- Mpi3-bwcompat - mailman3.cels.anl.gov

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

Solt, David George wrote on Tue, 17 May 2011 at 11:33:12 > I'll setup the meeting for the usual time this Friday. > > Did anyone send a 1-2 sentence explanation of MPI_Count to Rich? Just sent them to Rich. -Fab > > Thanks, > Dave > > -----Original Message----- > From: mpi3-bwcompat-bounces_at_[hidden] [mailto:mpi3- > bwcompat-bounces_at_[hidden]] On Behalf Of Fab Tillier > Sent: Tuesday, May 17, 2011 1:29 PM > To: MPI-3 backwards compatability WG > Subject: Re: [Mpi3-bwcompat] Phone conf schedule? > > Solt, David George wrote on Tue, 17 May 2011 at 11:24:37 > >> One other cleanup point. Rich requested: >> >>> One more thing. Those of you that have an item on the list (and those >>> that are adding items) - please send me a 1-2 sentence explanation that >>> we can use as a >> description." >> >> Has anyone sent him anything yet? > > I sent him my 3 new items. I don't know if Jeff sent him the timer > stuff or not. > >> I am open to working on other miscellaneous or backward compatibility >> related issues. I could schedule a one-time meeting to discuss our future >> and then decide if we should continue with a re-occurring meeting? > > Sure, that sounds good. Same time as usual, this Friday? > > -Fab > >> >> Thanks, >> Dave >> >> -----Original Message----- >> From: mpi3-bwcompat-bounces_at_[hidden] [mailto:mpi3- >> bwcompat-bounces_at_[hidden]] On Behalf Of Fab Tillier >> Sent: Tuesday, May 17, 2011 1:17 PM >> To: MPI-3 backwards compatability WG >> Subject: Re: [Mpi3-bwcompat] Phone conf schedule? >> >> Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15 >> >>> On May 17, 2011, at 1:01 PM, Fab Tillier wrote: >>> >>>> The meeting request I had expired 4/15/2011. Do we want to resume it? >>>> >>>> Couple topics of interest, actually on topic for our working group: - >>>> Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI >>>> 3.0. They are still available via MPI 2.2 support, and an >>>> implementation can easily provide concurrent support for MPI 2.2 and >>>> MPI 3.0, should that be required. - Defining the C++ bindings as >>>> optional in MPI 3.0. This allows implementations that do not ship >>>> C++ bindings to be standard compliant. - Removing C++ bindings from >>>> MPI 3.0. This one goes counter to the previous one, but follows the >>>> same logic as for the other deprecated functions. If we don't >>>> provide C++ bindings for new MPI 3.0 functionality, we should remove >>>> them from MPI 3.0. They can still be supported via MPI 2.2 >>>> compliance. >>> >>> Anything to discuss about const buffers? >> >> We passed the first reading, so we're going for 1st vote at the next >> meeting. There was one minor ticket0 wording change for the IN >> description, but nothing major. I don't think there's much to discuss >> at this point. >> >>> I've got someone working on extensions to the datatype construction >>> routines, creating a use case/justification for why we should add >>> those routines to 3.0. Adam Moody was going to create a ticket for >>> those, once I give him this use case document. [Dunno if we need to >>> talk about anything yet on this though] >> >> That would be nice to see. Jeff and I are also working on the MPI >> Timer request stuff, don't know if we want to discuss that in this >> working group, I'd be up for it, I like this crowd... >> >> -Fab >> >> _______________________________________________ Mpi3- bwcompat mailing >> list Mpi3-bwcompat_at_[hidden] >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat >> >> _______________________________________________ >> Mpi3-bwcompat mailing list >> Mpi3-bwcompat_at_[hidden] >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat > > _______________________________________________ Mpi3-bwcompat mailing > list Mpi3-bwcompat_at_[hidden] > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat > > _______________________________________________ > Mpi3-bwcompat mailing list > Mpi3-bwcompat_at_[hidden] > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat

1 0

Re: [Mpi3-bwcompat] [Mpi-forum] MPI 3.0
by unknown＠example.com 17 May '11

17 May '11

Hi Rich, Here's the description for the two items we have on the list: Const: make C bindings const-correct, except where backward compatibility would be broken. MPI_Count: Minimal support for large datatypes. -Fab From: mpi-forum-bounces_at_[hidden] [mailto:mpi-forum-bounces_at_[hidden]] On Behalf Of RICHARD GRAHAM Sent: Wednesday, May 11, 2011 8:30 PM To: mpi-forum_at_[hidden] Subject: Re: [Mpi-forum] MPI 3.0 One more thing. Those of you that have an item on the list (and those that are adding items) - please send me a 1-2 sentence explanation that we can use as a description. Thanks, Rich ________________________________ From: rlgraham32_at_[hidden] To: mpi-forum_at_[hidden] Date: Thu, 12 May 2011 01:46:58 +0100 Subject: [Mpi-forum] MPI 3.0 At the MPI forum meeting this week we decided to accelerate the release of MPI-3.0, and have a follow on MPI-3.1 when the more complete set of features is available. Currently, nonblocking collectives and MProbe have already been voted in. We have created a list of features that may be ready for first reading at the next couple of meetings, which will be used to determine the gating factors for the MPI-3.0 release. Here is the list we generated: - RMA - FT stabilization - Fortran bindings - MPI-T - Neighborhood collectives - Const - MPI Count - Nonblocking collective I/O - Hybrid: Shared memory communicator - Hybrid: Threads - Hybrid: Endpoints If people want to add more to this list, please let me know by end of the day Friday, 5/13/2011. Once we have the complete list, each eligible organization should vote for it's top three features, with respect to MPI-3.0. An organization is eligible to vote, if someone from that organization has participated in person at one or more meetings in 2011. The vote will take place 5/14-20/2011, and the top three items will be what gate the release of MPI-3.0. To vote, a representative of the organization should e-mail BOTH Jeff Squyres and myself with your vote. Other items that are voted in before 3.0 is final, will also be included. For MPI-3.1, we will change our process a bit, but that is still under discussion, and will be the subject of a future e-mail discussion. The eligible organizations include: - ANL - Bull - Cicso - Cray - Fujistsu - HDF - HLRS - HP - IBM - Indiana U. - INRIA - Intel - LLNL - LANL - Microsoft - NCSA/UIUC - ORNL - OSU - Oracle - SNL - U Alabama Birmingham - U Tennessee - U. Tokyo Rich _______________________________________________ mpi-forum mailing list mpi-forum_at_[hidden]<mailto:mpi-forum_at_[hidden]> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum *

1 0

MPI Backward Compatibility Working Group
by unknown＠example.com 17 May '11

17 May '11

When: Friday, May 20, 2011 12:00 PM-1:00 PM (GMT-06:00) Central Time (US & Canada). Where: 866.409.2889 id 3582114059 Note: The GMT offset above does not reflect daylight saving time adjustments. *~*~*~*~*~*~*~*~*~* U.S. toll number 702.696.4520 U.S. toll-free dial-in number International â see attached pdf 866.409.2889 - Conference Code: 3582114059 *

1 0

MPI Backward Compatibility Working Group
by unknown＠example.com 17 May '11

17 May '11

When: Occurs every 2 weeks on Friday effective 7/23/2010 until 5/21/2011 from 12:00 PM to 1:00 PM (GMT-06:00) Central Time (US & Canada). Where: 866.409.2889 id 3582114059 Note: The GMT offset above does not reflect daylight saving time adjustments. *~*~*~*~*~*~*~*~*~* U.S. toll number 702.696.4520 U.S. toll-free dial-in number International â see attached pdf 866.409.2889 - Conference Code: 3582114059 *

1 0

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

I'll setup the meeting for the usual time this Friday. Did anyone send a 1-2 sentence explanation of MPI_Count to Rich? Thanks, Dave -----Original Message----- From: mpi3-bwcompat-bounces_at_[hidden] [mailto:mpi3-bwcompat-bounces_at_[hidden]] On Behalf Of Fab Tillier Sent: Tuesday, May 17, 2011 1:29 PM To: MPI-3 backwards compatability WG Subject: Re: [Mpi3-bwcompat] Phone conf schedule? Solt, David George wrote on Tue, 17 May 2011 at 11:24:37 > One other cleanup point. Rich requested: > >> One more thing. Those of you that have an item on the list (and those >> that are adding items) - please send me a 1-2 sentence explanation that >> we can use as a > description." > > Has anyone sent him anything yet? I sent him my 3 new items. I don't know if Jeff sent him the timer stuff or not. > I am open to working on other miscellaneous or backward compatibility > related issues. I could schedule a one-time meeting to discuss our future > and then decide if we should continue with a re-occurring meeting? Sure, that sounds good. Same time as usual, this Friday? -Fab > > Thanks, > Dave > > -----Original Message----- > From: mpi3-bwcompat-bounces_at_[hidden] [mailto:mpi3- > bwcompat-bounces_at_[hidden]] On Behalf Of Fab Tillier > Sent: Tuesday, May 17, 2011 1:17 PM > To: MPI-3 backwards compatability WG > Subject: Re: [Mpi3-bwcompat] Phone conf schedule? > > Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15 > >> On May 17, 2011, at 1:01 PM, Fab Tillier wrote: >> >>> The meeting request I had expired 4/15/2011. Do we want to resume it? >>> >>> Couple topics of interest, actually on topic for our working group: - >>> Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI >>> 3.0. They are still available via MPI 2.2 support, and an >>> implementation can easily provide concurrent support for MPI 2.2 and >>> MPI 3.0, should that be required. - Defining the C++ bindings as >>> optional in MPI 3.0. This allows implementations that do not ship C++ >>> bindings to be standard compliant. - Removing C++ bindings from MPI >>> 3.0. This one goes counter to the previous one, but follows the same >>> logic as for the other deprecated functions. If we don't provide C++ >>> bindings for new MPI 3.0 functionality, we should remove them from MPI >>> 3.0. They can still be supported via MPI 2.2 compliance. >> >> Anything to discuss about const buffers? > > We passed the first reading, so we're going for 1st vote at the next meeting. > There was one minor ticket0 wording change for the IN description, but > nothing major. I don't think there's much to discuss at this point. > >> I've got someone working on extensions to the datatype construction >> routines, creating a use case/justification for why we should add those >> routines to 3.0. Adam Moody was going to create a ticket for those, >> once I give him this use case document. [Dunno if we need to talk >> about anything yet on this though] > > That would be nice to see. Jeff and I are also working on the MPI Timer > request stuff, don't know if we want to discuss that in this working group, I'd > be up for it, I like this crowd... > > -Fab > > _______________________________________________ Mpi3-bwcompat mailing > list Mpi3-bwcompat_at_[hidden] > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat > > _______________________________________________ > Mpi3-bwcompat mailing list > Mpi3-bwcompat_at_[hidden] > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat _______________________________________________ Mpi3-bwcompat mailing list Mpi3-bwcompat_at_[hidden] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat

1 0

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

Solt, David George wrote on Tue, 17 May 2011 at 11:24:37 > One other cleanup point. Rich requested: > >> One more thing. Those of you that have an item on the list (and those >> that are adding items) - please send me a 1-2 sentence explanation that >> we can use as a > description." > > Has anyone sent him anything yet? I sent him my 3 new items. I don't know if Jeff sent him the timer stuff or not. > I am open to working on other miscellaneous or backward compatibility > related issues. I could schedule a one-time meeting to discuss our future > and then decide if we should continue with a re-occurring meeting? Sure, that sounds good. Same time as usual, this Friday? -Fab > > Thanks, > Dave > > -----Original Message----- > From: mpi3-bwcompat-bounces_at_[hidden] [mailto:mpi3- > bwcompat-bounces_at_[hidden]] On Behalf Of Fab Tillier > Sent: Tuesday, May 17, 2011 1:17 PM > To: MPI-3 backwards compatability WG > Subject: Re: [Mpi3-bwcompat] Phone conf schedule? > > Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15 > >> On May 17, 2011, at 1:01 PM, Fab Tillier wrote: >> >>> The meeting request I had expired 4/15/2011. Do we want to resume it? >>> >>> Couple topics of interest, actually on topic for our working group: - >>> Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI >>> 3.0. They are still available via MPI 2.2 support, and an >>> implementation can easily provide concurrent support for MPI 2.2 and >>> MPI 3.0, should that be required. - Defining the C++ bindings as >>> optional in MPI 3.0. This allows implementations that do not ship C++ >>> bindings to be standard compliant. - Removing C++ bindings from MPI >>> 3.0. This one goes counter to the previous one, but follows the same >>> logic as for the other deprecated functions. If we don't provide C++ >>> bindings for new MPI 3.0 functionality, we should remove them from MPI >>> 3.0. They can still be supported via MPI 2.2 compliance. >> >> Anything to discuss about const buffers? > > We passed the first reading, so we're going for 1st vote at the next meeting. > There was one minor ticket0 wording change for the IN description, but > nothing major. I don't think there's much to discuss at this point. > >> I've got someone working on extensions to the datatype construction >> routines, creating a use case/justification for why we should add those >> routines to 3.0. Adam Moody was going to create a ticket for those, >> once I give him this use case document. [Dunno if we need to talk >> about anything yet on this though] > > That would be nice to see. Jeff and I are also working on the MPI Timer > request stuff, don't know if we want to discuss that in this working group, I'd > be up for it, I like this crowd... > > -Fab > > _______________________________________________ Mpi3-bwcompat mailing > list Mpi3-bwcompat_at_[hidden] > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat > > _______________________________________________ > Mpi3-bwcompat mailing list > Mpi3-bwcompat_at_[hidden] > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat

1 0

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

One other cleanup point. Rich requested: > One more thing. Those of you that have an item on the list (and those that are adding > items) - please send me a 1-2 sentence explanation that we can use as a description." Has anyone sent him anything yet? I am open to working on other miscellaneous or backward compatibility related issues. I could schedule a one-time meeting to discuss our future and then decide if we should continue with a re-occurring meeting? Thanks, Dave -----Original Message----- From: mpi3-bwcompat-bounces_at_[hidden] [mailto:mpi3-bwcompat-bounces_at_[hidden]] On Behalf Of Fab Tillier Sent: Tuesday, May 17, 2011 1:17 PM To: MPI-3 backwards compatability WG Subject: Re: [Mpi3-bwcompat] Phone conf schedule? Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15 > On May 17, 2011, at 1:01 PM, Fab Tillier wrote: > >> The meeting request I had expired 4/15/2011. Do we want to resume it? >> >> Couple topics of interest, actually on topic for our working group: >> - Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI >> 3.0. They are still available via MPI 2.2 support, and an implementation can >> easily provide concurrent support for MPI 2.2 and MPI 3.0, should that be >> required. >> - Defining the C++ bindings as optional in MPI 3.0. This allows >> implementations that do not ship C++ bindings to be standard compliant. >> - Removing C++ bindings from MPI 3.0. This one goes counter to the >> previous one, but follows the same logic as for the other deprecated >> functions. If we don't provide C++ bindings for new MPI 3.0 >> functionality, we should remove them from MPI 3.0. They can still be >> supported via MPI 2.2 compliance. > > Anything to discuss about const buffers? We passed the first reading, so we're going for 1st vote at the next meeting. There was one minor ticket0 wording change for the IN description, but nothing major. I don't think there's much to discuss at this point. > I've got someone working on extensions to the datatype > construction routines, creating a use case/justification for why we should add > those routines to 3.0. Adam Moody was going to create a ticket for those, > once I give him this use case document. [Dunno if we need to talk about > anything yet on this though] That would be nice to see. Jeff and I are also working on the MPI Timer request stuff, don't know if we want to discuss that in this working group, I'd be up for it, I like this crowd... -Fab _______________________________________________ Mpi3-bwcompat mailing list Mpi3-bwcompat_at_[hidden] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-bwcompat

1 0

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

Hi Fab, On May 17, 2011, at 1:16 PM, Fab Tillier wrote: > Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15 > >> On May 17, 2011, at 1:01 PM, Fab Tillier wrote: >> >>> The meeting request I had expired 4/15/2011. Do we want to resume it? >>> >>> Couple topics of interest, actually on topic for our working group: >>> - Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI >>> 3.0. They are still available via MPI 2.2 support, and an implementation can >>> easily provide concurrent support for MPI 2.2 and MPI 3.0, should that be >>> required. >>> - Defining the C++ bindings as optional in MPI 3.0. This allows >>> implementations that do not ship C++ bindings to be standard compliant. >>> - Removing C++ bindings from MPI 3.0. This one goes counter to the >>> previous one, but follows the same logic as for the other deprecated >>> functions. If we don't provide C++ bindings for new MPI 3.0 >>> functionality, we should remove them from MPI 3.0. They can still be >>> supported via MPI 2.2 compliance. >> >> Anything to discuss about const buffers? > > We passed the first reading, so we're going for 1st vote at the next meeting. There was one minor ticket0 wording change for the IN description, but nothing major. I don't think there's much to discuss at this point. > >> I've got someone working on extensions to the datatype >> construction routines, creating a use case/justification for why we should add >> those routines to 3.0. Adam Moody was going to create a ticket for those, >> once I give him this use case document. [Dunno if we need to talk about >> anything yet on this though] > > That would be nice to see. Jeff and I are also working on the MPI Timer request stuff, don't know if we want to discuss that in this working group, I'd be up for it, I like this crowd... Awww, we like you too, Fab. :-) :-) I've got two items I'd like to bounce around with folks: - How best to return an exit code from an MPI application? - What about adding a "quiet" flag to mpiexec to request that mpiexec (not the application) doesn't make any output? Quincey

1 0

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

Quincey Koziol wrote on Tue, 17 May 2011 at 11:09:15 > On May 17, 2011, at 1:01 PM, Fab Tillier wrote: > >> The meeting request I had expired 4/15/2011. Do we want to resume it? >> >> Couple topics of interest, actually on topic for our working group: >> - Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI >> 3.0. They are still available via MPI 2.2 support, and an implementation can >> easily provide concurrent support for MPI 2.2 and MPI 3.0, should that be >> required. >> - Defining the C++ bindings as optional in MPI 3.0. This allows >> implementations that do not ship C++ bindings to be standard compliant. >> - Removing C++ bindings from MPI 3.0. This one goes counter to the >> previous one, but follows the same logic as for the other deprecated >> functions. If we don't provide C++ bindings for new MPI 3.0 >> functionality, we should remove them from MPI 3.0. They can still be >> supported via MPI 2.2 compliance. > > Anything to discuss about const buffers? We passed the first reading, so we're going for 1st vote at the next meeting. There was one minor ticket0 wording change for the IN description, but nothing major. I don't think there's much to discuss at this point. > I've got someone working on extensions to the datatype > construction routines, creating a use case/justification for why we should add > those routines to 3.0. Adam Moody was going to create a ticket for those, > once I give him this use case document. [Dunno if we need to talk about > anything yet on this though] That would be nice to see. Jeff and I are also working on the MPI Timer request stuff, don't know if we want to discuss that in this working group, I'd be up for it, I like this crowd... -Fab

1 0

Re: [Mpi3-bwcompat] Phone conf schedule?
by unknown＠example.com 17 May '11

17 May '11

On May 17, 2011, at 1:01 PM, Fab Tillier wrote: > The meeting request I had expired 4/15/2011. Do we want to resume it? > > Couple topics of interest, actually on topic for our working group: > - Removing deprecated functions (e.g. MPI_UB, MPI_LB, etc.) from MPI 3.0. They are still available via MPI 2.2 support, and an implementation can easily provide concurrent support for MPI 2.2 and MPI 3.0, should that be required. > - Defining the C++ bindings as optional in MPI 3.0. This allows implementations that do not ship C++ bindings to be standard compliant. > - Removing C++ bindings from MPI 3.0. This one goes counter to the previous one, but follows the same logic as for the other deprecated functions. If we don't provide C++ bindings for new MPI 3.0 functionality, we should remove them from MPI 3.0. They can still be supported via MPI 2.2 compliance. Anything to discuss about const buffers? I've got someone working on extensions to the datatype construction routines, creating a use case/justification for why we should add those routines to 3.0. Adam Moody was going to create a ticket for those, once I give him this use case document. [Dunno if we need to talk about anything yet on this though] Quincey

1 0