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

Re: [MPIWG Fortran] Deprecate MPI_SIZEOF in MPI-4.0
by Jeff Squyres (jsquyres) 15 Apr '16

15 Apr '16

Re-sending because the reply-to on the original mail was incorrect (i.e., if you replied and didn't hand-edit the reply address, it likely bounced). > On Apr 15, 2016, at 11:57 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > > We brought this up forever ago (December 2014), and never progressed the issue (IIRC). > > It came up again today in a different context (in the Open MPI community). So I took the liberty of re-creating the issue in the Github MPI Forum system: > > https://github.com/mpi-forum/mpi-issues/issues/51 > > We still want to move forward with this, right? > > -- > 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)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] Deprecate MPI_SIZEOF in MPI-4.0
by Jeff Hammond 15 Apr '16

15 Apr '16

Yes. I migrated a lot of tickets but this wasn't one of them. Sent from my iPhone > On Apr 15, 2016, at 5:57 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > > We brought this up forever ago (December 2014), and never progressed the issue (IIRC). > > It came up again today in a different context (in the Open MPI community). So I took the liberty of re-creating the issue in the Github MPI Forum system: > > https://github.com/mpi-forum/mpi-issues/issues/51 > > We still want to move forward with this, right? > > -- > 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)mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran

1 0

Deprecate MPI_SIZEOF in MPI-4.0
by Jeff Squyres (jsquyres) 15 Apr '16

15 Apr '16

We brought this up forever ago (December 2014), and never progressed the issue (IIRC). It came up again today in a different context (in the Open MPI community). So I took the liberty of re-creating the issue in the Github MPI Forum system: https://github.com/mpi-forum/mpi-issues/issues/51 We still want to move forward with this, right? -- 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] remove HPF
by Bill Long 03 Mar '16

03 Mar '16

On Mar 1, 2016, at 2:28 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > > On Tue, Mar 1, 2016 at 2:12 PM, Bill Long <longb(a)cray.com> wrote: > > On Mar 1, 2016, at 2:03 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > > https://github.com/mpi-forum/mpi-issues/issues/43 > > > > any objections? > > Not from me. It seems like an excellent idea. > > > The only question is if MPI_Type_create_darray makes sense to be used with Fortran coarrays for MPI file I/O, but this is not something I can reason about, because I do not understand MPI I/O. Also, it is not clear to me that MPI_Type_create_darray is really what people want to use, given we have MPI_Type_subarray. I’m even less of an expert about MPI I/O. However, it would seem very unlikely that a coarray would be treated any different from an ordinary array. A coarray is just an ordinary array with the extra feature of being accessible from a different image. On the local image, it is just a ordinary array. Since I really doubt that MPI I/O is using coarray features inside, it seems unlikely that coarrays would be special in any way for MPI I/O. For defining coarray elements by reading a file, I would not use MPI I/O at all. Just have each image read in its part of the file - time tested solution. For writing, the array is memory is not modified, so even if MPI I/O is used, I would not see an issue of the output item were a coarray. Cheers, Bill > > Jeff > > Cheers, > Bill > > > > > > 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 > > Bill Long longb(a)cray.com > Fortran Technical Support & voice: 651-605-9024 > Bioinformatics Software Development fax: 651-605-9142 > Cray Inc./ Cray Plaza, Suite 210/ 380 Jackson St./ St. Paul, MN 55101 > > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran > > > > -- > 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 Bill Long longb(a)cray.com Fortran Technical Support & 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

Re: [MPIWG Fortran] remove HPF
by Jeff Hammond 01 Mar '16

01 Mar '16

On Tue, Mar 1, 2016 at 2:12 PM, Bill Long <longb(a)cray.com> wrote: > > On Mar 1, 2016, at 2:03 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > > > https://github.com/mpi-forum/mpi-issues/issues/43 > > > > any objections? > > Not from me. It seems like an excellent idea. > > The only question is if MPI_Type_create_darray makes sense to be used with Fortran coarrays for MPI file I/O, but this is not something I can reason about, because I do not understand MPI I/O. Also, it is not clear to me that MPI_Type_create_darray is really what people want to use, given we have MPI_Type_subarray. Jeff > Cheers, > Bill > > > > > > 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 > > Bill Long > longb(a)cray.com > Fortran Technical Support & voice: > 651-605-9024 > Bioinformatics Software Development fax: 651-605-9142 > Cray Inc./ Cray Plaza, Suite 210/ 380 Jackson St./ St. Paul, MN 55101 > > > _______________________________________________ > mpiwg-fortran mailing list > mpiwg-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran > -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [MPIWG Fortran] remove HPF
by Bill Long 01 Mar '16

01 Mar '16

On Mar 1, 2016, at 2:03 PM, Jeff Hammond <jeff.science(a)gmail.com> wrote: > https://github.com/mpi-forum/mpi-issues/issues/43 > > any objections? Not from me. It seems like an excellent idea. Cheers, Bill > > 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 Bill Long longb(a)cray.com Fortran Technical Support & 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

remove HPF
by Jeff Hammond 01 Mar '16

01 Mar '16

https://github.com/mpi-forum/mpi-issues/issues/43 any objections? jeff -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [MPIWG Fortran] Provide Fortran datatypes if Fortran bindings not provided?
by Bill Long 18 Feb '16

18 Feb '16

On Feb 18, 2016, at 4:09 PM, Craig Rasmusen <rasmus(a)cas.uoregon.edu> wrote: > >> (1) mpi.h does not provide a specific MPI Fortran datatype > > > I think JeffH and Jed want (1). > > > I think I want (1), too. > > I'm good with (1) too. This issue would not arise for Cray, but JeffH's argument to match the wording used for C++ is persuasive. So I am OK with (1) also. I agree with Rolf that the MPI spec should explicitly state the policy. Cheers, Bill > > -craig > > On Thu, Feb 18, 2016 at 1:31 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote: > On Feb 18, 2016, at 3:55 PM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > > > > I would prefer a solution where an application or middlware can check at > > runtime whether the Fortran MPI datatypes exist, i.e., I like mpich's > > solution. It is more simple than all methods that People must check prior > > to compile and must remove or Change parts of there software. > > The reason that JeffH and Jed hate this solution is because it makes it quite difficult for an application to know whether the MPI has Fortran support or not -- they basically have to compile an MPI application *and then run it* to see if the datatypes are actually MPI_DATATYPE_NULL. > > Doing this inside a GNU Autoconf/configure script is very, very difficult because of the huge differences in how you launch MPI applications across different HPC environments. > > Their (strong) preference is that you can have a simple compilation test in the GNU Autoconf/configure script that just tries to compile something like the following: > > ----- > #include <mpi.h> > MPI_Datatype foo = MPI_INTEGER; > ----- > > If that compiles, then your MPI has Fortran support. > If it doesn't compile, your MPI doesn't have Fortran support. > > > In the moment,in the case of a missing accompanying Fortran Compiler, > > an application or middleware that wants to handle > > MPI Fortran datatypes must handle both cases > > (1) mpi.h does not provide a specific MPI Fortran datatype > > (2) mpi.h provides this MPI Fortran datatype as MPI_DATATYPE_NULL > > > > The additional case, that it is mapped to a real datatype is not > > an exception, because it defines some sort of accompanying Fortran Compiler. > > > > I would prefer that the MPI standard defines, which of the two > > options should be implemented. > > I think JeffH and Jed want (1). > > I think I want (1), too. > > > As said at the beginning, I would prefer (1), > > - as mpich is doing it in general, and > > - OMPI is doing it partially, > > whereas (2) was not done by mpich nor by OMPI. > > I think you have (1) and (2) backwards in this paragraph. > > -- > 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 > > _______________________________________________ > 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 Support & 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

Re: [MPIWG Fortran] Provide Fortran datatypes if Fortran bindings not provided?
by Craig Rasmusen 18 Feb '16

18 Feb '16

>> (1) mpi.h does not provide a specific MPI Fortran datatype > I think JeffH and Jed want (1). > I think I want (1), too. I'm good with (1) too. -craig On Thu, Feb 18, 2016 at 1:31 PM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com > wrote: > On Feb 18, 2016, at 3:55 PM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> > wrote: > > > > I would prefer a solution where an application or middlware can check at > > runtime whether the Fortran MPI datatypes exist, i.e., I like mpich's > > solution. It is more simple than all methods that People must check prior > > to compile and must remove or Change parts of there software. > > The reason that JeffH and Jed hate this solution is because it makes it > quite difficult for an application to know whether the MPI has Fortran > support or not -- they basically have to compile an MPI application *and > then run it* to see if the datatypes are actually MPI_DATATYPE_NULL. > > Doing this inside a GNU Autoconf/configure script is very, very difficult > because of the huge differences in how you launch MPI applications across > different HPC environments. > > Their (strong) preference is that you can have a simple compilation test > in the GNU Autoconf/configure script that just tries to compile something > like the following: > > ----- > #include <mpi.h> > MPI_Datatype foo = MPI_INTEGER; > ----- > > If that compiles, then your MPI has Fortran support. > If it doesn't compile, your MPI doesn't have Fortran support. > > > In the moment,in the case of a missing accompanying Fortran Compiler, > > an application or middleware that wants to handle > > MPI Fortran datatypes must handle both cases > > (1) mpi.h does not provide a specific MPI Fortran datatype > > (2) mpi.h provides this MPI Fortran datatype as MPI_DATATYPE_NULL > > > > The additional case, that it is mapped to a real datatype is not > > an exception, because it defines some sort of accompanying Fortran > Compiler. > > > > I would prefer that the MPI standard defines, which of the two > > options should be implemented. > > I think JeffH and Jed want (1). > > I think I want (1), too. > > > As said at the beginning, I would prefer (1), > > - as mpich is doing it in general, and > > - OMPI is doing it partially, > > whereas (2) was not done by mpich nor by OMPI. > > I think you have (1) and (2) backwards in this paragraph. > > -- > 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 >

1 0

Re: [MPIWG Fortran] Provide Fortran datatypes if Fortran bindings not provided?
by Jeff Squyres (jsquyres) 18 Feb '16

18 Feb '16

On Feb 18, 2016, at 4:43 PM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > >> I think JeffH and Jed want (1). >> I think I want (1), too. > > I see the point. > I have not a strong opinion and would like to know, > what the other representatives in the Forum would like to see, > because > > (2) MPI_DATATYPE_NULL is what they mainly have today. The standard currently requires that the Fortran datatypes *exist* in mpi.h, no matter what (which seems to be in contradiction to what we stated for the C++ datatypes). That's why the implementations are currently this way. > (1) "not existing" would be a real change. Agreed. I'll ask Martin if there's plenary time to get a straw vote. -- Jeff Squyres jsquyres(a)cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

1 0