July 2013 - mpiwg-hybridpm - mailman3.cels.anl.gov

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Jeff Squyres (jsquyres) 30 Jul '13

30 Jul '13

On Jul 29, 2013, at 9:31 PM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > I prefer not changing the semantics of something that already exists. In the current semantics, once INITIALIZED/FINALIZED are set, they can never change their values. The more I think about this, the less I think that an MPI function indicating whether or not MPI is in an epoch is necessary: 1. One of the main ideas behind this proposal is that if you don't know if you can call MPI functions, just call MPI_INIT. 2. However, there are undoubtedly situations where you don't want to call MPI_INIT (e.g., where you need a local-only operation). But in this case, why wouldn't you use some local variable to know whether MPI has been initialized or not? Such a local variable could be protected with app-specific locking and therefore not be racy, like INITIALIZED/FINALIZED/IN_EPOCH. 3. To clarify #2, the original argument for INITIALIZED/FINALIZED was that library A in a process might not know that library B already called MPI_INIT/MPI_FINALIZE. But now every library will always be able to know this, because every library should call MPI_INIT/MPI_FINALIZE themselves. 4. Indeed, an MPI_IN_EPOCH type of function might accidentally lead to erroneous code. Consider something like this: T=0 Library A and library B (in different threads) call MPI_INIT T=N Library B shuts down all local MPI state and calls MPI_FINALIZE *** Note that library A has _not_ called MPI_FINALIZE T=N+1 Async thread in library B calls IN_EPOCH to see if MPI is still initialized Since IN_EPOCH will return true, the thread in library B may erroneously try to do something with the MPI state in library B that has already been torn down. My point (which mirrors #2, above): *it's not enough to know whether MPI is in an epoch or not;* each library will also need to know whether its internal MPI state (e.g., local communicators, ...etc.) is valid or not. *** So perhaps we should just deprecate INITIALIZED/FINALIZED and not replace them with anything. -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Re: [Mpi3-hybridpm] 7/29/13 meeting notes
by Antonio J. Peña 29 Jul '13

29 Jul '13

Sorry, the correct link is: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-07-… Thanks to Jeff Hammond for reporting. On Monday, July 29, 2013 07:18:34 PM Antonio J. Peña wrote: Hi all, The notes of today's meeting are in the following link: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-29-… [2] -- Antonio J. Peña Postdoctoral Appointee Mathematics and Computer Science Division Argonne National Laboratory 9700 South Cass Avenue, Bldg. 240, Of. 3148 Argonne, IL 60439-4847 (+1) 630-252-7928 apenya(a)mcs.anl.gov www.mcs.anl.gov/~apenya -- Antonio J. Peña Postdoctoral Appointee Mathematics and Computer Science Division Argonne National Laboratory 9700 South Cass Avenue, Bldg. 240, Of. 3148 Argonne, IL 60439-4847 (+1) 630-252-7928 apenya(a)mcs.anl.gov www.mcs.anl.gov/~apenya -------- [1] https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-29-… [2] https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-07-…

1 0

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Pavan Balaji 29 Jul '13

29 Jul '13

On 07/29/2013 04:19 PM, Jeff Squyres (jsquyres) wrote: > Yes, agreed -- I just put that box there because it's an obvious > thing we'll all see before we send this to the Forum. It can be > discussed on the call in 2 weeks (but I won't be there), or on the > call in 4 weeks (I'll be on that one). > > My $0.02: because of this proposal, INITIALIZED/FINALIZED *have* to > change: > > 1. they either have to reflect the epoch, like I proposed 2. or they > have to be specifically scoped to only apply to the *first* MPI > initialization/finalization (which seems really weird) > > I'm guessing you're advocating #2, right? If so, can you explain > why? I prefer not changing the semantics of something that already exists. In the current semantics, once INITIALIZED/FINALIZED are set, they can never change their values. -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

[Mpi3-hybridpm] 7/29/13 meeting notes
by Antonio J. Peña 29 Jul '13

29 Jul '13

Hi all, The notes of today's meeting are in the following link: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-29-… [2] -- Antonio J. Peña Postdoctoral Appointee Mathematics and Computer Science Division Argonne National Laboratory 9700 South Cass Avenue, Bldg. 240, Of. 3148 Argonne, IL 60439-4847 (+1) 630-252-7928 apenya(a)mcs.anl.gov www.mcs.anl.gov/~apenya -------- [1] https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-29-… [2] https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2013-07-…

1 0

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Jeff Squyres (jsquyres) 29 Jul '13

29 Jul '13

Yes, agreed -- I just put that box there because it's an obvious thing we'll all see before we send this to the Forum. It can be discussed on the call in 2 weeks (but I won't be there), or on the call in 4 weeks (I'll be on that one). My $0.02: because of this proposal, INITIALIZED/FINALIZED *have* to change: 1. they either have to reflect the epoch, like I proposed 2. or they have to be specifically scoped to only apply to the *first* MPI initialization/finalization (which seems really weird) I'm guessing you're advocating #2, right? If so, can you explain why? On Jul 29, 2013, at 3:10 PM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > Jeff, > > We need to decide on INITIALIZED/FINALIZED vs. IS_WITHIN_EPOCH in the working group and take the final decision to the Forum, instead of wasting the Forum's time on that. > > -- Pavan > > On 07/29/2013 11:43 AM, Jeff Squyres (jsquyres) wrote: >> I made a few more edits while Brian/Jim/Keith/Pavan are still arguing. :-) >> >> Here's a new copy... >> >> >> >> >> >> On Jul 29, 2013, at 12:32 PM, Jeff Squyres <jsquyres(a)cisco.com> >> wrote: >> >>> Here's a new copy of the slides; if we have time today, we can discuss. >>> >>> I took all feedback, except I disagreed with Steve's #5 and #8. >>> >>> >>> >>> On Jul 17, 2013, at 9:22 AM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: >>> >>> > Jeff, >>> > >>> > Some comments -- >>> > >>> > 1. Slide 2: The main goals need to be rephrased. The goals are to overcome limitations while using MPI in stacked and threaded libraries. For stacked libraries, the single INIT/FINALIZE limitation requires each library to have global knowledge on who else is using MPI. For threaded libraries, there are race conditions. >>> > >>> > 2. Slide 5: Sorry, I missed this earlier. I'm not in favor of changing the current definitions of IS_INITIALIZED and IS_FINALIZED. Better to deprecate them and add MPI_IS_WITHIN_EPOCH. >>> > >>> > 3. Slide 10: I don't think we agreed that the requested level will be ignored by subsequent calls. That's a quality of implementation issue. >>> > >>> > 4. Slide 15: We can say that our current stand is to remove that restriction that the "main thread" needs to finalize. >>> > >>> > Btw, I've requested for plenary time. >>> > >>> > -- Pavan >>> > >>> > On 07/15/2013 12:45 PM, Jeff Squyres (jsquyres) wrote: >>> >> Here's a better cut of slides that are more Forum-friendly. >>> >> >>> >> Comments welcome. >>> >> >>> >> (Pavan: I'm correct in thinking that you're getting plenary time from Martin for this, right?) >>> >> >>> >> >>> >> >>> >> _______________________________________________ >>> >> Mpi3-hybridpm mailing list >>> >> Mpi3-hybridpm(a)lists.mpi-forum.org >>> >>http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm >>> >> >>> > >>> > -- >>> > Pavan Balaji >>> >http://www.mcs.anl.gov/~balaji >>> >>> >>> -- >>> Jeff Squyres >>> jsquyres(a)cisco.com >>> For corporate legal information go to:http://www.cisco.com/web/about/doing_business/legal/cri/ >>> >>> <init-finalize-issues-for-forum.pptx> >> >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to: >> http://www.cisco.com/web/about/doing_business/legal/cri/ >> > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Pavan Balaji 29 Jul '13

29 Jul '13

Jeff, We need to decide on INITIALIZED/FINALIZED vs. IS_WITHIN_EPOCH in the working group and take the final decision to the Forum, instead of wasting the Forum's time on that. -- Pavan On 07/29/2013 11:43 AM, Jeff Squyres (jsquyres) wrote: > I made a few more edits while Brian/Jim/Keith/Pavan are still arguing. :-) > > Here's a new copy... > > > > > > On Jul 29, 2013, at 12:32 PM, Jeff Squyres <jsquyres(a)cisco.com> > wrote: > >> Here's a new copy of the slides; if we have time today, we can discuss. >> >> I took all feedback, except I disagreed with Steve's #5 and #8. >> >> >> >> On Jul 17, 2013, at 9:22 AM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: >> >> > Jeff, >> > >> > Some comments -- >> > >> > 1. Slide 2: The main goals need to be rephrased. The goals are to overcome limitations while using MPI in stacked and threaded libraries. For stacked libraries, the single INIT/FINALIZE limitation requires each library to have global knowledge on who else is using MPI. For threaded libraries, there are race conditions. >> > >> > 2. Slide 5: Sorry, I missed this earlier. I'm not in favor of changing the current definitions of IS_INITIALIZED and IS_FINALIZED. Better to deprecate them and add MPI_IS_WITHIN_EPOCH. >> > >> > 3. Slide 10: I don't think we agreed that the requested level will be ignored by subsequent calls. That's a quality of implementation issue. >> > >> > 4. Slide 15: We can say that our current stand is to remove that restriction that the "main thread" needs to finalize. >> > >> > Btw, I've requested for plenary time. >> > >> > -- Pavan >> > >> > On 07/15/2013 12:45 PM, Jeff Squyres (jsquyres) wrote: >> >> Here's a better cut of slides that are more Forum-friendly. >> >> >> >> Comments welcome. >> >> >> >> (Pavan: I'm correct in thinking that you're getting plenary time from Martin for this, right?) >> >> >> >> >> >> >> >> _______________________________________________ >> >> Mpi3-hybridpm mailing list >> >> Mpi3-hybridpm(a)lists.mpi-forum.org >> >>http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm >> >> >> > >> > -- >> > Pavan Balaji >> >http://www.mcs.anl.gov/~balaji >> >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to:http://www.cisco.com/web/about/doing_business/legal/cri/ >> >> <init-finalize-issues-for-forum.pptx> > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: > http://www.cisco.com/web/about/doing_business/legal/cri/ > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Jeff Squyres (jsquyres) 29 Jul '13

29 Jul '13

I made a few more edits while Brian/Jim/Keith/Pavan are still arguing. :-) Here's a new copy... On Jul 29, 2013, at 12:32 PM, Jeff Squyres <jsquyres(a)cisco.com> wrote: > Here's a new copy of the slides; if we have time today, we can discuss. > > I took all feedback, except I disagreed with Steve's #5 and #8. > > > > On Jul 17, 2013, at 9:22 AM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > > > Jeff, > > > > Some comments -- > > > > 1. Slide 2: The main goals need to be rephrased. The goals are to overcome limitations while using MPI in stacked and threaded libraries. For stacked libraries, the single INIT/FINALIZE limitation requires each library to have global knowledge on who else is using MPI. For threaded libraries, there are race conditions. > > > > 2. Slide 5: Sorry, I missed this earlier. I'm not in favor of changing the current definitions of IS_INITIALIZED and IS_FINALIZED. Better to deprecate them and add MPI_IS_WITHIN_EPOCH. > > > > 3. Slide 10: I don't think we agreed that the requested level will be ignored by subsequent calls. That's a quality of implementation issue. > > > > 4. Slide 15: We can say that our current stand is to remove that restriction that the "main thread" needs to finalize. > > > > Btw, I've requested for plenary time. > > > > -- Pavan > > > > On 07/15/2013 12:45 PM, Jeff Squyres (jsquyres) wrote: > >> Here's a better cut of slides that are more Forum-friendly. > >> > >> Comments welcome. > >> > >> (Pavan: I'm correct in thinking that you're getting plenary time from Martin for this, right?) > >> > >> > >> > >> _______________________________________________ > >> Mpi3-hybridpm mailing list > >> Mpi3-hybridpm(a)lists.mpi-forum.org > >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm > >> > > > > -- > > Pavan Balaji > > http://www.mcs.anl.gov/~balaji > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > > <init-finalize-issues-for-forum.pptx> -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Jeff Squyres (jsquyres) 29 Jul '13

29 Jul '13

Here's a new copy of the slides; if we have time today, we can discuss. I took all feedback, except I disagreed with Steve's #5 and #8. On Jul 17, 2013, at 9:22 AM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > Jeff, > > Some comments -- > > 1. Slide 2: The main goals need to be rephrased. The goals are to overcome limitations while using MPI in stacked and threaded libraries. For stacked libraries, the single INIT/FINALIZE limitation requires each library to have global knowledge on who else is using MPI. For threaded libraries, there are race conditions. > > 2. Slide 5: Sorry, I missed this earlier. I'm not in favor of changing the current definitions of IS_INITIALIZED and IS_FINALIZED. Better to deprecate them and add MPI_IS_WITHIN_EPOCH. > > 3. Slide 10: I don't think we agreed that the requested level will be ignored by subsequent calls. That's a quality of implementation issue. > > 4. Slide 15: We can say that our current stand is to remove that restriction that the "main thread" needs to finalize. > > Btw, I've requested for plenary time. > > -- Pavan > > On 07/15/2013 12:45 PM, Jeff Squyres (jsquyres) wrote: >> Here's a better cut of slides that are more Forum-friendly. >> >> Comments welcome. >> >> (Pavan: I'm correct in thinking that you're getting plenary time from Martin for this, right?) >> >> >> >> _______________________________________________ >> Mpi3-hybridpm mailing list >> Mpi3-hybridpm(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm >> > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

[Mpi3-hybridpm] WG Telecon Mon. 7/29 @ 11CT
by Jim Dinan 28 Jul '13

28 Jul '13

Hi All, Reminder that we'll have the hybrid WG telecon on Monday at 11am central. Toll-free dial-in number: 1-866-398-2885 Toll dial-in number: 1-719-457-6209 Participant Passcode: 3141592 Agenda: * Revisit endpoints draft. Revised draft: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/380 * Init/Finalize proposal ~Jim.

1 0

Re: [Mpi3-hybridpm] Better Forum INIT/FINALIZE slides
by Pavan Balaji 17 Jul '13

17 Jul '13

Jeff, Some comments -- 1. Slide 2: The main goals need to be rephrased. The goals are to overcome limitations while using MPI in stacked and threaded libraries. For stacked libraries, the single INIT/FINALIZE limitation requires each library to have global knowledge on who else is using MPI. For threaded libraries, there are race conditions. 2. Slide 5: Sorry, I missed this earlier. I'm not in favor of changing the current definitions of IS_INITIALIZED and IS_FINALIZED. Better to deprecate them and add MPI_IS_WITHIN_EPOCH. 3. Slide 10: I don't think we agreed that the requested level will be ignored by subsequent calls. That's a quality of implementation issue. 4. Slide 15: We can say that our current stand is to remove that restriction that the "main thread" needs to finalize. Btw, I've requested for plenary time. -- Pavan On 07/15/2013 12:45 PM, Jeff Squyres (jsquyres) wrote: > Here's a better cut of slides that are more Forum-friendly. > > Comments welcome. > > (Pavan: I'm correct in thinking that you're getting plenary time from Martin for this, right?) > > > > _______________________________________________ > Mpi3-hybridpm mailing list > Mpi3-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm > -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0