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

Re: [Mpi-forum] ABI
by Jeff Hammond 04 Dec '13

04 Dec '13

I know another way for all MPI implementations to have the same ABI ;-) Jeff On Wed, Dec 4, 2013 at 3:23 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > Pavan Balaji <balaji(a)mcs.anl.gov> writes: >> Also, it’s not like we’ll never change ABI again. We’ll just be more >> conservative and everyone in the “team” will change it at the same >> time. If it’s in the MPI standard, we lose that flexibility. > > It would make a _huge_ difference for distributions if MPICH and Open > MPI could have a compatible ABI. As it is, distributions generally > choose one to be primary and have some system for maintaining a suite of > packages for the other. This wastes a ton of maintainer time and causes > confusion for users. New users of MPI tend to encounter it on personal > computers and small clusters rather than on leadership class machines, > so I would argue that compatibility between MPICH and Open MPI may have > a bigger reward in terms of human time and user experience than > compatibility with the various commercial vendors (which I also think is > useful). > > However, the question of integer (MPICH) versus pointer (OMPI) handles > is a long-standing impasse. Pointer handles are great for users because > they get stronger type checking and it's easier to probe around in the > debugger. Meanwhile, integer handles can use more tag bits to optimize > certain operations. But I wonder if we could reach a middle ground that > would at least provide a common ABI on x86 systems. > > Suppose MPICH used intptr_t for handles. I believe pointers are treated > the same as integers for all the common calling conventions. This > leaves a question of the predefined constants, which are currently > compile-time constants of some sort (either integers or addresses of > privately-defined structs), but I'm not aware of text in the standard > that requires them to be defined in a header. That is, can we have > > > extern MPI_Datatype MPI_INT; > > in the header and leave it to the library to provide the value? > > Is anything else a showstopper for using intptr_t versus opaque_type* as > handles? > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Hammond jeff.science(a)gmail.com

1 0

Re: [Mpi-forum] ABI
by Jed Brown 04 Dec '13

04 Dec '13

Pavan Balaji <balaji(a)mcs.anl.gov> writes: > Also, it’s not like we’ll never change ABI again. We’ll just be more > conservative and everyone in the “team” will change it at the same > time. If it’s in the MPI standard, we lose that flexibility. It would make a _huge_ difference for distributions if MPICH and Open MPI could have a compatible ABI. As it is, distributions generally choose one to be primary and have some system for maintaining a suite of packages for the other. This wastes a ton of maintainer time and causes confusion for users. New users of MPI tend to encounter it on personal computers and small clusters rather than on leadership class machines, so I would argue that compatibility between MPICH and Open MPI may have a bigger reward in terms of human time and user experience than compatibility with the various commercial vendors (which I also think is useful). However, the question of integer (MPICH) versus pointer (OMPI) handles is a long-standing impasse. Pointer handles are great for users because they get stronger type checking and it's easier to probe around in the debugger. Meanwhile, integer handles can use more tag bits to optimize certain operations. But I wonder if we could reach a middle ground that would at least provide a common ABI on x86 systems. Suppose MPICH used intptr_t for handles. I believe pointers are treated the same as integers for all the common calling conventions. This leaves a question of the predefined constants, which are currently compile-time constants of some sort (either integers or addresses of privately-defined structs), but I'm not aware of text in the standard that requires them to be defined in a header. That is, can we have extern MPI_Datatype MPI_INT; in the header and leave it to the library to provide the value? Is anything else a showstopper for using intptr_t versus opaque_type* as handles?

1 0

Re: [Mpi-forum] ABI
by Pavan Balaji 04 Dec '13

04 Dec '13

On Dec 4, 2013, at 2:58 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > On Dec 4, 2013, at 3:54 PM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > >> And no, we are not proposing it for MPI-4. That’s a horrible idea. :-) > > So why are you proposing it at all, then? :-) I didn’t say it’s not useful. I just said it shouldn’t be standardized. :-) Also, it’s not like we’ll never change ABI again. We’ll just be more conservative and everyone in the “team” will change it at the same time. If it’s in the MPI standard, we lose that flexibility. -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [Mpi-forum] ABI
by Jeff Squyres (jsquyres) 04 Dec '13

04 Dec '13

On Dec 4, 2013, at 3:54 PM, Pavan Balaji <balaji(a)mcs.anl.gov> wrote: > And no, we are not proposing it for MPI-4. That’s a horrible idea. :-) So why are you proposing it at all, then? :-) -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

ABI
by Pavan Balaji 04 Dec '13

04 Dec '13

Sorry for sending an MPICH-specific thing to the Forum list. Thought some folks might be interested in this given all of the ABI stuff that went on in the Forum earlier. http://www.hpcwire.com/off-the-wire/argonne-researchers-establish-mpich-abi… And no, we are not proposing it for MPI-4. That’s a horrible idea. :-) — Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

Re: [Mpi-forum] [EXTERNAL] Re: Dates for March Meeting
by Schulz, Martin 04 Dec '13

04 Dec '13

Seems like this is the best solution, since the move is not convenient or not favored by most people who have spoken up and we have declared those dates as fixed. I'll keep this in mind and will schedule WG time for RMA on Monday and Tuesday morning. Hopefully at least a few WG members can stick around to represent the group in the remaining forum time. Martin On Dec 3, 2013, at 7:55 AM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > If we schedule all the RMA WG sessions on Monday and Tuesday, crazy > people can take the redeye east and arrive for the start of the SHMEM > workshop (Tuesday is just tutorials). > > Jeff > > On Tue, Dec 3, 2013 at 9:44 AM, Barrett, Brian W <bwbarre(a)sandia.gov> wrote: >> I'm undecided which one I'll attend, but Sandia will have a representative >> at the March meeting. I'm not sure it's worth the pain of moving the >> Forum meeting at this point. >> >> Brian >> >> On 12/2/13 6:37 PM, "Jeff Hammond" <jeff.science(a)gmail.com> wrote: >> >>> Jim and I are on the OpenSHMEM agenda. Probably Barrett is going, too. >>> That's a significant portion of the RMA WG. >>> >>> Given the OpenSHMEM workshop has already been rescheduled once, I really >>> doubt they can move, nor would they care to for our sake. >>> >>> Jeff >>> >>> Sent from my iPhone >>> >>>> On Dec 2, 2013, at 5:55 PM, "Schulz, Martin" <schulzm(a)llnl.gov> wrote: >>>> >>>> Hi all, >>>> >>>> The dates for our meeting in March had been set after the September >>>> meeting to March 3-6 and we declared those as final a few weeks after >>>> that. >>>> >>>> However, Rich sent me email today pointing out the following conflict: >>>> >>>>> Martin, >>>>> It turns out that the March MPI Forum meeting end up right on top of >>>>> the OpenSHMEM workshop, on the other side of the country. I expect >>>>> this conflict will impact those in the RMA working group. Do you want >>>>> to look into possibly moving the Forum forward or back by a week ? I >>>>> am guessing it is easier for us to move than for the OpenSHMEM meeting, >>>>> given the numbers of people involved. >>>>> >>>>> Rich >>>> >>>> Moving it a week later doesn't work (due to some already scheduled DOE >>>> meetings), but a week earlier could work. We did, however, already fix >>>> the March dates so people may have already tickets. Could you please let >>>> me know if this is the case for any of you? Also, for the RMA group - >>>> how much of a conflict is this and will this reduce attendance >>>> drastically? >>>> >>>> In general, I am hesitant to switch meeting dates after they have been >>>> fixed, but if it is advantageous for the meeting and no one would be >>>> negatively impacted (and we would need pretty much unanimous consensus), >>>> we can take this into considerations. >>>> >>>> Comments? >>>> >>>> Thanks! >>>> >>>> Martin >>>> >>>> >>>> ________________________________________________________________________ >>>> Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm >>>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >>>> >>>> >>>> >>>> _______________________________________________ >>>> mpi-forum mailing list >>>> mpi-forum(a)lists.mpi-forum.org >>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >>> _______________________________________________ >>> mpi-forum mailing list >>> mpi-forum(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >>> >> >> >> -- >> Brian W. Barrett >> Scalable System Software Group >> Sandia National Laboratories >> >> >> >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > > > > -- > Jeff Hammond > jeff.science(a)gmail.com > _______________________________________________ > 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

Re: [Mpi-forum] Agenda for December Meeting
by Jim Dinan 03 Dec '13

03 Dec '13

Hi Martin, In concert with ticket #349 (MPI_Aint_add), I would also like to show ticket #402, which uses MPI_Aint_add to clean up an RMA example in the standard. This won't count as a formal reading for #402, because there isn't enough lead time, but it would be nice to include a link in the agenda so folks can review it ahead of time and follow along. Unfortunately, we found a couple mistakes in the wording of ticket #381 (communicator info keys) and had to make changes this week. We would still like to read that ticket for feedback, but may not be able to count this as a formal reading, because of making changes so close to the meeting. Thanks, ~Jim. On Thu, Nov 28, 2013 at 4:30 PM, Schulz, Martin <schulzm(a)llnl.gov> wrote: > Hi all, > > I put the agenda for the next MPI forum on the web site. This should > include all the requests I have gotten, if not please let me know. Also, I > know this schedule is not ideal for everyone, but I have gotten many > constraints from people this time and I hope this is the best compromise. > > Thanks, > > Martin > > PS: Please note, that on Tuesday we only have a smaller room plus a > breakout room - the intention is to have real WG meetings on that day, so I > hope this won't be a problem. > > ________________________________________________________________________ > Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > > _______________________________________________ > 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] [EXTERNAL] Re: Dates for March Meeting
by Jeff Hammond 03 Dec '13

03 Dec '13

If we schedule all the RMA WG sessions on Monday and Tuesday, crazy people can take the redeye east and arrive for the start of the SHMEM workshop (Tuesday is just tutorials). Jeff On Tue, Dec 3, 2013 at 9:44 AM, Barrett, Brian W <bwbarre(a)sandia.gov> wrote: > I'm undecided which one I'll attend, but Sandia will have a representative > at the March meeting. I'm not sure it's worth the pain of moving the > Forum meeting at this point. > > Brian > > On 12/2/13 6:37 PM, "Jeff Hammond" <jeff.science(a)gmail.com> wrote: > >>Jim and I are on the OpenSHMEM agenda. Probably Barrett is going, too. >>That's a significant portion of the RMA WG. >> >>Given the OpenSHMEM workshop has already been rescheduled once, I really >>doubt they can move, nor would they care to for our sake. >> >>Jeff >> >>Sent from my iPhone >> >>> On Dec 2, 2013, at 5:55 PM, "Schulz, Martin" <schulzm(a)llnl.gov> wrote: >>> >>> Hi all, >>> >>> The dates for our meeting in March had been set after the September >>>meeting to March 3-6 and we declared those as final a few weeks after >>>that. >>> >>> However, Rich sent me email today pointing out the following conflict: >>> >>>> Martin, >>>> It turns out that the March MPI Forum meeting end up right on top of >>>>the OpenSHMEM workshop, on the other side of the country. I expect >>>>this conflict will impact those in the RMA working group. Do you want >>>>to look into possibly moving the Forum forward or back by a week ? I >>>>am guessing it is easier for us to move than for the OpenSHMEM meeting, >>>>given the numbers of people involved. >>>> >>>> Rich >>> >>> Moving it a week later doesn't work (due to some already scheduled DOE >>>meetings), but a week earlier could work. We did, however, already fix >>>the March dates so people may have already tickets. Could you please let >>>me know if this is the case for any of you? Also, for the RMA group - >>>how much of a conflict is this and will this reduce attendance >>>drastically? >>> >>> In general, I am hesitant to switch meeting dates after they have been >>>fixed, but if it is advantageous for the meeting and no one would be >>>negatively impacted (and we would need pretty much unanimous consensus), >>>we can take this into considerations. >>> >>> Comments? >>> >>> Thanks! >>> >>> Martin >>> >>> >>> ________________________________________________________________________ >>> Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm >>> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >>> >>> >>> >>> _______________________________________________ >>> mpi-forum mailing list >>> mpi-forum(a)lists.mpi-forum.org >>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >>_______________________________________________ >>mpi-forum mailing list >>mpi-forum(a)lists.mpi-forum.org >>http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >> > > > -- > Brian W. Barrett > Scalable System Software Group > Sandia National Laboratories > > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum -- Jeff Hammond jeff.science(a)gmail.com

1 0

Re: [Mpi-forum] [EXTERNAL] Re: Dates for March Meeting
by Barrett, Brian W 03 Dec '13

03 Dec '13

I'm undecided which one I'll attend, but Sandia will have a representative at the March meeting. I'm not sure it's worth the pain of moving the Forum meeting at this point. Brian On 12/2/13 6:37 PM, "Jeff Hammond" <jeff.science(a)gmail.com> wrote: >Jim and I are on the OpenSHMEM agenda. Probably Barrett is going, too. >That's a significant portion of the RMA WG. > >Given the OpenSHMEM workshop has already been rescheduled once, I really >doubt they can move, nor would they care to for our sake. > >Jeff > >Sent from my iPhone > >> On Dec 2, 2013, at 5:55 PM, "Schulz, Martin" <schulzm(a)llnl.gov> wrote: >> >> Hi all, >> >> The dates for our meeting in March had been set after the September >>meeting to March 3-6 and we declared those as final a few weeks after >>that. >> >> However, Rich sent me email today pointing out the following conflict: >> >>> Martin, >>> It turns out that the March MPI Forum meeting end up right on top of >>>the OpenSHMEM workshop, on the other side of the country. I expect >>>this conflict will impact those in the RMA working group. Do you want >>>to look into possibly moving the Forum forward or back by a week ? I >>>am guessing it is easier for us to move than for the OpenSHMEM meeting, >>>given the numbers of people involved. >>> >>> Rich >> >> Moving it a week later doesn't work (due to some already scheduled DOE >>meetings), but a week earlier could work. We did, however, already fix >>the March dates so people may have already tickets. Could you please let >>me know if this is the case for any of you? Also, for the RMA group - >>how much of a conflict is this and will this reduce attendance >>drastically? >> >> In general, I am hesitant to switch meeting dates after they have been >>fixed, but if it is advantageous for the meeting and no one would be >>negatively impacted (and we would need pretty much unanimous consensus), >>we can take this into considerations. >> >> Comments? >> >> Thanks! >> >> Martin >> >> >> ________________________________________________________________________ >> Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm >> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >> >> >> >> _______________________________________________ >> mpi-forum mailing list >> mpi-forum(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum >_______________________________________________ >mpi-forum mailing list >mpi-forum(a)lists.mpi-forum.org >http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum > -- Brian W. Barrett Scalable System Software Group Sandia National Laboratories

1 0

Re: [Mpi-forum] Dates for March Meeting
by William Gropp 03 Dec '13

03 Dec '13

I'm already triple booked the week of Feb 24-28th - moving the MPI Forum meeting there won't work for me. Bill William Gropp Director, Parallel Computing Institute Deputy Director for Research Institute for Advanced Computing Applications and Technologies Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Dec 3, 2013, at 1:55 AM, Schulz, Martin wrote: > Hi all, > > The dates for our meeting in March had been set after the September meeting to March 3-6 and we declared those as final a few weeks after that. > > However, Rich sent me email today pointing out the following conflict: > >> Martin, >> It turns out that the March MPI Forum meeting end up right on top of the OpenSHMEM workshop, on the other side of the country. I expect this conflict will impact those in the RMA working group. Do you want to look into possibly moving the Forum forward or back by a week ? I am guessing it is easier for us to move than for the OpenSHMEM meeting, given the numbers of people involved. >> >> Rich > > Moving it a week later doesn't work (due to some already scheduled DOE meetings), but a week earlier could work. We did, however, already fix the March dates so people may have already tickets. Could you please let me know if this is the case for any of you? Also, for the RMA group - how much of a conflict is this and will this reduce attendance drastically? > > In general, I am hesitant to switch meeting dates after they have been fixed, but if it is advantageous for the meeting and no one would be negatively impacted (and we would need pretty much unanimous consensus), we can take this into considerations. > > Comments? > > Thanks! > > Martin > > > ________________________________________________________________________ > Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

1 0