September 2012 - mpi-forum - mailman3.cels.anl.gov

Re: [Mpi-forum] Last changes to the draft standard
by Richard Graham 12 Sep '12

12 Sep '12

2 hours to cutoff .... I see some discussion going on, which is why this reminder is going out. From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Wednesday, September 12, 2012 8:20 AM To: mpi-forum(a)lists.mpi-forum.org Subject: [Mpi-forum] Last changes to the draft standard The opportunity to make last minute ticket-0 changes will be at 4pm EDT today, Wed Sept 12th. I expect to post the version that we will vote on next week sometime this evening. Rich

1 0

Re: [Mpi-forum] MPI 3.0 Implementation Question
by Dhabaleswar Panda 12 Sep '12

12 Sep '12

Hi Rich, You can put MPICH2 time frame and a `delta' of 3-6 months. Thanks, DK On Wed, 12 Sep 2012, Richard Graham wrote: > How do you want me to word this on the slide ? MPICH2 time frame + some delta ? If so, what is the order of magnitude of the delta, and do we just put the same time frame as MPICH2 ? > > -----Original Message----- > From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Dhabaleswar Panda > Sent: Wednesday, September 12, 2012 11:50 AM > To: Main MPI Forum mailing list > Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question > > Hi Rich, > > The upcoming releases of MVAPICH2 will be based on MPICH2 1.5 and successive versions and thus, will support several of the MPI-3 features supported by MPICH2. > > Thanks, > > DK > > On Wed, 12 Sep 2012, Richard Graham wrote: > >> As of now I have only received responses on this from Open MPI, MPICH, and SGI. Do the other implementations want this information included ? How about Intel, IBM, MVAPICH, NEC, Fujitsu, HP, .... ? >> >> Thanks, >> Rich >> >> -----Original Message----- >> From: mpi-forum-bounces(a)lists.mpi-forum.org >> [mailto:[email protected]] On Behalf Of Richard >> Graham >> Sent: Monday, September 03, 2012 9:52 PM >> To: Main MPI Forum mailing list >> Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question >> >> Feel free to add these to the list of your responses. >> >> Rich >> >> -----Original Message----- >> From: mpi-forum-bounces(a)lists.mpi-forum.org >> [mailto:[email protected]] On Behalf Of Pavan >> Balaji >> Sent: Monday, September 03, 2012 8:25 PM >> To: mpi-forum(a)lists.mpi-forum.org >> Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question >> >> >> Below is the list for MPICH -- >> >> On 09/03/2012 07:13 PM, Richard Graham wrote: >>> - Nonblocking collectives >>> - MPI Tool Interface >>> - Distributed Graph >>> - Noncollective communicator creation >> >> Already supported. >> >>> - RMA enhancements >>> - New Fortran bindings >>> - Sparse collectives >>> - Enhanced Datatype support >> >> November 2012. >> >> What about the remaining features that are voted in? >> >> - hindexed_block >> - const bindings >> - matched probe >> - intercomm modifications to not conflict with pt2pt tags >> - comm_split_type >> >> For MPICH, these are all already supported. >> >> -- Pavan >> >> -- >> Pavan Balaji >> http://www.mcs.anl.gov/~balaji >> _______________________________________________ >> 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 >> >> _______________________________________________ >> 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 > > _______________________________________________ > 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] MPI 3.0 Implementation Question
by Richard Graham 12 Sep '12

12 Sep '12

How do you want me to word this on the slide ? MPICH2 time frame + some delta ? If so, what is the order of magnitude of the delta, and do we just put the same time frame as MPICH2 ? -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Dhabaleswar Panda Sent: Wednesday, September 12, 2012 11:50 AM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Hi Rich, The upcoming releases of MVAPICH2 will be based on MPICH2 1.5 and successive versions and thus, will support several of the MPI-3 features supported by MPICH2. Thanks, DK On Wed, 12 Sep 2012, Richard Graham wrote: > As of now I have only received responses on this from Open MPI, MPICH, and SGI. Do the other implementations want this information included ? How about Intel, IBM, MVAPICH, NEC, Fujitsu, HP, .... ? > > Thanks, > Rich > > -----Original Message----- > From: mpi-forum-bounces(a)lists.mpi-forum.org > [mailto:[email protected]] On Behalf Of Richard > Graham > Sent: Monday, September 03, 2012 9:52 PM > To: Main MPI Forum mailing list > Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question > > Feel free to add these to the list of your responses. > > Rich > > -----Original Message----- > From: mpi-forum-bounces(a)lists.mpi-forum.org > [mailto:[email protected]] On Behalf Of Pavan > Balaji > Sent: Monday, September 03, 2012 8:25 PM > To: mpi-forum(a)lists.mpi-forum.org > Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question > > > Below is the list for MPICH -- > > On 09/03/2012 07:13 PM, Richard Graham wrote: >> - Nonblocking collectives >> - MPI Tool Interface >> - Distributed Graph >> - Noncollective communicator creation > > Already supported. > >> - RMA enhancements >> - New Fortran bindings >> - Sparse collectives >> - Enhanced Datatype support > > November 2012. > > What about the remaining features that are voted in? > > - hindexed_block > - const bindings > - matched probe > - intercomm modifications to not conflict with pt2pt tags > - comm_split_type > > For MPICH, these are all already supported. > > -- Pavan > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji > _______________________________________________ > 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 > > _______________________________________________ > 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

1 0

Re: [Mpi-forum] MPI 3.0 Implementation Question
by Dhabaleswar Panda 12 Sep '12

12 Sep '12

Hi Rich, The upcoming releases of MVAPICH2 will be based on MPICH2 1.5 and successive versions and thus, will support several of the MPI-3 features supported by MPICH2. Thanks, DK On Wed, 12 Sep 2012, Richard Graham wrote: > As of now I have only received responses on this from Open MPI, MPICH, and SGI. Do the other implementations want this information included ? How about Intel, IBM, MVAPICH, NEC, Fujitsu, HP, .... ? > > Thanks, > Rich > > -----Original Message----- > From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham > Sent: Monday, September 03, 2012 9:52 PM > To: Main MPI Forum mailing list > Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question > > Feel free to add these to the list of your responses. > > Rich > > -----Original Message----- > From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Pavan Balaji > Sent: Monday, September 03, 2012 8:25 PM > To: mpi-forum(a)lists.mpi-forum.org > Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question > > > Below is the list for MPICH -- > > On 09/03/2012 07:13 PM, Richard Graham wrote: >> - Nonblocking collectives >> - MPI Tool Interface >> - Distributed Graph >> - Noncollective communicator creation > > Already supported. > >> - RMA enhancements >> - New Fortran bindings >> - Sparse collectives >> - Enhanced Datatype support > > November 2012. > > What about the remaining features that are voted in? > > - hindexed_block > - const bindings > - matched probe > - intercomm modifications to not conflict with pt2pt tags > - comm_split_type > > For MPICH, these are all already supported. > > -- Pavan > > -- > Pavan Balaji > http://www.mcs.anl.gov/~balaji > _______________________________________________ > 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 > > _______________________________________________ > 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] MPI 3.0 Implementation Question
by Mark Pagel 12 Sep '12

12 Sep '12

Rich, Correction. I meant to say for the 3.0 features. We still do not plan to support Dynamic Process Management on Cray systems. Corrected below. Our goal is to have a fully compliant MPI (except dynamic process management) by Q4 2013 but it may be earlier. Some individual features may be released early as they are implemented in MPICH2 and verified to work on Cray systems. Mark Pagel Cray Inc. -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Wednesday, September 12, 2012 8:48 AM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question As of now I have only received responses on this from Open MPI, MPICH, and SGI. Do the other implementations want this information included ? How about Intel, IBM, MVAPICH, NEC, Fujitsu, HP, .... ? Thanks, Rich -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Monday, September 03, 2012 9:52 PM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Feel free to add these to the list of your responses. Rich -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Pavan Balaji Sent: Monday, September 03, 2012 8:25 PM To: mpi-forum(a)lists.mpi-forum.org Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Below is the list for MPICH -- On 09/03/2012 07:13 PM, Richard Graham wrote: > - Nonblocking collectives > - MPI Tool Interface > - Distributed Graph > - Noncollective communicator creation Already supported. > - RMA enhancements > - New Fortran bindings > - Sparse collectives > - Enhanced Datatype support November 2012. What about the remaining features that are voted in? - hindexed_block - const bindings - matched probe - intercomm modifications to not conflict with pt2pt tags - comm_split_type For MPICH, these are all already supported. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji _______________________________________________ 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 _______________________________________________ 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] MPI 3.0 Implementation Question
by Mark Pagel 12 Sep '12

12 Sep '12

Rich, Our goal is to have a fully compliant version by Q4 2013 but it may be earlier. Some individual features may be released early as they are implemented in MPICH2 and verified to work on Cray systems. Mark Pagel Cray Inc. -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Wednesday, September 12, 2012 8:48 AM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question As of now I have only received responses on this from Open MPI, MPICH, and SGI. Do the other implementations want this information included ? How about Intel, IBM, MVAPICH, NEC, Fujitsu, HP, .... ? Thanks, Rich -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Monday, September 03, 2012 9:52 PM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Feel free to add these to the list of your responses. Rich -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Pavan Balaji Sent: Monday, September 03, 2012 8:25 PM To: mpi-forum(a)lists.mpi-forum.org Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Below is the list for MPICH -- On 09/03/2012 07:13 PM, Richard Graham wrote: > - Nonblocking collectives > - MPI Tool Interface > - Distributed Graph > - Noncollective communicator creation Already supported. > - RMA enhancements > - New Fortran bindings > - Sparse collectives > - Enhanced Datatype support November 2012. What about the remaining features that are voted in? - hindexed_block - const bindings - matched probe - intercomm modifications to not conflict with pt2pt tags - comm_split_type For MPICH, these are all already supported. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji _______________________________________________ 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 _______________________________________________ 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] MPI 3.0 Implementation Question
by Richard Graham 12 Sep '12

12 Sep '12

As of now I have only received responses on this from Open MPI, MPICH, and SGI. Do the other implementations want this information included ? How about Intel, IBM, MVAPICH, NEC, Fujitsu, HP, .... ? Thanks, Rich -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Richard Graham Sent: Monday, September 03, 2012 9:52 PM To: Main MPI Forum mailing list Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Feel free to add these to the list of your responses. Rich -----Original Message----- From: mpi-forum-bounces(a)lists.mpi-forum.org [mailto:[email protected]] On Behalf Of Pavan Balaji Sent: Monday, September 03, 2012 8:25 PM To: mpi-forum(a)lists.mpi-forum.org Subject: Re: [Mpi-forum] MPI 3.0 Implementation Question Below is the list for MPICH -- On 09/03/2012 07:13 PM, Richard Graham wrote: > - Nonblocking collectives > - MPI Tool Interface > - Distributed Graph > - Noncollective communicator creation Already supported. > - RMA enhancements > - New Fortran bindings > - Sparse collectives > - Enhanced Datatype support November 2012. What about the remaining features that are voted in? - hindexed_block - const bindings - matched probe - intercomm modifications to not conflict with pt2pt tags - comm_split_type For MPICH, these are all already supported. -- Pavan -- Pavan Balaji http://www.mcs.anl.gov/~balaji _______________________________________________ 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

1 0

Last changes to the draft standard
by Richard Graham 12 Sep '12

12 Sep '12

The opportunity to make last minute ticket-0 changes will be at 4pm EDT today, Wed Sept 12th. I expect to post the version that we will vote on next week sometime this evening. Rich

1 0

Re: [Mpi-forum] Fortran change for discussion
by N.M. Maclaren 11 Sep '12

11 Sep '12

On Sep 11 2012, Jeremiah Willcock wrote: >On Tue, 11 Sep 2012, Jeff Squyres (jsquyres) wrote: > >> FWIW, we chose the copying language not because of the c language, but >> rather because we have to explain that the MPI data type is applied, and >> that seemed far easier to explain against a (virtual) contiguous buffer. > >Maybe the issue is that I'm not a Fortran programmer, but the copying >language has always been confusing to me (its effect on what user-defined >datatypes need to look like is another thing I asked about before). Am I >to understand the interface as something like this? > > For purposes of interpreting derived datatypes and other information > in communication operations, array slices are treated as if they are > stored contiguously in memory. I.e., it is "as if" some kind of > magic memory remapping occurs by the MPI implementation that maps the > user's non-contiguous array into a contiguous one without copying the > elements (in either direction). No, that's not the reason, because your wording is considerably closer to the Fortran model than the current wording. What you have said is a fairly precise paraphrase of what sequence association is. Realistically, I agree with your concerns - this is confusing. Your addendum would help, as would adding the phrase "at an unspecified time between the initial call and the wait". Personally, I would do both. Regards, Nick Maclaren.

1 0

Re: [Mpi-forum] Fortran change for discussion
by Jeremiah Willcock 11 Sep '12

11 Sep '12

Maybe the issue is that I'm not a Fortran programmer, but the copying language has always been confusing to me (its effect on what user-defined datatypes need to look like is another thing I asked about before). Am I to understand the interface as something like this? For purposes of interpreting derived datatypes and other information in communication operations, array slices are treated as if they are stored contiguously in memory. I.e., it is "as if" some kind of magic memory remapping occurs by the MPI implementation that maps the user's non-contiguous array into a contiguous one without copying the elements (in either direction). -- Jeremiah Willcock On Tue, 11 Sep 2012, Jeff Squyres (jsquyres) wrote: > FWIW, we chose the copying language not because of the c language, but rather because we have to explain that the MPI data type is applied, and that seemed far easier to explain against a (virtual) contiguous buffer. > > Sent from my phone. No type good. > > On Sep 11, 2012, at 3:38 AM, "N.M. Maclaren" <nmm1(a)cam.ac.uk> wrote: > >> On Sep 11 2012, Jeff Squyres wrote: >>> >>> You explain the situation in Fortran-eese, but the current text is an attempt to explain it in MPI-eese (i.e., in terms of copying). I think the MPI-reader-friendly text should be used. >> >> Sort-of, I agree. The problem with the current description is precisely >> that it uses copying rather than aliasing as a basis, but I assume that >> is for C programmers. There is definitely a problem in trying to explain >> something in C-friendly terms that is conceptually outside C's rudimentary >> array model. >> >>>> If the current text is kept, I think the clarification is important. >>> >>> Ok. >> >> The other possibility is to stress that the copying can occur at any >> time between the initial call and the Wait. That is, after, all precisely >> what asynchronous means in this context! >> >> >> Regards, >> Nick. >> >> _______________________________________________ >> 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 >

1 0