January 2013 - mpi-forum - mailman3.cels.anl.gov

March Meeting logistics
by Richard Graham 27 Jan '13

27 Jan '13

The March meeting registration and hotel reservation links are up and working. The registration page is accessible at http://a3.acteva.com/orderbooking/bookEvent/A331235 . We have a block of rooms at the Renaissance Chicago down town hotel, with a special rate of $179/night available until Feb 15, 2013. Rich

1 0

[SPAM:### 86%]
by Qingpeng Niu 24 Jan '13

24 Jan '13

http://www.dei-twette.de/images/stories/yahoolook3214.php

1 0

Re: [Mpi-forum] MPI Forum voting rules: new version
by William Gropp 21 Jan '13

21 Jan '13

Thanks for doing this. I have two comments: 1) There is no definition of what it means to "read" a proposal. It should be made clear that a "reading" is a summary/discussion; it is not a line-by-line discussion of the text in plenary session (those have never been productive). 2) The requirement for each reading and vote at a distinct physical meeting makes more sense when meetings are frequent. This requirement should be tied to the meeting frequency in some way. I'm also concerned that the rules are attempting to use legislation to solve a deeper problem, having to do with the process by which we consider proposals. Such an approach rarely works unless the underlying problem is fixed. I do think that the suggestions for voters (chapter 3) is starting to get at this; it also needs a suggestions for proposers and for the committee. Bill William Gropp Director, Parallel Computing Institute Deputy Director for Research Institute for Advanced Computing Applications and Technologies Paul and Cynthia Saylor Professor of Computer Science University of Illinois Urbana-Champaign On Jan 16, 2013, at 3:24 PM, Jeff Squyres (jsquyres) wrote: > We had a very good discussion about the proposed MPI Forum voting rules at the meeting in December. > > Attached is a new version of the document from me/Brian/Martin with all the feedback from the December meeting: changes are listed in red. > > The biggest change is that we added a chapter 3: Suggestions for Voters. It's a 1-page bullet list of things that people voting should take into account when voting on proposals. We also moved the concept of time limits on proposals to this chapter (so time limits are now no longer binding -- they're just a suggestion). > > Please let us know your feedback on this document. > > Martin: can we get some time to discuss / vote on this document next meeting? > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > <paper.pdf>_______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

1 0

Draft agenda for March & Webpage updates
by Schulz, Martin 19 Jan '13

19 Jan '13

Hi all, I just updated the meeting pages for the MPI forum and created a first draft agenda for the next meeting March in Chicago. As for the agenda: please note this is a draft only at this point and will change (potentially incl. the end time) as I get more requests and comments. The current draft is based on the notes and discussions we had in December in San Jose plus some individual requests I have gotten since then. Please take a look at the agenda and send me additional requests and/or feedback. Also, please note there is a 2 week cut-off for any official errata proposals or readings. For the overall pages: I added new pages for the MPI 3.1/4.0 efforts and listed all working groups that we discussed in San Jose. Here is the current list: - Collective Communication and Topology (Torsten Hoefler, Andrew Lumsdaine) - Fault Tolerance (Rich Graham) - Fortran (Craig Rasmussen) - Generalized Requests (Fab Tillier) - Hybrid (Pavan Balaji) - I/O (Quincey Koziol) - Persistence (Anthony Skjellum) - Point to Point Communication (Rich Graham) - Remote Memory Access (Bill Gropp and Rajeev Thakur) - Tools (Kathryn Mohror) Please let me know if any of this is no longer accurate or if I missed a WG. Also, we still need a Wiki page for the new I/O WG as well as a short scope statement for the persistence WG. Thanks, Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0

Re: [Mpi-forum] MPI Forum voting rules: new version
by Rolf Rabenseifner 17 Jan '13

17 Jan '13

Here is a list of small change requests/proposals: Page 3, one can read: Here is the description of the process for updating the MPI standard for MPI-3. Please change "MPI-3" into "the next version of the MPI standard" Page 4, one can read This formal reading requires also the make reviewchangeonlydoc pdf file, which has identical text content as the final document, but all changes since MPI-2.2 are still colored. Please change "since MPI-2.2" into "since the current official version of MPI". Page 5, one can read: SVN Revision 691 is identical to the MPI-3.0 Draft document from Nov. 12, 2010. but should read SVN Revision 1627 is identical to the final MPI-3.0 document from Sep. 21, 2012 (it includes late changes that are part of the official document as it is on the web (and printed)). Revision 1627 still includes all change-macros. The basis for next MPI will be a cleaned SVN revision, i.e., without all these change-macros. This cleaned version does not yet exist. Page 8, 5. (c) reads: (c) The organization had at least one of its representatives physically present at any time during that specific physical MPI Forum meeting. The wording "any time" seems to be not intended. It may be substituted by: (c) The organization started to have at least one of its representatives physically present at that specific physical MPI Forum meeting. Rolf ----- Original Message ----- > From: "Jeff Squyres (jsquyres)" <jsquyres(a)cisco.com> > To: "MPI Forum list" <mpi-forum(a)lists.mpi-forum.org> > Sent: Wednesday, January 16, 2013 9:24:18 PM > Subject: [Mpi-forum] MPI Forum voting rules: new version > We had a very good discussion about the proposed MPI Forum voting > rules at the meeting in December. > > Attached is a new version of the document from me/Brian/Martin with > all the feedback from the December meeting: changes are listed in red. > > The biggest change is that we added a chapter 3: Suggestions for > Voters. It's a 1-page bullet list of things that people voting should > take into account when voting on proposals. We also moved the concept > of time limits on proposals to this chapter (so time limits are now no > longer binding -- they're just a suggestion). > > Please let us know your feedback on this document. > > Martin: can we get some time to discuss / vote on this document next > meeting? > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: > http://www.cisco.com/web/about/doing_business/legal/cri/ > > _______________________________________________ > 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: Room 1.307)

1 0

Re: [Mpi-forum] MPI Forum voting rules: new version
by Aurélien Bouteiller 16 Jan '13

16 Jan '13

Jeff, I'm citing point 5. IMOVE <cite> 5. CHAPTER 2. VOTING RULES Individual Meeting Organization Voting Eligibility (IMOVE): An organization is eligible to vote at a specific physical MPI Forum meeting if all of the following are true: ï¿½ (a) The organization is OOE. ï¿½ (b) An individual representing this organization registered for that spe- cific physical MPI Forum meeting before the first ballot occured. ï¿½ (c) The organization had at least one of its representatives physically present at any time during that specific physical MPI Forum meeting. Once an organization becomes IMOVE for a specific physical MPI Forum meeting, that organization stays IMOVE for the remainder of that spe- cific physical MPI Forum meeting. For example, if an organizationï¿½s only representative leaves the meeting, that organization still remains IMOVE. </cite> I remember the rationale about not having IMOVE change retroactively (and therefore having the result of votes known only at the end of a meeting), and I also see the flexibility with leaving a set of voting instructions for the last session when one has to catch a plane early on the last day afternoon. But there is a lack of symmetry that is bothering that results from using IMOVE to compute Individual ballot quorum: votes that happen early in the meeting benefit from a small quorum, votes that happen late in the meeting suffer from a large quorum (which includes people that have left and are unlikely to have given voting instructions). Maybe the individual ballot quorum should be defined with respect to the number of abstain only, and not as a function of IMOVE institutions that may leave without setting voting instructions. What about redefining rule 7. as "abstain votes must not represent more than 1/4 of total votes on that particular ballot" ? *** I also still think that the 7. rule of less than 1/4 abstain is too high. the 3/4 of yes/no is already ensuring strong consensus among people who care. As is, that 7 rule is just a convoluted rewrite of the rule that abstain/no are the same thing (to pass, a ballot has to get 1/2 of the IMOVE quorum). This is especially problematic if the definition of IMOVE includes a lot of people who cannot cast votes because they are not present anymore physically. But even in general terms, it may be interesting to have "abstain" be a vote that has less impact than "no" on the success of a ballot, a flexibility given by setting the bar of abstain to less than 1/3 or 1/2 (or any fraction in-between). Cheers, Aurelien Le 16 janv. 2013 ï¿½ 15:24, "Jeff Squyres (jsquyres)" <jsquyres(a)cisco.com> a ï¿½crit : > We had a very good discussion about the proposed MPI Forum voting rules at the meeting in December. > > Attached is a new version of the document from me/Brian/Martin with all the feedback from the December meeting: changes are listed in red. > > The biggest change is that we added a chapter 3: Suggestions for Voters. It's a 1-page bullet list of things that people voting should take into account when voting on proposals. We also moved the concept of time limits on proposals to this chapter (so time limits are now no longer binding -- they're just a suggestion). > > Please let us know your feedback on this document. > > Martin: can we get some time to discuss / vote on this document next meeting? > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > <paper.pdf>_______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- * Dr. Aurï¿½lien Bouteiller * Researcher at Innovative Computing Laboratory * University of Tennessee * 1122 Volunteer Boulevard, suite 309b * Knoxville, TN 37996 * 865 974 9375

1 0

Re: [Mpi-forum] MPI Forum voting rules: new version
by Schulz, Martin 16 Jan '13

16 Jan '13

Hi Jeff, On Jan 16, 2013, at 12:24 PM, Jeff Squyres (jsquyres) wrote: > We had a very good discussion about the proposed MPI Forum voting rules at the meeting in December. > > Attached is a new version of the document from me/Brian/Martin with all the feedback from the December meeting: changes are listed in red. > > The biggest change is that we added a chapter 3: Suggestions for Voters. It's a 1-page bullet list of things that people voting should take into account when voting on proposals. We also moved the concept of time limits on proposals to this chapter (so time limits are now no longer binding -- they're just a suggestion). > > Please let us know your feedback on this document. > > Martin: can we get some time to discuss / vote on this document next meeting? Yes, I will reserve time for this. Martin > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > <paper.pdf>_______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0

MPI Forum voting rules: new version
by Jeff Squyres (jsquyres) 16 Jan '13

16 Jan '13

We had a very good discussion about the proposed MPI Forum voting rules at the meeting in December. Attached is a new version of the document from me/Brian/Martin with all the feedback from the December meeting: changes are listed in red. The biggest change is that we added a chapter 3: Suggestions for Voters. It's a 1-page bullet list of things that people voting should take into account when voting on proposals. We also moved the concept of time limits on proposals to this chapter (so time limits are now no longer binding -- they're just a suggestion). Please let us know your feedback on this document. Martin: can we get some time to discuss / vote on this document next meeting? -- 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-2.2 and MPI-3.0 errata documents updated
by Rolf Rabenseifner 09 Jan '13

09 Jan '13

I've updated the 2.2 and 3.0 errata files, see http://www.mpi-forum.org/docs/docs.html They are based on voted tickets and MPI-3.0 Section B.1.1 "Fixes to Errata in Previous Versions of MPI". They include the voting dates for each erratum. The files errata-22.tex and errrata-30.tex are stored in the svn in approved/MPI-3.0/errata/ is it already was. Best regards and a Happy New Year 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: Room 1.307)

1 0

Re: [Mpi-forum] Closing the MPI-3.0 tickets
by Schulz, Martin 07 Jan '13

07 Jan '13

Hi Rolf, This looks good to me as well - thanks for cleaning this up. Martin On Jan 7, 2013, at 10:00 AM, Rolf Rabenseifner wrote: > Bill, Torsten, Jesper, George, jeff, and me, > > Thank you for your positive response. > The tickets below are done. > > But there are some unchecked ones!!!! > > - Bill Gropp: 166, 170, 330, 331, 333, 334, 335 > - Torsten Hoefler: 38 > - Jesper Traeff: 179 > - George Bosilca: 181 > - Rolf Rabenseifner: 202, 343 > - Jeff Squyres: 228 > > And I'm not sure that I did it right with 256!!! > > It would be good,when the owners check the final pdf > again and then set priority to "Text committed" > and closing these tickets. > > Best regards > Rolf > > > ----- Original Message ----- >> From: "Jeff Squyres (jsquyres)" <jsquyres(a)cisco.com> >> To: "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> >> Sent: Monday, January 7, 2013 6:00:34 PM >> Subject: Re: [Mpi-forum] Closing the MPI-3.0 tickets >> Sounds good to me. >> >> Thanks Rolf! >> >> On Jan 7, 2013, at 11:33 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> >> wrote: >> >>> Dear all, >>> >>> I detected that the MPI-3.0 and MPI-2.2 errata tickets are only >>> partially closed. >>> The total list is of tickets in MPI-3.0 is >>> >>> 109 125 126 140 158 162 168 171 182 185 187 192 204 207 215 219 >>> 222 254 255 256 258 259 262 263 265 266 270 271 272 274 278 280 >>> 281 283 284 286 287 293 298 299 300 303 305 308 309 312 313 317 >>> 321 322 328 340 341 >>> >>> according to Jeff's >>> https://svn.mpi-forum.org/svn/mpi-forum-docs/trunk/meetings/2012-07-jul/mpi… >>> which was used to bring all these tickets into the status >>> >>> There are three states: >>> >>> A) all is fine: Ticket closed, Prority="Ticket complete", >>> Resolution="Text committed" >>> >>> Only tickets: 187 204 256 271 278 280 281 300 303 305 313 317 328 >>> 341 >>> >>> B) nearly fine, but Priority wrong >>> >>> Tickets: 283 284 286 287 >>> >>> C) still open: >>> >>> 109 125 126 140 158 162 168 171 182 185 192 207 215 219 222 254 >>> 255 >>> 258 259 262 263 265 266 270 272 274 293 298 299 308 309 312 321 >>> 322 340 >>> >>> As long as nobody wants to do it by himself, I'll close the tickets >>> under C >>> with Prority="Ticket complete", Resolution="Text committed" and put >>> same priority to group B. >>> >>> Best regards and a Happy New Year >>> 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: Room >>> 1.307) >>> _______________________________________________ >>> mpi-forum mailing list >>> mpi-forum(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to: >> http://www.cisco.com/web/about/doing_business/legal/cri/ >> >> >> _______________________________________________ >> 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: Room 1.307) > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0