MPI Forum] #352: replace require with request in MPI_INIT_THREAD text
by Jeff Hammond 10 Feb '14

10 Feb '14

What do I need to do to get this voted on for 3.1? I can't remember if the chapter committee alone can approve this. Jeff ---------- Forwarded message ---------- From: MPI Forum <mpi-forum(a)lists.mpi-forum.org> Date: Mon, Feb 10, 2014 at 9:27 AM Subject: Re: [MPI Forum] #352: replace require with request in MPI_INIT_THREAD text To: #352: replace require with request in MPI_INIT_THREAD text -------------------------------------+------------------------------------- Reporter: | Owner: jhammond | balaji Type: Text | Status: new (only) changes | Milestone: Priority: Forum | 2012/12/03 California, USA feedback requested | Resolution: Version: MPI | Implementation status: 3.0 | Unnecessary Keywords: | Author: Rich Graham: 0 Author: Bill Gropp: 0 | Author: Torsten Hoefler: 1 Author: Adam Moody: 0 | Author: Jesper Larsson Traeff: 0 Author: Dick Treumann: 0 | Author: David Solt: 0 Author: George Bosilca: 0 | Author: Rajeev Thakur: 0 Author: Bronis de Supinski: 0 | Author: Alexander Supalov: 0 Author: Jeff Squyres: 0 | Author: Rolf Rabenseifner: 0 | -------------------------------------+------------------------------------- Description changed by jhammond: Old description: > In MPI-3 12.4.3, > {{{ > "Different processes in MPI_COMM_WORLD may require different levels > of thread support." > }}} > should be changed to > {{{ > "Different processes in MPI_COMM_WORLD may request different levels > of thread support." > }}} > > The first statement has no well-defined meaning as I read it. First, the > function itself takes a "requested" argument and thus "request" is the > appropriate word by symmetry. Second, it is no different than text that > says "Different processes may require different amounts of memory," which > has no connection to what an implementation is required to do. > > I would also recommend that this text be supplemented as follows so that > it is unambiguous to users what is being said: > > {{{ > Different processes in MPI_COMM_WORLD may request different levels of > thread > support. For example, an application might request MPI_THREAD_MULTIPLE > on > one rank and MPI_THREAD_SINGLE on another. A high-quality implementation > will provide only the requested level on each process, but it may also > provide the > maximum requested level on all processes. > }}} > > If the additional text is considered controversial, then these two > suggestions should be separated. I think the first one is trivial to > implement but important for clarity. New description: In MPI-3 12.4.3, {{{ "Different processes in MPI_COMM_WORLD may require different levels of thread support." }}} should be changed to {{{ "Different processes in MPI_COMM_WORLD may request different levels of thread support." }}} The first statement has no well-defined meaning as I read it. First, the function itself takes a "requested" argument and thus "request" is the appropriate word by symmetry. Second, it is no different than text that says "Different processes may require different amounts of memory," which has no connection to what an implementation is required to do. -- -- Ticket URL: <https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/352#comment:4> MPI Forum <https://svn.mpi-forum.org/> MPI Forum -- Jeff Hammond jeff.science(a)gmail.com

1 0