June 2014 - mpiwg-fortran - mailman3.cels.anl.gov

Re: [MPIWG Fortran] errata
by Jeff Squyres (jsquyres) 17 Jun '14

17 Jun '14

Looks like Rolf already handled this on the ticket. On Jun 17, 2014, at 7:44 AM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > See https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/432 > > Jeff > > -- > Jeff Hammond > jeff.science(a)gmail.com > http://jeffhammond.github.io/ > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

errata
by Jeff Hammond 17 Jun '14

17 Jun '14

See https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/432 Jeff -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [MPIWG Fortran] MPI-3.0 Fortran errata latex integration: Reviews requested
by Jeff Squyres (jsquyres) 16 Jun '14

16 Jun '14

I’ve received one review so far (from Hubert; he found a mistake). Can others please review this text to ensure I did the merge properly? It’s a LOT of text to merge — I would really appreciate some additional eyeballs on it to ensure it was done properly. Thanks! On May 14, 2014, at 12:54 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > I have integrated all the MPI-3.0 Fortran-related errata as of this morning into the MPI-3.1 tree latex. It is not committed in the SVN MPI-3.1 tree because I'd like to get reviews first. > > Per Forum direction, I did *not* use the latex change macros. Instead, we're relying on diffs to review the changes. > > Once I get 2-3 reviews to ensure that I have applied the errata properly, I'll commit all of it to the official MPI-3.1 SVN tree. > > ----- > > Here's how I've made the changes available for review: > > 0. All the files listed below are available here: > > http://www.open-mpi.org/~jsquyres/mpi-forum/ > > 1. A PDF of the entire standard is there; all Fortran-related errata have been applied. You should proof this document. To know exactly what to proof in this PDF, however, see the follow documents... > > 2. I also put my copy of errata-30.pdf there, including BRIGHT RED annotations indicating which errata items I incorporated into the PDF from #1. > > 3. I also posted a tarball of a git tree where I did my work (NOTE: because this git tree includes the entire latex source, it is behind a trivial password. Email me to get it). There are 36 commits; each commit is the application of an individual errata item. You can use normal git commands to view the diffs. For example: > > tar xf mpi-3.1-tree-with-fortran-errata-applied-git.tar.bz2 > cd mpi-3.1 > git log > # ...see all my individual commits... > git show c126721 > # ...see the trivial addition of Type(MPI_Message) into the annex... > > 4. Or, if you don't like/know git, the individual patch files (numbered in order) are also available. For example: > > See a list of all the individual patches at > http://www.open-mpi.org/~jsquyres/mpi-forum/individual-patches/ > > See an individual patch: > wget http://www.open-mpi.org/~jsquyres/mpi-forum/individual-patches/0033-Add-TYP… > > -- > Jeff Squyres > jsquyres(a)cisco.com > For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/ > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0

Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by N.M. Maclaren 15 Jun '14

15 Jun '14

On Jun 14 2014, William Gropp wrote: > > My feeling on this is (a) there have already been too many oversights > because of the tremendous scope of the changes and (b) the mpif.h > interface, in practice, should and will be used mostly for legacy > applications - thus either defining it as similar to the original "basic" > support, with any new constants, will satisfy the users without the risk > of finding another MPI_SIZEOF issue. At this point I have no confidence > that there isn't something else lurking in the wings, and that argues for > as conservative an approach as possible. In my view, that is the only reasonable approach. Sane people use it only in 'legacy' code or because the code is extremely defensive. In neither case are major new facilities relevant. Regards, Nick Maclaren. Regards, Nick Maclaren.

1 0

Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by William Gropp 14 Jun '14

14 Jun '14

My feeling on this is (a) there have already been too many oversights because of the tremendous scope of the changes and (b) the mpif.h interface, in practice, should and will be used mostly for legacy applications - thus either defining it as similar to the original "basic" support, with any new constants, will satisfy the users without the risk of finding another MPI_SIZEOF issue. At this point I have no confidence that there isn't something else lurking in the wings, and that argues for as conservative an approach as possible. Bill William Gropp Director, Parallel Computing Institute Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Jun 14, 2014, at 12:49 PM, Rolf Rabenseifner wrote: > Yes, the "Extended Support" listed of about 5 additional routines, > but as far as I understand, MPI_SIZEOF is the only one > that needs an INTERFACE description in mpif.h. > > All other routines are without any implementation-problems. > > Therefore, it would be enough to allow in the special case > that MODULE (and INTERFACE) are not supported by the target compiler, > that > - the Fortran Support requires only mpif.h (same as in MPI-2.0), > and > - MPI_SIZEOF is not provided. > > Additionally, we can restrict the MPI_SIZEOF support only 23 dimensions > - in mpif.h > - or generally if TS29113 or other non-Standard methods > are not available, > supposed that "DOUBLE PRECISION" is the longest > Fortran type Name: > > 123456789012345678901234567890123456789012345678901234567890123456789012 > SUBROUTINE MPI_SIZEOF(x,n) > DOUBLE PRECISION x(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,*) > ! 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 > What did mpich and OpenMPI in the past with MPI_SIZEOF? > > Rolf > > ----- Original Message ----- >> From: "William Gropp" <wgropp(a)illinois.edu> >> To: "MPI-WG Fortran working group" <mpiwg-fortran(a)lists.mpi-forum.org> >> Sent: Saturday, June 14, 2014 6:36:35 AM >> Subject: Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question) >> >> >> Not quite true. See my email, which gives specific locations in the >> standard for the relevant text. Note also that MPI_SIZEOF is *not* >> the only function that is not included in the "basic" Fortran >> support - the routines to create a type based on precision and range >> are also not required for basic support. >> >> >> I think there are two options for the so-called Fortran 77 support. >> Continue to support the "basic" support, as defined in MPI-2, or >> delete it entirely. Anything else is likely to leave errors and >> mistakes. And while I agree with Jeff that its not good practice to >> remove something like F77 support at a point release, I believe that >> the language in the standard removed Fortran 77 with 3.0. We're >> just now recognizing that. >> >> >> Bill >> >> >> >> >> >> >> >> William Gropp >> Director, Parallel Computing Institute Thomas M. Siebel Chair in >> Computer Science >> >> >> University of Illinois Urbana-Champaign >> >> >> >> >> >> >> >> On Jun 13, 2014, at 10:32 AM, Rolf Rabenseifner wrote: >> >> >> 2) The requirement of having this function in mpif.h was changed: >> >> MPI-2.0 through 2.2 requiered MPI_SIZEOF in mpif.h >> "unless the target compiler does not support modules >> or KIND-parametrized types." >> >> This restriction was removed. >> >> >> _______________________________________________ >> mpiwg-fortran mailing list >> mpiwg-fortran(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran > > -- > 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) > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran

1 0

Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by Rolf Rabenseifner 14 Jun '14

14 Jun '14

Yes, the "Extended Support" listed of about 5 additional routines, but as far as I understand, MPI_SIZEOF is the only one that needs an INTERFACE description in mpif.h. All other routines are without any implementation-problems. Therefore, it would be enough to allow in the special case that MODULE (and INTERFACE) are not supported by the target compiler, that - the Fortran Support requires only mpif.h (same as in MPI-2.0), and - MPI_SIZEOF is not provided. Additionally, we can restrict the MPI_SIZEOF support only 23 dimensions - in mpif.h - or generally if TS29113 or other non-Standard methods are not available, supposed that "DOUBLE PRECISION" is the longest Fortran type Name: 123456789012345678901234567890123456789012345678901234567890123456789012 SUBROUTINE MPI_SIZEOF(x,n) DOUBLE PRECISION x(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,*) ! 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 What did mpich and OpenMPI in the past with MPI_SIZEOF? Rolf ----- Original Message ----- > From: "William Gropp" <wgropp(a)illinois.edu> > To: "MPI-WG Fortran working group" <mpiwg-fortran(a)lists.mpi-forum.org> > Sent: Saturday, June 14, 2014 6:36:35 AM > Subject: Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question) > > > Not quite true. See my email, which gives specific locations in the > standard for the relevant text. Note also that MPI_SIZEOF is *not* > the only function that is not included in the "basic" Fortran > support - the routines to create a type based on precision and range > are also not required for basic support. > > > I think there are two options for the so-called Fortran 77 support. > Continue to support the "basic" support, as defined in MPI-2, or > delete it entirely. Anything else is likely to leave errors and > mistakes. And while I agree with Jeff that its not good practice to > remove something like F77 support at a point release, I believe that > the language in the standard removed Fortran 77 with 3.0. We're > just now recognizing that. > > > Bill > > > > > > > > William Gropp > Director, Parallel Computing Institute Thomas M. Siebel Chair in > Computer Science > > > University of Illinois Urbana-Champaign > > > > > > > > On Jun 13, 2014, at 10:32 AM, Rolf Rabenseifner wrote: > > > 2) The requirement of having this function in mpif.h was changed: > > MPI-2.0 through 2.2 requiered MPI_SIZEOF in mpif.h > "unless the target compiler does not support modules > or KIND-parametrized types." > > This restriction was removed. > > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran -- 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: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by Rolf Rabenseifner 14 Jun '14

14 Jun '14

Bill, Yes, you are right, I oversaw the advice to users; my apologies. Rolf ----- Original Message ----- > From: "William Gropp" <wgropp(a)illinois.edu> > To: "MPI-WG Fortran working group" <mpiwg-fortran(a)lists.mpi-forum.org> > Sent: Saturday, June 14, 2014 6:31:39 AM > Subject: Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question) > > > > > But the advice to users that follows it specifically describes what > happens when an array, rather than a scalar, is passed, and this > advice appears at the time of the original definition, in MPI 2.0. > From that, it is clear that the *intent* of the Forum was to permit > arrays to be passed to MPI_SIZEOF; otherwise, this text would never > have been written. > > > So I'll agree that the interface hasn't changed - it has *always* > permitted an array as an argument. > > > Bill > > > > > > > William Gropp > Director, Parallel Computing Institute Thomas M. Siebel Chair in > Computer Science > > > University of Illinois Urbana-Champaign > > > > > > > > On Jun 13, 2014, at 10:32 AM, Rolf Rabenseifner wrote: > > > Back to MPI_SIZEOF in mpif.h: > > A few clarifications: > > 1) The interface of Argument x was not changed between > MPI-2.0, 2.1, 2.2 and MPI-3.0: > It is always "<type> x" for mpif.h, i.e., without a dimension. > > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran -- 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: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by William Gropp 13 Jun '14

13 Jun '14

Not quite true. See my email, which gives specific locations in the standard for the relevant text. Note also that MPI_SIZEOF is *not* the only function that is not included in the "basic" Fortran support - the routines to create a type based on precision and range are also not required for basic support. I think there are two options for the so-called Fortran 77 support. Continue to support the "basic" support, as defined in MPI-2, or delete it entirely. Anything else is likely to leave errors and mistakes. And while I agree with Jeff that its not good practice to remove something like F77 support at a point release, I believe that the language in the standard removed Fortran 77 with 3.0. We're just now recognizing that. Bill William Gropp Director, Parallel Computing Institute Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Jun 13, 2014, at 10:32 AM, Rolf Rabenseifner wrote: > 2) The requirement of having this function in mpif.h was changed: > > MPI-2.0 through 2.2 requiered MPI_SIZEOF in mpif.h > "unless the target compiler does not support modules > or KIND-parametrized types." > > This restriction was removed.

1 0

Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by William Gropp 13 Jun '14

13 Jun '14

But the advice to users that follows it specifically describes what happens when an array, rather than a scalar, is passed, and this advice appears at the time of the original definition, in MPI 2.0. From that, it is clear that the *intent* of the Forum was to permit arrays to be passed to MPI_SIZEOF; otherwise, this text would never have been written. So I'll agree that the interface hasn't changed - it has *always* permitted an array as an argument. Bill William Gropp Director, Parallel Computing Institute Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Jun 13, 2014, at 10:32 AM, Rolf Rabenseifner wrote: > Back to MPI_SIZEOF in mpif.h: > > A few clarifications: > > 1) The interface of Argument x was not changed between > MPI-2.0, 2.1, 2.2 and MPI-3.0: > It is always "<type> x" for mpif.h, i.e., without a dimension.

1 0

Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question)
by Bill Long 13 Jun '14

13 Jun '14

On Jun 13, 2014, at 3:32 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > > 2) The requirement of having this function in mpif.h was changed: > > MPI-2.0 through 2.2 requiered MPI_SIZEOF in mpif.h > "unless the target compiler does not support modules > or KIND-parametrized types." > > This restriction was removed. On the surface that is reasonable, since 1) I don’t see how support for modules is relevant to mpif.h anyway, and 2) It’s a stretch to imagine there are any current compilers that do not support modules and parameterized types. > > How to solve these problems? > > I would expect that we should add two errata: > > A) MPI_SIZEOF needs not to be interceptable, > the idea of ticket #428. Definitely. > > B) The Fortran 77 rule on p609:47 - p610:2 > should be extended to allow mpif.h with > all MPI-3.0 routines that are defined with an mpif.h > interface, but with the exception of MPI_SIZEOF, > because MPI_SIZEOF is the only routine that requires > (several overloaded) INTERFACE specifications, > which did not came before Fortran 90. I’d prefer to just delete the "For Fortran 77 with some extensions:” part and start with "Fortran 90 with some extensions” and move the extension "An MPI subroutine with a choice argument may be called with different argument types.” to the Fortran 90 block. Separately, the initial sentence of 17.1.6 reads; "This section describes which Fortran interface functionality can be provided for different versions of the Fortran standard.” This is technically flawed in that there are no “different versions” of the Fortran standard. There is only one version at any given time. This could be corrected by changing “different” to a more accurate word like deleted, cancelled, obsolete, legacy, …, and change the end of the sentence from “the Fortran standard” to “Fortran as well as standard Fortran”. Resulting in something like: "This section describes which Fortran interface functionality can be provided for obsolete versions of Fortran as well as standard Fortran." Cheers, Bill > > Rolf > > ----- Original Message ----- >> From: "Jeff Squyres (jsquyres)" <jsquyres(a)cisco.com> >> To: "MPI-WG Fortran working group" <mpiwg-fortran(a)lists.mpi-forum.org> >> Sent: Thursday, June 12, 2014 9:19:27 PM >> Subject: Re: [MPIWG Fortran] MPI_SIZEOF: Errata required (was: Another MPI_SIZEOF question) >> >> On Jun 11, 2014, at 4:25 PM, Bill Long <longb(a)cray.com> wrote: >> >>>> Actually, this doesn't work, either. If you have an mpif.h that >>>> contains an "INTERFACE MPI_Sizeof", the compiler will treat it >>>> differently depending on whether it is included from source.f or >>>> source.f90 (and probably other variants, too). Gaahhh! >>> >>> Not sure why you say this. Unless the lines get too long because >>> of long names, you can write an include file so that it works for >>> both fixed and free source forms. The assumed source form is the >>> only distinction to the compiler between source files that end in >>> .f and .f90. >> >> Argh -- my Fortran ignorance rears its ugly head. Yes, indenting >> properly makes that problem go away. >> >> It's a tight squeeze, but the longest interface I can find is with >> COMPLEX(REAL128) and DIMENSION for 15 ranks; it barely fits into 72 >> columns: >> >> COMPLEX(REAL128),DIMENSION(1,1,1,1,1,1,1,1,1,1,1,1,1,1,*)::x >> >> Quite ugly, of course. But with a perl script to generate all the >> possibilities (for maximum portability), it works. Yuck. >> >>> As you might recall from earlier conference calls, I prefer that >>> nothing be added to mpif.h that was not there before. None of the >>> new functions or constants. It should be frozen at MPI 2.2 for >>> compatibility with legacy codes. If programmers want to use new >>> features, they have to modify their sources anyway, and should >>> change to using a module at the same time. >> >> There was unfortunately pushback from the Forum on this issue; we >> lost that battle. >> >> -- >> Jeff Squyres >> jsquyres(a)cisco.com >> For corporate legal information go to: >> http://www.cisco.com/web/about/doing_business/legal/cri/ >> >> _______________________________________________ >> mpiwg-fortran mailing list >> mpiwg-fortran(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran >> > > -- > 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) > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran Bill Long longb(a)cray.com Fortran Technical Suport & voice: 651-605-9024 Bioinformatics Software Development fax: 651-605-9142 Cray Inc./ Cray Plaza, Suite 210/ 380 Jackson St./ St. Paul, MN 55101

1 0