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

Re: [mpiwg-hybridpm] Info Keys Ticket
by Jim Dinan 20 Nov '14

20 Nov '14

Jeff pointed out that he already did this for us on #381 (thanks!). I copied that info over to the new ticket: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/461#comment:2 ~Jim. On Thu, Nov 20, 2014 at 2:50 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Hi Dan, > > Sure, let's gather what we can. For these info keys, the benefit is > fairly straightforward. For communicators with wildcard restrictions you > can use a structure that is much more efficient to search than a single > ordered list. In addition, when you have multiple communication channels > (e.g. shared memory, multiple rails, etc) no_any_source removes a lot of > complications and overheads. > > ~Jim. > > On Thu, Nov 20, 2014 at 2:09 PM, Daniel Holmes <dholmes(a)epcc.ed.ac.uk> > wrote: > >> Hi All, >> >> One of the objections last time was that these keys seem like a good idea >> for performance but there's no implementation/proof. We'll need to collate >> some of the existing research that shows this gives a performance benefit >> in order to address that objection. >> >> Cheers, >> Dan. >> >> >> On 20/11/2014 13:03, Jim Dinan wrote: >> >> Hi All, >> >> I would like to bring forward the following ticket on communicator info >> keys for a formal reading at the upcoming meeting: >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/461. >> >> This ticket contains the no_any_tag and no_any_source info keys from >> #381 on which we were able to reach consensus. There are lots of other >> good ideas on #318, and we will continue to work on moving those ahead. >> >> Please send feedback. The deadline for changes to the proposal is >> Sunday. >> >> Thanks, >> ~Jim. >> >> >> _______________________________________________ >> mpiwg-hybridpm mailing [email protected]://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm >> >> >> -- >> 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(a)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-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] Info Keys Ticket
by Jim Dinan 20 Nov '14

20 Nov '14

Hi Dan, Sure, let's gather what we can. For these info keys, the benefit is fairly straightforward. For communicators with wildcard restrictions you can use a structure that is much more efficient to search than a single ordered list. In addition, when you have multiple communication channels (e.g. shared memory, multiple rails, etc) no_any_source removes a lot of complications and overheads. ~Jim. On Thu, Nov 20, 2014 at 2:09 PM, Daniel Holmes <dholmes(a)epcc.ed.ac.uk> wrote: > Hi All, > > One of the objections last time was that these keys seem like a good idea > for performance but there's no implementation/proof. We'll need to collate > some of the existing research that shows this gives a performance benefit > in order to address that objection. > > Cheers, > Dan. > > > On 20/11/2014 13:03, Jim Dinan wrote: > > Hi All, > > I would like to bring forward the following ticket on communicator info > keys for a formal reading at the upcoming meeting: > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/461. > > This ticket contains the no_any_tag and no_any_source info keys from > #381 on which we were able to reach consensus. There are lots of other > good ideas on #318, and we will continue to work on moving those ahead. > > Please send feedback. The deadline for changes to the proposal is > Sunday. > > Thanks, > ~Jim. > > > _______________________________________________ > mpiwg-hybridpm mailing [email protected]://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm > > > -- > 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(a)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-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] Info Keys Ticket
by Daniel Holmes 20 Nov '14

20 Nov '14

Hi All, One of the objections last time was that these keys seem like a good idea for performance but there's no implementation/proof. We'll need to collate some of the existing research that shows this gives a performance benefit in order to address that objection. Cheers, Dan. On 20/11/2014 13:03, Jim Dinan wrote: > Hi All, > > I would like to bring forward the following ticket on communicator > info keys for a formal reading at the upcoming meeting: > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/461. > > This ticket contains the no_any_tag and no_any_source info keys from > #381 on which we were able to reach consensus. There are lots of other > good ideas on #318, and we will continue to work on moving those ahead. > > Please send feedback. The deadline for changes to the proposal is Sunday. > > Thanks, > ~Jim. > > > _______________________________________________ > mpiwg-hybridpm mailing list > mpiwg-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm -- 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(a)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.

1 0

Info Keys Ticket
by Jim Dinan 20 Nov '14

20 Nov '14

Hi All, I would like to bring forward the following ticket on communicator info keys for a formal reading at the upcoming meeting: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/461. This ticket contains the no_any_tag and no_any_source info keys from #381 on which we were able to reach consensus. There are lots of other good ideas on #318, and we will continue to work on moving those ahead. Please send feedback. The deadline for changes to the proposal is Sunday. Thanks, ~Jim.

1 0

Re: [mpiwg-hybridpm] Meeting Today
by Jim Dinan 17 Nov '14

17 Nov '14

All, I noticed that I had a small (but very good) edit lingering on my hard drive from the September Forum meeting. This change was given by Aurelien (thanks!). Updated proposal is mpi-report.9.pdf. Index: chap-context/context.tex =================================================================== --- chap-context/context.tex (revision 1824) +++ chap-context/context.tex (working copy) @@ -1241,7 +1241,7 @@ \mpiarg{new\_comm\_handles} argument is ignored. If a process provides a valid \mpiarg{my\_num\_ep} argument, but the MPI implementation is not able to create a new communicator because of the -\mpiarg{my\_num\_ep} argument at this process, this function will return +\mpiarg{my\_num\_ep} argument at this process, this function will raise an exception of type \error{MPI\_ERR\_ENDPOINTS}. ~Jim. On Mon, Nov 17, 2014 at 12:20 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Looks like today's meeting isn't going to happen. Either the Webex is > kaput or everyone is occupied with SC-related activities. :) > > Here are the items that the WG has lined up for the December meeting, > along with TODOs for WG members: > > 1) Dramatic reading of endpoints (mpi-report.8.pdf @ > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/380) > TODO: All WG members, please re-read and look closely at the > MPI_ERR_ENDPOINTS text > 2) Dan will present his endpoints work > 3) Jim will present his SC '14 paper on endpoints > 4) WG session > TODO: Send items for the agenda (e.g. ticket numbers) > > Please respond with any additions, comments, and feedback. > > See you in December, > ~Jim. > > On Mon, Nov 17, 2014 at 12:09 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > >> All, >> >> Please let me know if you are having trouble connecting. The web link >> doesn't seem to be working, but I was able to connect via the dial-in >> number. >> >> ~Jim. >> >> On Mon, Nov 17, 2014 at 10:29 AM, Jim Dinan <james.dinan(a)gmail.com> >> wrote: >> >>> Hi All, >>> >>> This is a reminder that we are meeting today at 12pm ET / 11am CT / 9am >>> PT. This is our last WG meeting scheduled before the agenda cutoff for the >>> upcoming December F2F. >>> >>> See you then, >>> ~Jim. >>> >>> =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= >>> >>> Meeting Info: >>> >>> >>> https://cisco.webex.com/ciscosales/j.php?ED=236535652&UID=0&PW=NOGE0NDk5MmV… >>> >>> +1-866-432-9903 >>> Meeting ID: 206095536 >>> >>> https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid >>> >>> =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= >>> >> >> >

1 0

Re: [mpiwg-hybridpm] Meeting Today
by Jim Dinan 17 Nov '14

17 Nov '14

Looks like today's meeting isn't going to happen. Either the Webex is kaput or everyone is occupied with SC-related activities. :) Here are the items that the WG has lined up for the December meeting, along with TODOs for WG members: 1) Dramatic reading of endpoints (mpi-report.8.pdf @ https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/380) TODO: All WG members, please re-read and look closely at the MPI_ERR_ENDPOINTS text 2) Dan will present his endpoints work 3) Jim will present his SC '14 paper on endpoints 4) WG session TODO: Send items for the agenda (e.g. ticket numbers) Please respond with any additions, comments, and feedback. See you in December, ~Jim. On Mon, Nov 17, 2014 at 12:09 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > All, > > Please let me know if you are having trouble connecting. The web link > doesn't seem to be working, but I was able to connect via the dial-in > number. > > ~Jim. > > On Mon, Nov 17, 2014 at 10:29 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > >> Hi All, >> >> This is a reminder that we are meeting today at 12pm ET / 11am CT / 9am >> PT. This is our last WG meeting scheduled before the agenda cutoff for the >> upcoming December F2F. >> >> See you then, >> ~Jim. >> >> =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= >> >> Meeting Info: >> >> >> https://cisco.webex.com/ciscosales/j.php?ED=236535652&UID=0&PW=NOGE0NDk5MmV… >> >> +1-866-432-9903 >> Meeting ID: 206095536 >> >> https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid >> >> =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= >> > >

1 0

Re: [mpiwg-hybridpm] Meeting Today
by Jim Dinan 17 Nov '14

17 Nov '14

All, Please let me know if you are having trouble connecting. The web link doesn't seem to be working, but I was able to connect via the dial-in number. ~Jim. On Mon, Nov 17, 2014 at 10:29 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Hi All, > > This is a reminder that we are meeting today at 12pm ET / 11am CT / 9am > PT. This is our last WG meeting scheduled before the agenda cutoff for the > upcoming December F2F. > > See you then, > ~Jim. > > =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= > > Meeting Info: > > > https://cisco.webex.com/ciscosales/j.php?ED=236535652&UID=0&PW=NOGE0NDk5MmV… > > +1-866-432-9903 > Meeting ID: 206095536 > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid > > =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= >

1 0

Meeting Today
by Jim Dinan 17 Nov '14

17 Nov '14

Hi All, This is a reminder that we are meeting today at 12pm ET / 11am CT / 9am PT. This is our last WG meeting scheduled before the agenda cutoff for the upcoming December F2F. See you then, ~Jim. =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= Meeting Info: https://cisco.webex.com/ciscosales/j.php?ED=236535652&UID=0&PW=NOGE0NDk5MmV… +1-866-432-9903 Meeting ID: 206095536 https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

1 0

Re: [mpiwg-hybridpm] Fwd: [Mpi-forum] December 2014 Meeting / Agenda
by Jim Dinan 10 Nov '14

10 Nov '14

Sounds good. Thanks! On Mon, Nov 10, 2014 at 3:01 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > I'd like to ask for some WG time too. > > So reading + plenary + WG time? > > -- Pavan > > Sent from my iPhone > > On Nov 9, 2014, at 8:29 PM, Jim Dinan <james.dinan(a)gmail.com> wrote: > > I haven't figure out yet whether I will be attending. If I do, I would > like to do a plenary session to present my SC paper on endpoints. We're > planning to do the formal reading of endpoints at this meeting. > > On Tue, Nov 4, 2014 at 1:11 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > >> >> Do we need WG or plenary time at the next meeting? >> >> — Pavan >> >> Begin forwarded message: >> >> *Date: *November 4, 2014 at 12:09:51 AM CST >> *From: *Schulz Martin <schulzm(a)llnl.gov> >> *To: *Main mailing list <mpi-forum(a)lists.mpi-forum.org> >> *Subject: **[Mpi-forum] December 2014 Meeting / Agenda* >> *Reply-To: *Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org> >> >> Hi all, >> >> I put up an initial roster for the December meeting in San Jose. As >> usual, I am planning to arrange WG meeting in the beginning of the meeting >> (until Tuesday noon or evening depending on the requests that I get), >> followed by the plenaries. Votes will be Thursday morning. >> >> For the plenaries, I would like to try something different, based on >> the experience at the forum in Japan: instead of having a fixed schedule, >> I’ll put up a list of topics that we need to cover and we’ll go through >> them in order, but not necessarily with a fixed schedule. We basically >> followed this scheme in Japan (out of necessity since we had to redo the >> agenda on the fly due to the missed quorum) and this worked fairly well and >> allowed for enough discussions for the topics that needed them. I’ll still >> try to keep track of overall time to make sure we get through with the >> agenda, in particular with the items to be voted on, but I would like to >> avoid that this dictates the schedule. >> >> For the WG leads: please send me any requests for WG time or plenary >> items as soon as you can (even if you need to revise that later), so we can >> start building a more concrete agenda. Also, please let me know of any >> tickets that you want read or voted on as part of errata. I have >> pre-populated the agenda with the second and first votes, since those are >> fixed – but please also double check those. The deadline for announcing >> tickets for readings or votes is as usual 2 weeks before the forum, which >> is the Monday 11/24, I.e., the Monday of Thanksgiving week. >> >> 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 >> >> >> -- >> Pavan Balaji ✉️ >> http://www.mcs.anl.gov/~balaji >> >> >> _______________________________________________ >> mpiwg-hybridpm mailing list >> mpiwg-hybridpm(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm >> > > _______________________________________________ > mpiwg-hybridpm mailing list > mpiwg-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm > > > _______________________________________________ > 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] Fwd: [Mpi-forum] December 2014 Meeting / Agenda
by Balaji, Pavan 10 Nov '14

10 Nov '14

I'd like to ask for some WG time too. So reading + plenary + WG time? -- Pavan Sent from my iPhone On Nov 9, 2014, at 8:29 PM, Jim Dinan <james.dinan(a)gmail.com<mailto:[email protected]>> wrote: I haven't figure out yet whether I will be attending. If I do, I would like to do a plenary session to present my SC paper on endpoints. We're planning to do the formal reading of endpoints at this meeting. On Tue, Nov 4, 2014 at 1:11 AM, Balaji, Pavan <balaji(a)anl.gov<mailto:[email protected]>> wrote: Do we need WG or plenary time at the next meeting? — Pavan Begin forwarded message: Date: November 4, 2014 at 12:09:51 AM CST From: Schulz Martin <schulzm(a)llnl.gov<mailto:[email protected]>> To: Main mailing list <mpi-forum(a)lists.mpi-forum.org<mailto:[email protected]>> Subject: [Mpi-forum] December 2014 Meeting / Agenda Reply-To: Main MPI Forum mailing list <mpi-forum(a)lists.mpi-forum.org<mailto:[email protected]>> Hi all, I put up an initial roster for the December meeting in San Jose. As usual, I am planning to arrange WG meeting in the beginning of the meeting (until Tuesday noon or evening depending on the requests that I get), followed by the plenaries. Votes will be Thursday morning. For the plenaries, I would like to try something different, based on the experience at the forum in Japan: instead of having a fixed schedule, I’ll put up a list of topics that we need to cover and we’ll go through them in order, but not necessarily with a fixed schedule. We basically followed this scheme in Japan (out of necessity since we had to redo the agenda on the fly due to the missed quorum) and this worked fairly well and allowed for enough discussions for the topics that needed them. I’ll still try to keep track of overall time to make sure we get through with the agenda, in particular with the items to be voted on, but I would like to avoid that this dictates the schedule. For the WG leads: please send me any requests for WG time or plenary items as soon as you can (even if you need to revise that later), so we can start building a more concrete agenda. Also, please let me know of any tickets that you want read or voted on as part of errata. I have pre-populated the agenda with the second and first votes, since those are fixed – but please also double check those. The deadline for announcing tickets for readings or votes is as usual 2 weeks before the forum, which is the Monday 11/24, I.e., the Monday of Thanksgiving week. Thanks! Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov<mailto:[email protected]>, http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA _______________________________________________ mpi-forum mailing list mpi-forum(a)lists.mpi-forum.org<mailto:[email protected]> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Pavan Balaji ✉️ http://www.mcs.anl.gov/~balaji _______________________________________________ mpiwg-hybridpm mailing list mpiwg-hybridpm(a)lists.mpi-forum.org<mailto:[email protected]> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm _______________________________________________ mpiwg-hybridpm mailing list mpiwg-hybridpm(a)lists.mpi-forum.org<mailto:[email protected]> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm

1 0