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

Re: [mpiwg-hybridpm] Hybrid telecon tomorrow
by Jim Dinan 24 Mar '14

24 Mar '14

Updated copy with notes, decisions and todo items from today's meeting. On Mon, Mar 24, 2014 at 10:36 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > I attached a set of notes for discussion. Once we have agreed on > semantics, we can generate the needed text. > > I'm home with the kiddo today. I'm going to try to call in to the > meeting, but unfortunately can't make any guarantees. > > ~Jim. > > > On Sun, Mar 23, 2014 at 5:42 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > >> >> The plan for tomorrow's hybrid telecon was to discuss the updated text >> for the Endpoints proposal. >> >> Jim: are you ready with this? >> >> -- 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] Hybrid telecon tomorrow
by Balaji, Pavan 24 Mar '14

24 Mar '14

Jim, Can you let us know if you can’t join? It’ll not be very useful to discuss the endpoints proposal without you on the call. — Pavan On Mar 24, 2014, at 9:36 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > I attached a set of notes for discussion. Once we have agreed on semantics, we can generate the needed text. > > I'm home with the kiddo today. I'm going to try to call in to the meeting, but unfortunately can't make any guarantees. > > ~Jim. > > > On Sun, Mar 23, 2014 at 5:42 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > The plan for tomorrow’s hybrid telecon was to discuss the updated text for the Endpoints proposal. > > Jim: are you ready with this? > > — Pavan > > _______________________________________________ > mpiwg-hybridpm mailing list > mpiwg-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm > > <MPI Endpoints Notes.pdf>_______________________________________________ > 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] Hybrid telecon tomorrow
by Jim Dinan 24 Mar '14

24 Mar '14

I attached a set of notes for discussion. Once we have agreed on semantics, we can generate the needed text. I'm home with the kiddo today. I'm going to try to call in to the meeting, but unfortunately can't make any guarantees. ~Jim. On Sun, Mar 23, 2014 at 5:42 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > The plan for tomorrow's hybrid telecon was to discuss the updated text for > the Endpoints proposal. > > Jim: are you ready with this? > > -- Pavan > > _______________________________________________ > mpiwg-hybridpm mailing list > mpiwg-hybridpm(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm >

1 0

Hybrid telecon tomorrow
by Balaji, Pavan 23 Mar '14

23 Mar '14

The plan for tomorrow’s hybrid telecon was to discuss the updated text for the Endpoints proposal. Jim: are you ready with this? — Pavan

1 0

Re: [mpiwg-hybridpm] Hybrid telecon on Monday
by Balaji, Pavan 10 Mar '14

10 Mar '14

Here are the meeting minutes: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/notes-2014-03-… — Pavan On Mar 10, 2014, at 11:02 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > Webex just booted me off and now it says that I already have one meeting in progress. > > — Pavan > > On Mar 6, 2014, at 11:34 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > >> >> We’ll have our regular telecon this coming Monday (03/10; 11am central). >> >> Ticket 357 had a successful formal reading. We don’t need to spend any more time on that. >> >> There were a bunch of comments on 380 and one issue to fix/figure-out for 411. We can those two on the call. >> >> Thanks, >> >> — Pavan >> >> _______________________________________________ >> 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] Hybrid telecon on Monday
by Balaji, Pavan 10 Mar '14

10 Mar '14

Webex just booted me off and now it says that I already have one meeting in progress. — Pavan On Mar 6, 2014, at 11:34 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > We’ll have our regular telecon this coming Monday (03/10; 11am central). > > Ticket 357 had a successful formal reading. We don’t need to spend any more time on that. > > There were a bunch of comments on 380 and one issue to fix/figure-out for 411. We can those two on the call. > > 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

Hybrid telecon on Monday
by Balaji, Pavan 06 Mar '14

06 Mar '14

We’ll have our regular telecon this coming Monday (03/10; 11am central). Ticket 357 had a successful formal reading. We don’t need to spend any more time on that. There were a bunch of comments on 380 and one issue to fix/figure-out for 411. We can those two on the call. Thanks, — Pavan

1 0

Re: [mpiwg-hybridpm] Endpoints
by Balaji, Pavan 06 Mar '14

06 Mar '14

I prefer the last option. — Pavan On Mar 6, 2014, at 11:10 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Eek, ok.. I guess there are a few options. We could restrict what you can do in a callback, but that would break backward compatibility. We could add a caveat that MPI_Comm_free needs to be called concurrently on all ranks if there are any attribute callbacks that might want to use a collective. Worst case, we could remove the concurrency exception for MPI_Comm_free on an endpoints communicator. > > ~Jim. > > > On Thu, Mar 6, 2014 at 10:59 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > There are (currently) no restrictions on what can happen in attribute callback functions. > > See Fab's presentation from several months ago (I don't remember when, exactly, but the meeting was in San Jose/Cisco) about all the problems w/ attributes -- this is one of them (unrestricted access in callback functions). > > > > On Mar 6, 2014, at 7:51 AM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > > > > > I’m pretty sure that’s not correct. You can have large collective ops and other operations in the callbacks. HDF5 and PETSc do all sorts of stuff in these callbacks. > > > > — Pavan > > > > On Mar 6, 2014, at 9:39 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > > > >> I believe that the copy and free attribute callbacks are intended to be local functions that deal with objects the user has attached as attributes, that might need to be copied or freed. Can't find text that confirms this explicitly, though.. > >> > >> > >> On Wed, Mar 5, 2014 at 6:31 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > >> > >> The endpoints proposal says that Comm_free is not fully synchronizing. What happens if I have an attribute callback on that communicator which does a Bcast? How will that work? > >> > >> — Pavan > >> > >> _______________________________________________ > >> 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 > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > > _______________________________________________ > 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] Endpoints
by Jim Dinan 06 Mar '14

06 Mar '14

Eek, ok.. I guess there are a few options. We could restrict what you can do in a callback, but that would break backward compatibility. We could add a caveat that MPI_Comm_free needs to be called concurrently on all ranks if there are any attribute callbacks that might want to use a collective. Worst case, we could remove the concurrency exception for MPI_Comm_free on an endpoints communicator. ~Jim. On Thu, Mar 6, 2014 at 10:59 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com > wrote: > There are (currently) no restrictions on what can happen in attribute > callback functions. > > See Fab's presentation from several months ago (I don't remember when, > exactly, but the meeting was in San Jose/Cisco) about all the problems w/ > attributes -- this is one of them (unrestricted access in callback > functions). > > > > On Mar 6, 2014, at 7:51 AM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > > > > > I'm pretty sure that's not correct. You can have large collective ops > and other operations in the callbacks. HDF5 and PETSc do all sorts of > stuff in these callbacks. > > > > -- Pavan > > > > On Mar 6, 2014, at 9:39 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > > > >> I believe that the copy and free attribute callbacks are intended to be > local functions that deal with objects the user has attached as attributes, > that might need to be copied or freed. Can't find text that confirms this > explicitly, though.. > >> > >> > >> On Wed, Mar 5, 2014 at 6:31 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > >> > >> The endpoints proposal says that Comm_free is not fully synchronizing. > What happens if I have an attribute callback on that communicator which > does a Bcast? How will that work? > >> > >> -- Pavan > >> > >> _______________________________________________ > >> 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 > > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: > http://www.cisco.com/web/about/doing_business/legal/cri/ > > _______________________________________________ > 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] Endpoints
by Jeff Squyres (jsquyres) 06 Mar '14

06 Mar '14

There are (currently) no restrictions on what can happen in attribute callback functions. See Fab's presentation from several months ago (I don't remember when, exactly, but the meeting was in San Jose/Cisco) about all the problems w/ attributes -- this is one of them (unrestricted access in callback functions). On Mar 6, 2014, at 7:51 AM, "Balaji, Pavan" <balaji(a)anl.gov> wrote: > > I’m pretty sure that’s not correct. You can have large collective ops and other operations in the callbacks. HDF5 and PETSc do all sorts of stuff in these callbacks. > > — Pavan > > On Mar 6, 2014, at 9:39 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > >> I believe that the copy and free attribute callbacks are intended to be local functions that deal with objects the user has attached as attributes, that might need to be copied or freed. Can't find text that confirms this explicitly, though.. >> >> >> On Wed, Mar 5, 2014 at 6:31 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: >> >> The endpoints proposal says that Comm_free is not fully synchronizing. What happens if I have an attribute callback on that communicator which does a Bcast? How will that work? >> >> — Pavan >> >> _______________________________________________ >> 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 -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0