May 2011 - mpi-forum - mailman3.cels.anl.gov

Re: [Mpi-forum] MPI 3.0 voting
by Rolf Rabenseifner 17 May '11

17 May '11

I made a short analysis about the interest shown in the answers in the questionnaire the MPI Forum started at SC09 although this questionnaire cannot be representative. Attached. ----- Original Message ----- > From: "RICHARD GRAHAM" <rlgraham32(a)msn.com> > To: mpi-forum(a)lists.mpi-forum.org > Sent: Tuesday, May 17, 2011 7:35:58 AM > Subject: [Mpi-forum] MPI 3.0 voting > Attached is the list of proposals that are under consideration for key > features that will gate the MPI-3.0 release. Each organization should > vote for 3 items on this list, by sending a note to Jeff Squyres and > myself, by end of the day Sunday, 5/22/2011. The top three items will > be the gating items for release 3.0. > > Rich > > - RMA - Significant enhancements to MPI's one-sided communication to > incorporate requests from multiple user communities and > codification > of established practice > > - FT stabilization - A proposal that defines semantics for MPI to > continue and function upon failure, due to process failure. > > - Fortran bindings - improved Fortran bindings, taking into account > Fortran 2003 and 2008 features. > > - MPI-T - The MPI_T interface, proposed in ticket #266, provides a > mechanism for > the MPI implementation to expose a set of variables, each of which > represent a > particular property, setting, or performance measurement from within > the MPI > implementation. Tools can use this information to help users > understand and > optimize the configuration of the MPI library and/or the performance > of their > codes. > > - Neighborhood collectives - scalable sparse collective support > > - Const - Make the MPI binding const correct > > - MPI Count - Add the mpi count type > > - Nonblocking collective I/O - extend nonblocking collective support > to > include MPI-I/O > > - Hybrid: Shared memory communicator, Threads, and Endpoint proposals > > - Persistence - expand the support for persistent communications, for > optimal > performance. > > > The eligible organizations include: > - ANL > - Bull > - Cisco > - Cray > - Fujitsu > - HDF > - HLRS > - HP > - IBM > - Indiana U. > - INRIA > - Intel > - LLNL > - LANL > - Microsoft > - NCSA/UIUC > - ORNL > - OSU > - Oracle > - SNL > - U Alabama Birmingham > - U Tennessee > - U. Tokyo > > > _______________________________________________ > 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: Allmandring 30)

1 0

MPI 3.0 voting
by RICHARD GRAHAM 17 May '11

17 May '11

Attached is the list of proposals that are under consideration for key features that will gate the MPI-3.0 release. Each organization should vote for 3 items on this list, by sending a note to Jeff Squyres and myself, by end of the day Sunday, 5/22/2011. The top three items will be the gating items for release 3.0. Rich - RMA - Significant enhancements to MPI's one-sided communication to incorporate requests from multiple user communities and codification of established practice - FT stabilization - A proposal that defines semantics for MPI to continue and function upon failure, due to process failure. - Fortran bindings - improved Fortran bindings, taking into account Fortran 2003 and 2008 features. - MPI-T - The MPI_T interface, proposed in ticket #266, provides a mechanism for the MPI implementation to expose a set of variables, each of which represent a particular property, setting, or performance measurement from within the MPI implementation. Tools can use this information to help users understand and optimize the configuration of the MPI library and/or the performance of their codes. - Neighborhood collectives - scalable sparse collective support - Const - Make the MPI binding const correct - MPI Count - Add the mpi count type - Nonblocking collective I/O - extend nonblocking collective support to include MPI-I/O - Hybrid: Shared memory communicator, Threads, and Endpoint proposals - Persistence - expand the support for persistent communications, for optimal performance. The eligible organizations include: - ANL - Bull - Cisco - Cray - Fujitsu - HDF - HLRS - HP - IBM - Indiana U. - INRIA - Intel - LLNL - LANL - Microsoft - NCSA/UIUC - ORNL - OSU - Oracle - SNL - U Alabama Birmingham - U Tennessee - U. Tokyo

1 0

Re: [Mpi-forum] MPI 3.0
by Pavan Balaji 11 May '11

11 May '11

On 05/11/2011 07:46 PM, RICHARD GRAHAM wrote: > - Hybrid: Shared memory communicator > - Hybrid: Threads > - Hybrid: Endpoints We had decided to merge these. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [Mpi-forum] MPI 3.0
by RICHARD GRAHAM 11 May '11

11 May '11

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(a)msn.com To: mpi-forum(a)lists.mpi-forum.org 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(a)lists.mpi-forum.org http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

1 0

Re: [Mpi-forum] 1st reading results
by Jeff Squyres 11 May '11

11 May '11

On May 11, 2011, at 5:16 PM, Jeff Squyres wrote: > PASSED (i.e., can move on to 1st vote) > - RMA > - neighborhood collectives > - MPI_Count > - const > - remove NBC C++ bindings And Mprobe's cleanup ticket, too. Sorry for forgetting to mention that. > WENT BACK FOR REVISION (i.e., reset the clock) > - MPIT > - New Fortran bindings -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

MPI 3.0
by RICHARD GRAHAM 11 May '11

11 May '11

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

1 0

1st reading results
by Jeff Squyres 11 May '11

11 May '11

We had several 1st readings at this meeting: PASSED (i.e., can move on to 1st vote) - RMA - neighborhood collectives - MPI_Count - const - remove NBC C++ bindings WENT BACK FOR REVISION (i.e., reset the clock) - MPIT - New Fortran bindings -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Slides at the Forum
by Jeff Squyres 11 May '11

11 May '11

Please send me all slides that you have presented (or will present today) at the Forum; thanks. -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

MProbe proposal question
by Barrett, Brian W 06 May '11

06 May '11

All - The MProbe proposal (which was voted into MPI-3 in March) accidentally neglected to update the list of handle manipulation functions in Section 16.3.4 (Transfer of Handles) to include two new lines: MPI_Message MPI_Message_f2c(MPI_Fint message) MPI_Fint MPI_Message_c2f(MPI_Message message) Unfortunately, while I prototyped the implementation in Open MPI, I neglected to implement the Fortran bindings (They're simple wrappers, who needs 'em? :) ) until recently. So this slipped through the cracks when we had the reading and votes on the original proposal. Consistent with the other handle transfer functions (for Requests, Windows, etc.), those are the only two lines which need to be changed. I've talked to Torsten, and we propose suspending the 2 week rule for a first reading at next week's Forum for of a proposal to add the two lines. I don't particularly like the idea on principle, but it's a very simple change and we're running out of meetings before the October meeting (which is when I assume we will vote on releasing a draft for SC). If people don't like that, any other thoughts on the best way to correct the omission? Brian -- Brian W. Barrett Dept. 1423: Scalable System Software Sandia National Laboratories

1 0

Fortran Official Reading Tickets #229 - #253 at May 9-11, 2011 meeting
by Rolf Rabenseifner 06 May '11

06 May '11

Dear all, two weeks ago, I send out a version for the formal reading. In the last two weeks i made several (most editorial) updates. I expect, that you represent an organization that uses Fortran or that has Fortran customers. Therefore, you may be heavily involved. On ticket https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/229 I have prepared a new version of the set of pdf files: Please can you print https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/229/mpi-repo… and take it with you to our meeting at Cisco. It contains most of the solutions of the Fortran tickets and can be perfectly used for comments. It is a version without the deleted text and therefore best readable. If you have already printed and read and commented the version from April 21, 2011, then this is also fine. If you have not yet read and commented, then you should take this newest version with you. In the ticket's attachment zone, you find also a compilable version of the new module: https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/229/mpi-3.0-… Best regards and until Monday Rolf -- 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: Allmandring 30)

1 0