[Mpi-forum] Deadline Sep.21 -- Reviews for New Fortran bindings - Formal reading

Moody, Adam T. moody20 at llnl.gov
Thu Sep 22 10:30:11 CDT 2011


Hi Rolf,
I found a few things in the collectives chapter:

Why not list sendtype and recvtype on same line?
  MPI_Gather
  MPI_Gatherv
  MPI_Scatter
  MPI_Scatterv
  MPI_Allgather
  MPI_Allgatherv
  MPI_Alltoall
  MPI_Alltoallv
  
At least the count, displacement, and datatype arrays on icollectives need to be marked as ASYNCHRONOUS due to page 202, line 2-5 of mpi-report-F2008-2011-09-08-reviewdoc.pdf

"Once initiated, all associated send buffers and buffers associated with input arguments (such
as arrays of counts, displacements, or datatypes in the vector versions of the collectives)
should not be modifed, and all associated receive buffers should not be accessed, until the
collective operation completes."

MPI_Igatherv
  recvcounts, displs
MPI_Iscatterv
  sendcounts, displs
MPI_Iallgatherv
  recvcounts, displs
MPI_Ialltoallv
  sendcounts, sdispls, recvcounts, rdispls
MPI_Ialltoallw
  sendcounts, sdispls, recvcounts, rdispls, sendtypes, recvtypes
MPI_Ireduce_scatter
  recvcounts

-Adam


________________________________________
From: mpi-forum-bounces at lists.mpi-forum.org [mpi-forum-bounces at lists.mpi-forum.org] On Behalf Of Rolf Rabenseifner [rabenseifner at hlrs.de]
Sent: Tuesday, September 20, 2011 11:53 PM
To: Main MPI Forum mailing list
Subject: [Mpi-forum] Deadline Sep.21 -- Reviews for New Fortran bindings - Formal reading

Dear all, especially,
dear owners of new MPI-3 tickets, and
dear chapter author,

If a review is missing, we need to execute a line-by-line
formal reading. I expect that nobody want this.

I hope, you all will return your reviews today (Sep. 21)
that I can prepare the ppt with the review-results
for the MPI Forum meeting (Sep. 22).

For most chapters, it is mainly to check the
argument lists of the new bindings.

Responsible:
Special Fortran reviewers:
  "Jeff Squyres" <jsquyres at cisco.com>;
  "Craig E Rasmussen" <rasmussn at lanl.gov>;  (IN SANTORINI)
  "Hubert Ritzdorf" <Hubert.Ritzdorf at EMEA.NEC.COM>;
  "Purushotham Bangalore" <puri at cis.uab.edu>;
  "Bill Gropp" <wgropp at uiuc.edu>;
  "Reinhold Bader" <Reinhold.Bader at lrz.de>
Chapter authors from
 http://meetings.mpi-forum.org/mpi3.0_chapter_wgs.php
 - Front matter : Bill Gropp
 - Introduction : Bill Gropp
 - Terms and Definitions : Rolf Rabenseifner (IN SANTORINI)
 - Point-to-Point Communication : Richard Graham (IN SANTORINI)
 - Datatypes: George Bosilca (IN SANTORINI)
 - Collective Communication : Adam Moody (IN SANTORINI), Torsten Hoefler (IN SANTORINI),
    Yutaka Ishikawa, Hideyuki Jitsumoto
    Who is responsible?
 - Process Topologies : Torsten Hoefler (IN SANTORINI)
 - Groups, Contexts, and Communicators : Bill Gropp
 - Environmental Management : George Bosilca (IN SANTORINI)
 - Info-Object : Darius Buntinas (IN SANTORINI)
 - Process Creation and Management : David Solt
 - One-Sided Communications : Bill Gropp + Rajeev Thakur
 - External Interfaces : Bronis R. de Supinski
 - I/O : Rajeev Thakur(c)
 - Profiling interface : Bronis R. de Supinski + Martin Schulz
 - Deprecated Functions : Rolf Rabenseifner (IN SANTORINI)
 - Language Bindings : Jeffrey M. Squyres
 - Annex Language Bindings Summary : Rolf Rabenseifner (IN SANTORINI)
 - Annex Change-Log : Rolf Rabenseifner (IN SANTORINI)
 - Bibliography : Bill Gropp

Best regards
Rolf


----- Original Message -----
> From: "Rolf Rabenseifner" <rabenseifner at hlrs.de>
> To: "Jeff Squyres" <jsquyres at cisco.com>
> Cc: "William Gropp" <wgropp at illinois.edu>, "Hubert Ritzdorf" <Hubert.Ritzdorf at EMEA.NEC.COM>, "Purushotham Bangalore"
> <puri at cis.uab.edu>, "Bill Gropp" <wgropp at uiuc.edu>, "Reinhold Bader" <Reinhold.Bader at lrz.de>, "Richard Graham"
> <rlgraham at ornl.gov>, "George Bosilca" <bosilca at eecs.utk.edu>, "Adam Moody" <moody20 at llnl.gov>, "Torsten Hoefler"
> <htor at illinois.edu>, "David George Solt" <david.solt at hp.com>, "Rajeev Thakur" <thakur at mcs.anl.gov>, "Bronis R. de
> Supinski" <bronis at llnl.gov>, "Martin Schulz" <schulzm at llnl.gov>, "Darius Buntinas" <buntinas at mcs.anl.gov>, "Craig E
> Rasmussen" <rasmussn at lanl.gov>
> Sent: Friday, September 9, 2011 8:54:33 AM
> Subject: Re: Pre-review - New Fortran bindings - Formal reading
> Dear all,
>
> It is a really huge ticket that needs the review-support of
> all chapter responsibles. I try to sort it out:
>
> A) Normal chapter-reviewers:
> Officially, reviewers have to review always the reviewdoc, i.e.,
> mpi-report-F2008-2011-09-08-reviewdoc.pdf
> This means, one must mainly review the diff, i.e., the red portions
> marked with ticket. This is recommended for the normal chapter
> reviewers.
>
> B) Fortran-Chapter-Reviewers:
> For Section 16.2 "Fortran Support", it is strongly recommended to
> review
> mpi-report-F2008-2011-09-08-changeonlyplustickets.pdf
> because this is the only readable form of this section.
> This means one must read the total Section 16.2 as a whole (black and
> red).
>
> C) For owners of MPI-3 tickets:
> Please take
> mpi-report-F2008-2011-09-08-changeonlyplustickets.pdf
> and check for Annex B.4.
> The interfaces are always two:
> - the new mpif_f08
> - and the old mpi.
> The old one (i.e., the second form) must be checked whether
> it is identical to your latest version of your ticket.
> The new one (i.e., the first form) must be checked
> whether it is correctly mapped to mpi_f08.
>
> D) All:
> If one wants to make handwritten marks, for example on travel,
> then I recommend to print
> mpi-report-F2008-2011-09-08-changeonlyplustickets_majorpages.pdf
> and in
> mpi-report-F2008-2011-09-08-changeonlyplustickets.pdf
> the Annexes A.3 (p.553-695) and B.4 (p.753-770).
>
> As already said, it is huge. It touches 140 pages (...majorpages.pdf)
> and contains 43 pages of new interfaces for MPI-2.2 (Annex A.3)
> and additionally 18 pages for MPI-3.0 tickets (Annex B.4)
>
> E) Compilation-Check:
> If you want compile the module interface, then you must download
> mpi-3.0-all-interfaces-08-f2003.f90
>
> F) If you want to see, how it looks like in the future
> if you have a F2008+TR29113 compiler, then you must look at
> mpi-3.0-all-interfaces-12-f2008TR29113.f90
>
> G) and finally, the file
> mpi-3.0-all-interfaces-07-f2008.f90
> contains exactly what is part of the pdf files Annex A.3.
> To be more exact, this is the starting point of all
> the interfaces in the pdf files and of the other two
> .f90 files. All is done by a set of macros, translating this
> only one basis into Latex and real Fortran F2003 and F2008+TR29113.
> (the .2 version is a double uploading, it should be removed
> but I have no rights to remove such duplicated files
> Jeff may help?)
>
> Best regards
> Rolf
>
> ----- Original Message -----
> > From: "Jeff Squyres" <jsquyres at cisco.com>
> > To: "Rolf Rabenseifner" <rabenseifner at hlrs.de>
> > Cc: "William Gropp" <wgropp at illinois.edu>, "Hubert Ritzdorf"
> > <Hubert.Ritzdorf at EMEA.NEC.COM>, "Purushotham Bangalore"
> > <puri at cis.uab.edu>, "Bill Gropp" <wgropp at uiuc.edu>, "Reinhold Bader"
> > <Reinhold.Bader at lrz.de>, "Richard Graham"
> > <rlgraham at ornl.gov>, "George Bosilca" <bosilca at eecs.utk.edu>, "Adam
> > Moody" <moody20 at llnl.gov>, "Torsten Hoefler"
> > <htor at illinois.edu>, "David George Solt" <david.solt at hp.com>,
> > "Rajeev Thakur" <thakur at mcs.anl.gov>, "Bronis R. de
> > Supinski" <bronis at llnl.gov>, "Martin Schulz" <schulzm at llnl.gov>,
> > "Darius Buntinas" <buntinas at mcs.anl.gov>, "Craig E
> > Rasmussen" <rasmussn at lanl.gov>
> > Sent: Friday, September 9, 2011 12:28:23 AM
> > Subject: Re: Pre-review - New Fortran bindings - Formal reading
> > Rolf --
> >
> > I see 7 files attached to the ticket from today. Which one are we
> > supposed to review?
> >
> >
> > mpi-report-F2008-2011-09-08-changeonlyplustickets_majorpages.pdf
> > (843.5 KB) - added by RolfRabenseifner 2 hours ago.
> > Updated version for reading/reviewing at MPI Forum meeting Sep 2011
> > -
> > deleted text is invisible - major pages for printing
> > mpi-report-F2008-2011-09-08-changeonlyplustickets.pdf (4.0 MB) -
> > added
> > by RolfRabenseifner 2 hours ago.
> > Updated version for reading/reviewing at MPI Forum meeting Sep 2011
> > -
> > deleted text is invisible
> > mpi-report-F2008-2011-09-08-reviewdoc.pdf (4.1 MB) - added by
> > RolfRabenseifner 2 hours ago.
> > Updated version for reading/reviewing at MPI Forum meeting Sep 2011
> > -
> > deleted text is still visible
> > mpi-3.0-all-interfaces-07-f2008.f90 (90.3 KB) - added by
> > RolfRabenseifner 2 hours ago.
> > mpi_f08 binding used in the MPI draft 2011-09-08
> > mpi-3.0-all-interfaces-07-f2008.2.f90 (90.3 KB) - added by
> > RolfRabenseifner 2 hours ago.
> > mpi_f08 binding used in the MPI draft 2011-09-08
> > mpi-3.0-all-interfaces-08-f2003.f90 (184.1 KB) - added by
> > RolfRabenseifner 2 hours ago.
> > mpi_f08 module, preliminary implementation with Fortran 2003
> > mpi-3.0-all-interfaces-12-f2008TR29113.f90 (175.8 KB) - added by
> > RolfRabenseifner 2 hours ago.
> > mpi_f08 module, final implementation for Fortran2008+TR29113
> > compiler
> >
> >
> > On Sep 8, 2011, at 4:51 PM, Rolf Rabenseifner wrote:
> >
> > >
> > > Dear reviewers,
> > >
> > > Many thanks for the pre-reviews especially to
> > > Jeff, Reinhold, Craig, Hubert, Puri, Martin, Rajeev, Darius,
> > > George,
> > > Bill, and Bronis.
> > >
> > > My apologies that I had to few time for answering individually.
> > >
> > > All (or nearly all) is included in this final version.
> > > I expect there are still bugs but I hope that we can get them
> > > fixed
> > > after your final reviews in a new ticket.
> > >
> > > This version should go into your reviews, in the formal reading
> > > and through the two votes.
> > >
> > > All files are attached to the ticket
> > > https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/229/
> > >
> > > The review should be based only on the latest pdf-files from
> > > 2011-09-08.
> > > The content of the tickets is not updated, it tells only some
> > > reasons
> > > in a very early state. Some of these reasons are now rationales
> > > in the document.
> > >
> > > Please finish your review on the day before the meeting starts,
> > > i.e., not later than
> > >
> > >               September 21.
> > >
> > > Most of you are responsible for one or more chapters.
> > > If you do not speak Fortran, I expect, that you organized
> > > already a colleague who assists you in reviewing.
> > >
> > > If new MPI-3 tickets will go into your chapter,
> > > you should also review Annex B.4 that contains already
> > > interfaces for those tickets.
> > >
> > > Responsible:
> > > Special Fortran reviewers:
> > >  "Jeff Squyres" <jsquyres at cisco.com>;
> > >  "Craig E Rasmussen" <rasmussn at lanl.gov>;
> > >  "Hubert Ritzdorf" <Hubert.Ritzdorf at EMEA.NEC.COM>;
> > >  "Purushotham Bangalore" <puri at cis.uab.edu>;
> > >  "Bill Gropp" <wgropp at uiuc.edu>;
> > >  "Reinhold Bader" <Reinhold.Bader at lrz.de>
> > > Chapter authors from
> > >  http://meetings.mpi-forum.org/mpi3.0_chapter_wgs.php
> > > - Front matter : Bill Gropp
> > > - Introduction : Bill Gropp
> > > - Terms and Definitions : Rolf Rabenseifner
> > > - Point-to-Point Communication : Richard Graham
> > > - Datatypes: George Bosilca
> > > - Collective Communication : Adam Moody, Torsten Hoefler, Yutaka
> > > Ishikawa, Hideyuki Jitsumoto
> > >                             Who is responsible?
> > > - Process Topologies : Torsten Hoefler
> > > - Groups, Contexts, and Communicators : Bill Gropp
> > > - Environmental Management : George Bosilca
> > > - Info-Object : Darius Buntinas
> > > - Process Creation and Management : David Solt
> > > - One-Sided Communications : Bill Gropp + Rajeev Thakur
> > > - External Interfaces : Bronis R. de Supinski
> > > - I/O : Rajeev Thakur(c)
> > > - Profiling interface : Bronis R. de Supinski + Martin Schulz
> > > - Deprecated Functions : Rolf Rabenseifner
> > > - Language Bindings : Jeffrey M. Squyres
> > > - Annex Language Bindings Summary : Rolf Rabenseifner
> > > - Annex Change-Log : Rolf Rabenseifner
> > > - Bibliography : Bill Gropp
> > >
> > > Best regards
> > > Rolf
> >
> >
> > --
> > Jeff Squyres
> > jsquyres at cisco.com
> > For corporate legal information go to:
> > http://www.cisco.com/web/about/doing_business/legal/cri/
>
> --

--
Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner at 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)
_______________________________________________
mpi-forum mailing list
mpi-forum at lists.mpi-forum.org
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum




More information about the mpi-forum mailing list