September 2014 - mpi-forum - mailman3.cels.anl.gov

Slides from Kobe Meeting
by Jim Dinan 22 Sep '14

22 Sep '14

Hi All, Could folks who presented slides at the Kobe meeting please send them to Jeff Squyres so we can post them on the meeting website? Thank you to our gracious hosts for a very enjoyable and productive meeting! Best, ~Jim.

1 0

MPI Forum Agenda for Kobe Meeting
by Martin Schulz 12 Sep '14

12 Sep '14

Hi all, A few logistics issues for the Kobe meeting: as you know, we won¹t meet the quorum this time, so we won¹t be able to any errata or ticket votes. However, there are still many issues that we can discuss (we actually have a fairly good attendance in total) and so I am sure we can make this a productive meeting. Without the need of having a rigid voting time, though, and without the need for parallel tracks, we can make the schedule a bit more flexible. We¹ll start at 10am with the tools WG. For those who want to skip the tools WG, we¹ll assemble at noon for lunch (meeting point is at RIKEN as well) and then we¹ll go to a nearby restaurant. After that, my suggestion is just to work through the agenda (incl. discussion of some of the more controversial errata items) as it is listed, but not worry about timing. I got a few more requests/suggestions for plenary discussions and I will add those to the agenda later today, but aside from that I am not planning to modify the website much more we¹ll just go in order and we may finish early or not, depending on how things go. The only fixed item is still the K computer tour on Tuesday for those who are interested. See you on Monday in Kobe, Martin

1 0

Re: [Mpi-forum] September MPI Forum / Agenda and Reminder to Register
by Jeff Squyres (jsquyres) 07 Sep '14

07 Sep '14

Martin and I are working through the quorum issue, so let's leave that out of the discussion for the moment. *Assuming* the meeting meets quorum, you can have a separate formal vote at the meeting to get your ticket 0 changes included as part of the #380 text. See http://meetings.mpi-forum.org/procedures-2013-03-15.pdf, section 1.3.2, bullet 4. Basically: you present the changes, the Forum votes, and assuming all quorums are met, if there are zero "no" votes, your changes can be included in #380. On Sep 4, 2014, at 10:16 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > To state this a little more clearly, we should be able to vote on the version of #380 that was read at the last meeting. We have ticket 0 changes to add to this version of the proposal. Procedurally speaking, we could make this work by reading the ticket 0 changes this time as an FYI and reading them again formally at the next meeting in concert with a second vote. This will be confusing, so I move to just forgive my tardiness. :) > > ~Jim. > > > On Thu, Sep 4, 2014 at 8:56 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Jeff, > > The version of the endpoints proposal that was read at the last meeting is proceeding for a vote, with the addition of ticket 0 changes that the working group made in response to feedback at the last reading. > > My apologies for being a day late on the deadline. The timing was a little tricky this time around with the deadline landing on a Monday and a US holiday. > > ~Jim. > > > On Tue, Sep 2, 2014 at 6:23 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > Jim -- > > I'm sorry, but your mail missed the deadline (I also see you just added a new PDF to the ticket about 15 minutes ago). > > I think you have the following options: > > - proceed with the last version of the ticket that was announced to the Forum (as of last meeting?) > - submit mpi-report.8.pdf for the December meeting > > > > On Sep 2, 2014, at 5:27 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > > > Hi Martin, > > > > Apologies for missing the 9/1 deadline. Endpoints #380 should be up for a first vote in Japan. The working group has made several changes to the ticket based on feedback from the prior reading. We feel that these changes are ticket 0, and would like to read them prior to a vote. If the forum feels the changes are not ticket 0, we would like to re-read at the Japan meeting. > > > > We do have an unmet implementation requirement, which we will satisfy prior to the second vote. > > > > Cheers, > > ~Jim. > > > > > > On Tue, Sep 2, 2014 at 11:44 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Hi Martin, > > > > This is not a vote, just a formal reading. We were asked to make some changes and read it again at the last meeting. > > > > — Pavan > > > > On Sep 2, 2014, at 10:30 AM, Schulz Martin <schulzm(a)llnl.gov> wrote: > > > > > Hi Pavan, > > > > > > I think we can still add this, despite the two week deadline, since we did > > > talk about this in the last meeting. What I don’t remember, though, is the > > > state of the ticket: it’s listed as scheduled for vote and also listed as > > > a reading in the June notes, but the ticket has two options on it and it’s > > > not clear which one is the intended one. What do you want to do with it? > > > > > > Thanks! > > > > > > Martin > > > > > > > > > ________________________________________________________________________ > > > Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > > > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > > > > > > > > > > > > > > > > On 9/2/14, 6:12 AM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > > > > > >> Hi Martin, > > >> > > >> You missed 411: > > >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 > > >> > > >> We had decided to postpone it to this meeting last time. > > >> > > >> Can you add it too? > > >> > > >> Thanks, > > >> > > >> — Pavan > > >> > > >> On Sep 2, 2014, at 3:59 AM, Schulz Martin <schulzm(a)llnl.gov> wrote: > > >> > > >>> Hi Rolf, all, > > >>> > > >>> I have updated the agenda with all tickets to be read and voted on - I > > >>> only got Rolf’s request below, please let me know asap in case I missed > > >>> something (plus I moved the first votes from last time up to the second > > >>> vote stage). With that, the agenda should be complete. > > >>> > > >>> Rolf: all the tickets are on the agenda. There are a few, though, as you > > >>> also noted, that are ticket 0 changes - I grouped them separately with > > >>> the > > >>> intent that we don’t vote on them (but if we decide otherwise, they > > >>> count > > >>> as announced). I know we are a bit sparse in the chapter committees > > >>> right > > >>> now, but we need to fix this anyway to make MPI 3.1 happen. I am > > >>> planning > > >>> on pinging people separately about this in the next days - if they are > > >>> any > > >>> volunteers right away, please let me know. > > >>> > > >>> Thanks! > > >>> > > >>> Martin > > >>> > > >>> ________________________________________________________________________ > > >>> Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > > >>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> On 8/28/14, 8:30 AM, "Rolf Rabenseifner" <rabenseifner(a)hlrs.de> wrote: > > >>> > > >>>> Dear Martin, > > >>>> > > >>>> please add the following errata tickets to the agenda : > > >>>> > > >>>> #456 MPI 3.0 Errata: MPI RMA synchronization for MPI shared memory > > >>>> > > >>>> This is a large ticket that needs discussion > > >>>> first in the RMA-WG, second in the plenary, and > > >>>> hopefully some voting (corrctions and final) > > >>>> > > >>>> ____ > > >>>> Small Fortran corrections are: > > >>>> > > >>>> #433 MPI 3.0 Errata: MPI_WIN_ATTACH must only be for simply contiguous > > >>>> buffers > > >>>> #440 Fortran cross-reference correction (ticket 0 / errata) > > >>>> #441 Missing text for Fortran MPI_SUBARRAYS_SUPPORTED equals .FALSE. > > >>>> (errata) > > >>>> #442 Fortran "vector subscipts" and MPI_BOTTOM (ticket 0 / errata) > > >>>> #443 Fortran MPI_BOTTOM not for initialization or assignment (ticket 0 > > >>>> / > > >>>> errata) > > >>>> #444 Fortran and Persistent Requests (ticket 0 / errata) > > >>>> #445 Fortran and Permament Data Movement: "All" nonblocking (ticket 0 / > > >>>> errata) > > >>>> #446 Fortran cross-reference update (ticket-0/errata) > > >>>> > > >>>> ____ > > >>>> Fortran and absolute address text corrections: > > >>>> #431 MPI 3.0 Errata: MPI Address Errata > > >>>> > > >>>> ____ > > >>>> Errata and ticket-0 additions to tickets #349, #402, #404 > > >>>> (current status of the triple 1st vote): > > >>>> #421 MPI_Aint addressing arithmetic (ticket-0 additions) > > >>>> > > >>>> ____ > > >>>> Small errata corrections to Chap.2: > > >>>> > > >>>> #447 Chap.2 errata: Tools section is an exception: no Fortran interface > > >>>> #448 Chap.2 errata: Semantic term "matching" is missing > > >>>> #449 Chap.2 errata: Missing "in C" in wording about handle-comparison > > >>>> #450 Chap.2 errata: Outdated cross-reference to MPI-1 > > >>>> #451 Chap.2 errata: Correcting description of nonblocking > > >>>> #452 Chap.2 errata: Wording correction about deprecated and removed > > >>>> functions > > >>>> #453 Chap.2 errata: Avoidance of mpi_ / pmpi_ in C due to Fortran > > >>>> reasons > > >>>> > > >>>> #455 Chap.2 errata: Removal of duplicated text on MPI_Aint and > > >>>> MPI_Offset > > >>>> > > >>>> > > >>>> _____________________________________ > > >>>> The rest of this mail are only hints. > > >>>> ____ > > >>>> I also expect that Jeff may want to update, discuss and vote: > > >>>> #428 Add more non-profile-able MPI routines > > >>>> > > >>>> ____ > > >>>> I also expect that Kathryn may want to update, discuss and vote: > > >>>> #383 Consistency of tools information interface variable information > > >>>> #368 MPI_T Minor Errata > > >>>> > > >>>> ____ > > >>>> I also expect that Bill may want to update, discuss and vote: > > >>>> #438 Interpretation of MPI_WIN_BASE for shared memory windows > > >>>> > > >>>> ____ > > >>>> Other MPI-3.0 errata tickets that may or should be handled before > > >>>> MPI-3.1 > > >>>> is finalized: > > >>>> > > >>>> #420 Bug in neighbourhood collective example > > >>>> - due to missing chapter committee action, > > >>>> it shoud be voted in > > >>>> > > >>>> ____ > > >>>> The only MPI-3.0-errata ticket that may need to go MPI-4.0 is: > > >>>> #374 Comm_free/Type_free/Win_free: clarify attribute destructor > > >>>> semantics > > >>>> > > >>>> > > >>>> ____ > > >>>> Tickets #428, #368, #368, #420 may need a milestone-change to "Kobe". > > >>>> Otherwise, they are invisible in the Kobe tickets list: > > >>>> > > >>>> https://svn.mpi-forum.org/trac/mpi-forum-web/query?status=accepted&statu > > >>>> s= > > >>>> > > >>>> assigned&status=new&status=reopened&milestone=2014%2F09%2F15+Kobe%2C+Jap > > >>>> an > > >>>> > > >>>> &col=id&col=summary&col=milestone&col=status&col=type&col=priority&col=v > > >>>> er > > >>>> sion&order=priority > > >>>> > > >>>> Best regards > > >>>> Rolf > > >>>> > > >>>> ----- Original Message ----- > > >>>>> From: "Schulz Martin" <schulzm(a)llnl.gov> > > >>>>> To: "Main mailing list" <mpi-forum(a)lists.mpi-forum.org> > > >>>>> Sent: Thursday, August 28, 2014 9:53:01 AM > > >>>>> Subject: [Mpi-forum] September MPI Forum / Agenda and Reminder to > > >>>>> Register > > >>>>> > > >>>>> Hi all, > > >>>>> > > >>>>> Based on the (sparse) feedback that I have gotten, I have assembled > > >>>>> an > > >>>>> initial agenda and uploaded it to the server. Please check the > > >>>>> agenda, > > >>>>> especially if you sent me a request, and double check that I didn¹t > > >>>>> miss > > >>>>> anything or anybody. If something is missing, please let me know > > >>>>> asap. > > >>>>> > > >>>>> As usual, I tried to split working groups and plenary on separate > > >>>>> days, > > >>>>> with votes on day three. However, so far I have only received one WG > > >>>>> meeting request, which would make Monday relatively sparse. We could > > >>>>> move > > >>>>> parts of Tuesday to that day and probably finish on Tuesday evening. > > >>>>> Some > > >>>>> people have voiced interest in doing this. However, this would move > > >>>>> votes > > >>>>> onto the same day as the second half of the plenary and I would like > > >>>>> some > > >>>>> feedback on this (since the voting times are typically fixed to the > > >>>>> last > > >>>>> day). So the options are: > > >>>>> > > >>>>> Monday WG time (with open, general WG time), Tuesday plenary, > > >>>>> Wednesday > > >>>>> votes > > >>>>> > > >>>>> -or- > > >>>>> > > >>>>> Monday WG and plenary time (errata), Tuesday plenary and votes, > > >>>>> Wednesday > > >>>>> off > > >>>>> > > >>>>> Please let me know if you have a preference one way or the other. If > > >>>>> I > > >>>>> don¹t hear anything or if there are objections to moving the voting > > >>>>> block, > > >>>>> we¹ll leave things as is (first option). > > >>>>> > > >>>>> Thanks! > > >>>>> > > >>>>> Martin > > >>>>> > > >>>>> PS: Please don¹t forget to register, if you haven¹t done so, by the > > >>>>> end of > > >>>>> August (just a few days left). We need an exact attendee list, so > > >>>>> people > > >>>>> can get access to Riken. > > >>>>> > > >>>>> > > >>>>> _______________________________________________________________________ > > >>>>> _ > > >>>>> Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > > >>>>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > >>>>> > > >>>>> > > >>>>> _______________________________________________ > > >>>>> mpi-forum mailing list > > >>>>> mpi-forum(a)lists.mpi-forum.org > > >>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > >>>>> > > >>>> > > >>>> -- > > >>>> Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > > >>>> High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > > >>>> University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > > >>>> Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > > >>>> Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > > >>>> _______________________________________________ > > >>>> mpi-forum mailing list > > >>>> mpi-forum(a)lists.mpi-forum.org > > >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > >>> > > >>> > > >>> _______________________________________________ > > >>> mpi-forum mailing list > > >>> mpi-forum(a)lists.mpi-forum.org > > >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > >> > > >> -- > > >> Pavan Balaji ✉️ > > >> http://www.mcs.anl.gov/~balaji > > >> > > >> _______________________________________________ > > >> mpi-forum mailing list > > >> mpi-forum(a)lists.mpi-forum.org > > >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > > > > > > _______________________________________________ > > > mpi-forum mailing list > > > mpi-forum(a)lists.mpi-forum.org > > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > -- > > Pavan Balaji ✉️ > > http://www.mcs.anl.gov/~balaji > > > > _______________________________________________ > > mpi-forum mailing list > > mpi-forum(a)lists.mpi-forum.org > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > _______________________________________________ > > mpi-forum mailing list > > mpi-forum(a)lists.mpi-forum.org > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Quorum at the Japan Meeting
by Martin Schulz 06 Sep '14

06 Sep '14

Hi all, Bad news: I just got the most recent attendee list for the MPI forum in Japan. I would like Jeff to confirm, but based on my count, we will miss the quorum for this meeting. We currently have 23 eligible organizations (OOE) and, while 18 overall organizations will be attending, only 14 of the 23 will be OOE, leaving us two organizations shy of the quorum, which is 16. 9 of the OOE, i.e., organizations that have been at the last two meetings and therefore are eligible to vote won¹t be in Japan, which causes us to miss the quorum. As a consequence, we cannot have any votes or readings during this meeting, only discussions preparing for readings or errata votes next time. Also, if we wanted to stick with the original MPI 3.1 timeline, only tickets and errata items that are voted in right now would be part of the MPI 3.1, which leaves many of the intended additions out of MPI 3.1. The latter issue can be fixed by moving the MPI 3.1 schedule, as previous discussed for other reasons, and I have not heard any objections to this, so this may not be a big issue. However, we need to figure out how to handle this meeting without a quorum. Tuesday is mostly not affected (aside from the voting block), but I am not sure if talking about the errata tickets on Monday makes much sense. Martin

1 0

Re: [Mpi-forum] September MPI Forum / Agenda and Reminder to Register
by Jim Dinan 04 Sep '14

04 Sep '14

To state this a little more clearly, we should be able to vote on the version of #380 that was read at the last meeting. We have ticket 0 changes to add to this version of the proposal. Procedurally speaking, we could make this work by reading the ticket 0 changes this time as an FYI and reading them again formally at the next meeting in concert with a second vote. This will be confusing, so I move to just forgive my tardiness. :) ~Jim. On Thu, Sep 4, 2014 at 8:56 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Jeff, > > The version of the endpoints proposal that was read at the last meeting is > proceeding for a vote, with the addition of ticket 0 changes that the > working group made in response to feedback at the last reading. > > My apologies for being a day late on the deadline. The timing was a > little tricky this time around with the deadline landing on a Monday and a > US holiday. > > ~Jim. > > > On Tue, Sep 2, 2014 at 6:23 PM, Jeff Squyres (jsquyres) < > jsquyres(a)cisco.com> wrote: > >> Jim -- >> >> I'm sorry, but your mail missed the deadline (I also see you just added a >> new PDF to the ticket about 15 minutes ago). >> >> I think you have the following options: >> >> - proceed with the last version of the ticket that was announced to the >> Forum (as of last meeting?) >> - submit mpi-report.8.pdf for the December meeting >> >> >> >> On Sep 2, 2014, at 5:27 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: >> >> > Hi Martin, >> > >> > Apologies for missing the 9/1 deadline. Endpoints #380 should be up >> for a first vote in Japan. The working group has made several changes to >> the ticket based on feedback from the prior reading. We feel that these >> changes are ticket 0, and would like to read them prior to a vote. If the >> forum feels the changes are not ticket 0, we would like to re-read at the >> Japan meeting. >> > >> > We do have an unmet implementation requirement, which we will satisfy >> prior to the second vote. >> > >> > Cheers, >> > ~Jim. >> > >> > >> > On Tue, Sep 2, 2014 at 11:44 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: >> > Hi Martin, >> > >> > This is not a vote, just a formal reading. We were asked to make some >> changes and read it again at the last meeting. >> > >> > — Pavan >> > >> > On Sep 2, 2014, at 10:30 AM, Schulz Martin <schulzm(a)llnl.gov> wrote: >> > >> > > Hi Pavan, >> > > >> > > I think we can still add this, despite the two week deadline, since >> we did >> > > talk about this in the last meeting. What I don’t remember, though, >> is the >> > > state of the ticket: it’s listed as scheduled for vote and also >> listed as >> > > a reading in the June notes, but the ticket has two options on it and >> it’s >> > > not clear which one is the intended one. What do you want to do with >> it? >> > > >> > > Thanks! >> > > >> > > Martin >> > > >> > > >> > > >> ________________________________________________________________________ >> > > Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ >> > > CASC @ Lawrence Livermore National Laboratory, Livermore, USA >> > > >> > > >> > > >> > > >> > > >> > > On 9/2/14, 6:12 AM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: >> > > >> > >> Hi Martin, >> > >> >> > >> You missed 411: >> > >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 >> > >> >> > >> We had decided to postpone it to this meeting last time. >> > >> >> > >> Can you add it too? >> > >> >> > >> Thanks, >> > >> >> > >> — Pavan >> > >> >> > >> On Sep 2, 2014, at 3:59 AM, Schulz Martin <schulzm(a)llnl.gov> wrote: >> > >> >> > >>> Hi Rolf, all, >> > >>> >> > >>> I have updated the agenda with all tickets to be read and voted on >> - I >> > >>> only got Rolf’s request below, please let me know asap in case I >> missed >> > >>> something (plus I moved the first votes from last time up to the >> second >> > >>> vote stage). With that, the agenda should be complete. >> > >>> >> > >>> Rolf: all the tickets are on the agenda. There are a few, though, >> as you >> > >>> also noted, that are ticket 0 changes - I grouped them separately >> with >> > >>> the >> > >>> intent that we don’t vote on them (but if we decide otherwise, they >> > >>> count >> > >>> as announced). I know we are a bit sparse in the chapter committees >> > >>> right >> > >>> now, but we need to fix this anyway to make MPI 3.1 happen. I am >> > >>> planning >> > >>> on pinging people separately about this in the next days - if they >> are >> > >>> any >> > >>> volunteers right away, please let me know. >> > >>> >> > >>> Thanks! >> > >>> >> > >>> Martin >> > >>> >> > >>> >> ________________________________________________________________________ >> > >>> Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ >> > >>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >> > >>> >> > >>> >> > >>> >> > >>> >> > >>> >> > >>> On 8/28/14, 8:30 AM, "Rolf Rabenseifner" <rabenseifner(a)hlrs.de> >> wrote: >> > >>> >> > >>>> Dear Martin, >> > >>>> >> > >>>> please add the following errata tickets to the agenda : >> > >>>> >> > >>>> #456 MPI 3.0 Errata: MPI RMA synchronization for MPI shared memory >> > >>>> >> > >>>> This is a large ticket that needs discussion >> > >>>> first in the RMA-WG, second in the plenary, and >> > >>>> hopefully some voting (corrctions and final) >> > >>>> >> > >>>> ____ >> > >>>> Small Fortran corrections are: >> > >>>> >> > >>>> #433 MPI 3.0 Errata: MPI_WIN_ATTACH must only be for simply >> contiguous >> > >>>> buffers >> > >>>> #440 Fortran cross-reference correction (ticket 0 / errata) >> > >>>> #441 Missing text for Fortran MPI_SUBARRAYS_SUPPORTED equals >> .FALSE. >> > >>>> (errata) >> > >>>> #442 Fortran "vector subscipts" and MPI_BOTTOM (ticket 0 / errata) >> > >>>> #443 Fortran MPI_BOTTOM not for initialization or assignment >> (ticket 0 >> > >>>> / >> > >>>> errata) >> > >>>> #444 Fortran and Persistent Requests (ticket 0 / errata) >> > >>>> #445 Fortran and Permament Data Movement: "All" nonblocking >> (ticket 0 / >> > >>>> errata) >> > >>>> #446 Fortran cross-reference update (ticket-0/errata) >> > >>>> >> > >>>> ____ >> > >>>> Fortran and absolute address text corrections: >> > >>>> #431 MPI 3.0 Errata: MPI Address Errata >> > >>>> >> > >>>> ____ >> > >>>> Errata and ticket-0 additions to tickets #349, #402, #404 >> > >>>> (current status of the triple 1st vote): >> > >>>> #421 MPI_Aint addressing arithmetic (ticket-0 additions) >> > >>>> >> > >>>> ____ >> > >>>> Small errata corrections to Chap.2: >> > >>>> >> > >>>> #447 Chap.2 errata: Tools section is an exception: no Fortran >> interface >> > >>>> #448 Chap.2 errata: Semantic term "matching" is missing >> > >>>> #449 Chap.2 errata: Missing "in C" in wording about >> handle-comparison >> > >>>> #450 Chap.2 errata: Outdated cross-reference to MPI-1 >> > >>>> #451 Chap.2 errata: Correcting description of nonblocking >> > >>>> #452 Chap.2 errata: Wording correction about deprecated and removed >> > >>>> functions >> > >>>> #453 Chap.2 errata: Avoidance of mpi_ / pmpi_ in C due to Fortran >> > >>>> reasons >> > >>>> >> > >>>> #455 Chap.2 errata: Removal of duplicated text on MPI_Aint and >> > >>>> MPI_Offset >> > >>>> >> > >>>> >> > >>>> _____________________________________ >> > >>>> The rest of this mail are only hints. >> > >>>> ____ >> > >>>> I also expect that Jeff may want to update, discuss and vote: >> > >>>> #428 Add more non-profile-able MPI routines >> > >>>> >> > >>>> ____ >> > >>>> I also expect that Kathryn may want to update, discuss and vote: >> > >>>> #383 Consistency of tools information interface variable >> information >> > >>>> #368 MPI_T Minor Errata >> > >>>> >> > >>>> ____ >> > >>>> I also expect that Bill may want to update, discuss and vote: >> > >>>> #438 Interpretation of MPI_WIN_BASE for shared memory windows >> > >>>> >> > >>>> ____ >> > >>>> Other MPI-3.0 errata tickets that may or should be handled before >> > >>>> MPI-3.1 >> > >>>> is finalized: >> > >>>> >> > >>>> #420 Bug in neighbourhood collective example >> > >>>> - due to missing chapter committee action, >> > >>>> it shoud be voted in >> > >>>> >> > >>>> ____ >> > >>>> The only MPI-3.0-errata ticket that may need to go MPI-4.0 is: >> > >>>> #374 Comm_free/Type_free/Win_free: clarify attribute destructor >> > >>>> semantics >> > >>>> >> > >>>> >> > >>>> ____ >> > >>>> Tickets #428, #368, #368, #420 may need a milestone-change to >> "Kobe". >> > >>>> Otherwise, they are invisible in the Kobe tickets list: >> > >>>> >> > >>>> >> https://svn.mpi-forum.org/trac/mpi-forum-web/query?status=accepted&statu >> > >>>> s= >> > >>>> >> > >>>> >> assigned&status=new&status=reopened&milestone=2014%2F09%2F15+Kobe%2C+Jap >> > >>>> an >> > >>>> >> > >>>> >> &col=id&col=summary&col=milestone&col=status&col=type&col=priority&col=v >> > >>>> er >> > >>>> sion&order=priority >> > >>>> >> > >>>> Best regards >> > >>>> Rolf >> > >>>> >> > >>>> ----- Original Message ----- >> > >>>>> From: "Schulz Martin" <schulzm(a)llnl.gov> >> > >>>>> To: "Main mailing list" <mpi-forum(a)lists.mpi-forum.org> >> > >>>>> Sent: Thursday, August 28, 2014 9:53:01 AM >> > >>>>> Subject: [Mpi-forum] September MPI Forum / Agenda and Reminder to >> > >>>>> Register >> > >>>>> >> > >>>>> Hi all, >> > >>>>> >> > >>>>> Based on the (sparse) feedback that I have gotten, I have >> assembled >> > >>>>> an >> > >>>>> initial agenda and uploaded it to the server. Please check the >> > >>>>> agenda, >> > >>>>> especially if you sent me a request, and double check that I >> didn¹t >> > >>>>> miss >> > >>>>> anything or anybody. If something is missing, please let me know >> > >>>>> asap. >> > >>>>> >> > >>>>> As usual, I tried to split working groups and plenary on separate >> > >>>>> days, >> > >>>>> with votes on day three. However, so far I have only received one >> WG >> > >>>>> meeting request, which would make Monday relatively sparse. We >> could >> > >>>>> move >> > >>>>> parts of Tuesday to that day and probably finish on Tuesday >> evening. >> > >>>>> Some >> > >>>>> people have voiced interest in doing this. However, this would >> move >> > >>>>> votes >> > >>>>> onto the same day as the second half of the plenary and I would >> like >> > >>>>> some >> > >>>>> feedback on this (since the voting times are typically fixed to >> the >> > >>>>> last >> > >>>>> day). So the options are: >> > >>>>> >> > >>>>> Monday WG time (with open, general WG time), Tuesday plenary, >> > >>>>> Wednesday >> > >>>>> votes >> > >>>>> >> > >>>>> -or- >> > >>>>> >> > >>>>> Monday WG and plenary time (errata), Tuesday plenary and votes, >> > >>>>> Wednesday >> > >>>>> off >> > >>>>> >> > >>>>> Please let me know if you have a preference one way or the other. >> If >> > >>>>> I >> > >>>>> don¹t hear anything or if there are objections to moving the >> voting >> > >>>>> block, >> > >>>>> we¹ll leave things as is (first option). >> > >>>>> >> > >>>>> Thanks! >> > >>>>> >> > >>>>> Martin >> > >>>>> >> > >>>>> PS: Please don¹t forget to register, if you haven¹t done so, by >> the >> > >>>>> end of >> > >>>>> August (just a few days left). We need an exact attendee list, so >> > >>>>> people >> > >>>>> can get access to Riken. >> > >>>>> >> > >>>>> >> > >>>>> >> _______________________________________________________________________ >> > >>>>> _ >> > >>>>> Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ >> > >>>>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >> > >>>>> >> > >>>>> >> > >>>>> _______________________________________________ >> > >>>>> mpi-forum mailing list >> > >>>>> mpi-forum(a)lists.mpi-forum.org >> > >>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > >>>>> >> > >>>> >> > >>>> -- >> > >>>> Dr. Rolf Rabenseifner . . . . . . . . . .. email >> rabenseifner(a)hlrs.de >> > >>>> High Performance Computing Center (HLRS) . phone >> ++49(0)711/685-65530 >> > >>>> University of Stuttgart . . . . . . . . .. fax ++49(0)711 / >> 685-65832 >> > >>>> Head of Dpmt Parallel Computing . . . >> www.hlrs.de/people/rabenseifner >> > >>>> Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room >> 1.307) >> > >>>> _______________________________________________ >> > >>>> mpi-forum mailing list >> > >>>> mpi-forum(a)lists.mpi-forum.org >> > >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > >>> >> > >>> >> > >>> _______________________________________________ >> > >>> mpi-forum mailing list >> > >>> mpi-forum(a)lists.mpi-forum.org >> > >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > >> >> > >> -- >> > >> Pavan Balaji ✉️ >> > >> http://www.mcs.anl.gov/~balaji >> > >> >> > >> _______________________________________________ >> > >> mpi-forum mailing list >> > >> mpi-forum(a)lists.mpi-forum.org >> > >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > > >> > > >> > > _______________________________________________ >> > > mpi-forum mailing list >> > > mpi-forum(a)lists.mpi-forum.org >> > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > >> > -- >> > Pavan Balaji ✉️ >> > http://www.mcs.anl.gov/~balaji >> > >> > _______________________________________________ >> > mpi-forum mailing list >> > mpi-forum(a)lists.mpi-forum.org >> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > >> > _______________________________________________ >> > mpi-forum mailing list >> > mpi-forum(a)lists.mpi-forum.org >> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to: >> http://www.cisco.com/web/about/doing_business/legal/cri/ >> >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > >

1 0

Re: [Mpi-forum] September MPI Forum / Agenda and Reminder to Register
by Jim Dinan 04 Sep '14

04 Sep '14

Jeff, The version of the endpoints proposal that was read at the last meeting is proceeding for a vote, with the addition of ticket 0 changes that the working group made in response to feedback at the last reading. My apologies for being a day late on the deadline. The timing was a little tricky this time around with the deadline landing on a Monday and a US holiday. ~Jim. On Tue, Sep 2, 2014 at 6:23 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > Jim -- > > I'm sorry, but your mail missed the deadline (I also see you just added a > new PDF to the ticket about 15 minutes ago). > > I think you have the following options: > > - proceed with the last version of the ticket that was announced to the > Forum (as of last meeting?) > - submit mpi-report.8.pdf for the December meeting > > > > On Sep 2, 2014, at 5:27 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > > > Hi Martin, > > > > Apologies for missing the 9/1 deadline. Endpoints #380 should be up for > a first vote in Japan. The working group has made several changes to the > ticket based on feedback from the prior reading. We feel that these > changes are ticket 0, and would like to read them prior to a vote. If the > forum feels the changes are not ticket 0, we would like to re-read at the > Japan meeting. > > > > We do have an unmet implementation requirement, which we will satisfy > prior to the second vote. > > > > Cheers, > > ~Jim. > > > > > > On Tue, Sep 2, 2014 at 11:44 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Hi Martin, > > > > This is not a vote, just a formal reading. We were asked to make some > changes and read it again at the last meeting. > > > > — Pavan > > > > On Sep 2, 2014, at 10:30 AM, Schulz Martin <schulzm(a)llnl.gov> wrote: > > > > > Hi Pavan, > > > > > > I think we can still add this, despite the two week deadline, since we > did > > > talk about this in the last meeting. What I don’t remember, though, is > the > > > state of the ticket: it’s listed as scheduled for vote and also listed > as > > > a reading in the June notes, but the ticket has two options on it and > it’s > > > not clear which one is the intended one. What do you want to do with > it? > > > > > > Thanks! > > > > > > Martin > > > > > > > > > > ________________________________________________________________________ > > > Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > > > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > > > > > > > > > > > > > > > > On 9/2/14, 6:12 AM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > > > > > >> Hi Martin, > > >> > > >> You missed 411: > > >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/411 > > >> > > >> We had decided to postpone it to this meeting last time. > > >> > > >> Can you add it too? > > >> > > >> Thanks, > > >> > > >> — Pavan > > >> > > >> On Sep 2, 2014, at 3:59 AM, Schulz Martin <schulzm(a)llnl.gov> wrote: > > >> > > >>> Hi Rolf, all, > > >>> > > >>> I have updated the agenda with all tickets to be read and voted on - > I > > >>> only got Rolf’s request below, please let me know asap in case I > missed > > >>> something (plus I moved the first votes from last time up to the > second > > >>> vote stage). With that, the agenda should be complete. > > >>> > > >>> Rolf: all the tickets are on the agenda. There are a few, though, as > you > > >>> also noted, that are ticket 0 changes - I grouped them separately > with > > >>> the > > >>> intent that we don’t vote on them (but if we decide otherwise, they > > >>> count > > >>> as announced). I know we are a bit sparse in the chapter committees > > >>> right > > >>> now, but we need to fix this anyway to make MPI 3.1 happen. I am > > >>> planning > > >>> on pinging people separately about this in the next days - if they > are > > >>> any > > >>> volunteers right away, please let me know. > > >>> > > >>> Thanks! > > >>> > > >>> Martin > > >>> > > >>> > ________________________________________________________________________ > > >>> Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > > >>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> On 8/28/14, 8:30 AM, "Rolf Rabenseifner" <rabenseifner(a)hlrs.de> > wrote: > > >>> > > >>>> Dear Martin, > > >>>> > > >>>> please add the following errata tickets to the agenda : > > >>>> > > >>>> #456 MPI 3.0 Errata: MPI RMA synchronization for MPI shared memory > > >>>> > > >>>> This is a large ticket that needs discussion > > >>>> first in the RMA-WG, second in the plenary, and > > >>>> hopefully some voting (corrctions and final) > > >>>> > > >>>> ____ > > >>>> Small Fortran corrections are: > > >>>> > > >>>> #433 MPI 3.0 Errata: MPI_WIN_ATTACH must only be for simply > contiguous > > >>>> buffers > > >>>> #440 Fortran cross-reference correction (ticket 0 / errata) > > >>>> #441 Missing text for Fortran MPI_SUBARRAYS_SUPPORTED equals .FALSE. > > >>>> (errata) > > >>>> #442 Fortran "vector subscipts" and MPI_BOTTOM (ticket 0 / errata) > > >>>> #443 Fortran MPI_BOTTOM not for initialization or assignment > (ticket 0 > > >>>> / > > >>>> errata) > > >>>> #444 Fortran and Persistent Requests (ticket 0 / errata) > > >>>> #445 Fortran and Permament Data Movement: "All" nonblocking (ticket > 0 / > > >>>> errata) > > >>>> #446 Fortran cross-reference update (ticket-0/errata) > > >>>> > > >>>> ____ > > >>>> Fortran and absolute address text corrections: > > >>>> #431 MPI 3.0 Errata: MPI Address Errata > > >>>> > > >>>> ____ > > >>>> Errata and ticket-0 additions to tickets #349, #402, #404 > > >>>> (current status of the triple 1st vote): > > >>>> #421 MPI_Aint addressing arithmetic (ticket-0 additions) > > >>>> > > >>>> ____ > > >>>> Small errata corrections to Chap.2: > > >>>> > > >>>> #447 Chap.2 errata: Tools section is an exception: no Fortran > interface > > >>>> #448 Chap.2 errata: Semantic term "matching" is missing > > >>>> #449 Chap.2 errata: Missing "in C" in wording about > handle-comparison > > >>>> #450 Chap.2 errata: Outdated cross-reference to MPI-1 > > >>>> #451 Chap.2 errata: Correcting description of nonblocking > > >>>> #452 Chap.2 errata: Wording correction about deprecated and removed > > >>>> functions > > >>>> #453 Chap.2 errata: Avoidance of mpi_ / pmpi_ in C due to Fortran > > >>>> reasons > > >>>> > > >>>> #455 Chap.2 errata: Removal of duplicated text on MPI_Aint and > > >>>> MPI_Offset > > >>>> > > >>>> > > >>>> _____________________________________ > > >>>> The rest of this mail are only hints. > > >>>> ____ > > >>>> I also expect that Jeff may want to update, discuss and vote: > > >>>> #428 Add more non-profile-able MPI routines > > >>>> > > >>>> ____ > > >>>> I also expect that Kathryn may want to update, discuss and vote: > > >>>> #383 Consistency of tools information interface variable information > > >>>> #368 MPI_T Minor Errata > > >>>> > > >>>> ____ > > >>>> I also expect that Bill may want to update, discuss and vote: > > >>>> #438 Interpretation of MPI_WIN_BASE for shared memory windows > > >>>> > > >>>> ____ > > >>>> Other MPI-3.0 errata tickets that may or should be handled before > > >>>> MPI-3.1 > > >>>> is finalized: > > >>>> > > >>>> #420 Bug in neighbourhood collective example > > >>>> - due to missing chapter committee action, > > >>>> it shoud be voted in > > >>>> > > >>>> ____ > > >>>> The only MPI-3.0-errata ticket that may need to go MPI-4.0 is: > > >>>> #374 Comm_free/Type_free/Win_free: clarify attribute destructor > > >>>> semantics > > >>>> > > >>>> > > >>>> ____ > > >>>> Tickets #428, #368, #368, #420 may need a milestone-change to > "Kobe". > > >>>> Otherwise, they are invisible in the Kobe tickets list: > > >>>> > > >>>> > https://svn.mpi-forum.org/trac/mpi-forum-web/query?status=accepted&statu > > >>>> s= > > >>>> > > >>>> > assigned&status=new&status=reopened&milestone=2014%2F09%2F15+Kobe%2C+Jap > > >>>> an > > >>>> > > >>>> > &col=id&col=summary&col=milestone&col=status&col=type&col=priority&col=v > > >>>> er > > >>>> sion&order=priority > > >>>> > > >>>> Best regards > > >>>> Rolf > > >>>> > > >>>> ----- Original Message ----- > > >>>>> From: "Schulz Martin" <schulzm(a)llnl.gov> > > >>>>> To: "Main mailing list" <mpi-forum(a)lists.mpi-forum.org> > > >>>>> Sent: Thursday, August 28, 2014 9:53:01 AM > > >>>>> Subject: [Mpi-forum] September MPI Forum / Agenda and Reminder to > > >>>>> Register > > >>>>> > > >>>>> Hi all, > > >>>>> > > >>>>> Based on the (sparse) feedback that I have gotten, I have assembled > > >>>>> an > > >>>>> initial agenda and uploaded it to the server. Please check the > > >>>>> agenda, > > >>>>> especially if you sent me a request, and double check that I didn¹t > > >>>>> miss > > >>>>> anything or anybody. If something is missing, please let me know > > >>>>> asap. > > >>>>> > > >>>>> As usual, I tried to split working groups and plenary on separate > > >>>>> days, > > >>>>> with votes on day three. However, so far I have only received one > WG > > >>>>> meeting request, which would make Monday relatively sparse. We > could > > >>>>> move > > >>>>> parts of Tuesday to that day and probably finish on Tuesday > evening. > > >>>>> Some > > >>>>> people have voiced interest in doing this. However, this would move > > >>>>> votes > > >>>>> onto the same day as the second half of the plenary and I would > like > > >>>>> some > > >>>>> feedback on this (since the voting times are typically fixed to the > > >>>>> last > > >>>>> day). So the options are: > > >>>>> > > >>>>> Monday WG time (with open, general WG time), Tuesday plenary, > > >>>>> Wednesday > > >>>>> votes > > >>>>> > > >>>>> -or- > > >>>>> > > >>>>> Monday WG and plenary time (errata), Tuesday plenary and votes, > > >>>>> Wednesday > > >>>>> off > > >>>>> > > >>>>> Please let me know if you have a preference one way or the other. > If > > >>>>> I > > >>>>> don¹t hear anything or if there are objections to moving the voting > > >>>>> block, > > >>>>> we¹ll leave things as is (first option). > > >>>>> > > >>>>> Thanks! > > >>>>> > > >>>>> Martin > > >>>>> > > >>>>> PS: Please don¹t forget to register, if you haven¹t done so, by the > > >>>>> end of > > >>>>> August (just a few days left). We need an exact attendee list, so > > >>>>> people > > >>>>> can get access to Riken. > > >>>>> > > >>>>> > > >>>>> > _______________________________________________________________________ > > >>>>> _ > > >>>>> Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > > >>>>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > >>>>> > > >>>>> > > >>>>> _______________________________________________ > > >>>>> mpi-forum mailing list > > >>>>> mpi-forum(a)lists.mpi-forum.org > > >>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > >>>>> > > >>>> > > >>>> -- > > >>>> Dr. Rolf Rabenseifner . . . . . . . . . .. email > rabenseifner(a)hlrs.de > > >>>> High Performance Computing Center (HLRS) . phone > ++49(0)711/685-65530 > > >>>> University of Stuttgart . . . . . . . . .. fax ++49(0)711 / > 685-65832 > > >>>> Head of Dpmt Parallel Computing . . . > www.hlrs.de/people/rabenseifner > > >>>> Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room > 1.307) > > >>>> _______________________________________________ > > >>>> mpi-forum mailing list > > >>>> mpi-forum(a)lists.mpi-forum.org > > >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > >>> > > >>> > > >>> _______________________________________________ > > >>> mpi-forum mailing list > > >>> mpi-forum(a)lists.mpi-forum.org > > >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > >> > > >> -- > > >> Pavan Balaji ✉️ > > >> http://www.mcs.anl.gov/~balaji > > >> > > >> _______________________________________________ > > >> mpi-forum mailing list > > >> mpi-forum(a)lists.mpi-forum.org > > >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > > > > > > _______________________________________________ > > > mpi-forum mailing list > > > mpi-forum(a)lists.mpi-forum.org > > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > -- > > Pavan Balaji ✉️ > > http://www.mcs.anl.gov/~balaji > > > > _______________________________________________ > > mpi-forum mailing list > > mpi-forum(a)lists.mpi-forum.org > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > _______________________________________________ > > mpi-forum mailing list > > mpi-forum(a)lists.mpi-forum.org > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: > http://www.cisco.com/web/about/doing_business/legal/cri/ > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >

1 0

Re: [Mpi-forum] Timeline for MPI 3.1
by Jeff Hammond 03 Sep '14

03 Sep '14

I support getting it right late as opposed to the timely alternative. There is no point in having a 3.1 release that doesn't cover all the critical errata. Jeff On Tue, Sep 2, 2014 at 5:03 PM, Schulz Martin <schulzm(a)llnl.gov> wrote: > Hi all, > > After discussing with the steering committee, I wanted to bring up the > following: > > In one of the previous meetings we decided on the timeline for MPI 3.1: the > intent is to ratify the complete MPI 3.1 standard by the December meeting, > which makes the upcoming September meeting the last meeting to allow for > errata votes (actual additions had to be read in March). The main goal of > the MPI 3.1 document is/was to create a “clean” standard that integrates > outstanding errata items and the small additions we worked on. While the > latter seem to be more or less done, we have multiple outstanding errata > items (in particular in RMA, but also a few in tools, potentially also in > other areas) that won’t get done by the September meeting (the respective WG > chairs felt their wasn’t enough time or participation in the Japan meeting > for their WG to allow for enough discussion). > > If we stick with the current timeline, we’ll have an MPI 3.1 and yet again a > good number of errata items next to it, which is counter the intention for > publishing an intermediate version before MPI 4.0. This means we are either > back in the same situation as now with MPI 3.0 for quite a while longer or > we have to go for an MPI 3.2. Neither option seems good. > > Therefore, I would like to propose to not rush it and to slip the schedule > by one (or if needed by two) meetings to make sure we conclude the > outstanding errata items. Otherwise, we’ll end up with a standard that we > know is broken by the time we vote for it (we would have the final document > vote and the MPI 3.1 errata votes in the same meeting). > > Opinions? > Thanks, > > Martin > > > ________________________________________________________________________ > Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [Mpi-forum] September MPI Forum / Agenda and Reminder to Register
by Schulz Martin 03 Sep '14

03 Sep '14

That sounds fine then - I honestly didn’t remember what happened back then and it was not obvious from the ticket. It should be noted in the ticket (I think you said you would do that anyway,so that’s good). Thanks! Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA On 9/2/14, 5:33 PM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: >Hi Martin, > >On Sep 2, 2014, at 6:44 PM, Schulz Martin <schulzm(a)llnl.gov> wrote: >> I found the same things confusing (for a reading) as Jeff. Nevertheless, >> it may be good to add this to agenda to discuss the two options, unless >> you want to lead that discussion yourself next time. > >We already discussed this two meetings ago and decided to only go with >solution 2 (there wasn’t even a debate at that point). I’m happy to >present both options again, if needed, but I think it’s unnecessary to do >it again. > > — Pavan > >-- >Pavan Balaji ✉️ >http://www.mcs.anl.gov/~balaji > >_______________________________________________ >mpi-forum mailing list >mpi-forum(a)lists.mpi-forum.org >http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

1 0

Re: [Mpi-forum] September MPI Forum / Agenda and Reminder to Register
by Balaji, Pavan 02 Sep '14

02 Sep '14

Hi Martin, On Sep 2, 2014, at 6:44 PM, Schulz Martin <schulzm(a)llnl.gov> wrote: > I found the same things confusing (for a reading) as Jeff. Nevertheless, > it may be good to add this to agenda to discuss the two options, unless > you want to lead that discussion yourself next time. We already discussed this two meetings ago and decided to only go with solution 2 (there wasn’t even a debate at that point). I’m happy to present both options again, if needed, but I think it’s unnecessary to do it again. — Pavan -- Pavan Balaji ✉️ http://www.mcs.anl.gov/~balaji

1 0

Timeline for MPI 3.1
by Schulz Martin 02 Sep '14

02 Sep '14

Hi all, After discussing with the steering committee, I wanted to bring up the following: In one of the previous meetings we decided on the timeline for MPI 3.1: the intent is to ratify the complete MPI 3.1 standard by the December meeting, which makes the upcoming September meeting the last meeting to allow for errata votes (actual additions had to be read in March). The main goal of the MPI 3.1 document is/was to create a ³clean² standard that integrates outstanding errata items and the small additions we worked on. While the latter seem to be more or less done, we have multiple outstanding errata items (in particular in RMA, but also a few in tools, potentially also in other areas) that won¹t get done by the September meeting (the respective WG chairs felt their wasn¹t enough time or participation in the Japan meeting for their WG to allow for enough discussion). If we stick with the current timeline, we¹ll have an MPI 3.1 and yet again a good number of errata items next to it, which is counter the intention for publishing an intermediate version before MPI 4.0. This means we are either back in the same situation as now with MPI 3.0 for quite a while longer or we have to go for an MPI 3.2. Neither option seems good. Therefore, I would like to propose to not rush it and to slip the schedule by one (or if needed by two) meetings to make sure we conclude the outstanding errata items. Otherwise, we¹ll end up with a standard that we know is broken by the time we vote for it (we would have the final document vote and the MPI 3.1 errata votes in the same meeting). Opinions? Thanks, Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0