August 2021 - mpi-forum - mailman3.cels.anl.gov

Tools WG: Plenary time at Sep'21 meeting
by Marc-André Hermanns 30 Aug '21

30 Aug '21

Dear all, the Tools WG would like to get some plenary feedback on the following topics: Tools WG PR: Function Pointer Interception (QMPI) https://github.com/mpiwg-tools/mpi-standard/pull/10 Forum Issue 519: Add functions to check multiple request statuses at once https://github.com/mpi-forum/mpi-issues/issues/519 Forum 520: Add compound datatypes for large integers for MPI_(MIN|MAX)LOC https://github.com/mpi-forum/mpi-issues/issues/520 Cheers, Marc-Andre -- Dr. rer. nat. Marc-André Hermanns IT Center Group: High Performance Computing Division: Computational Science and Engineering RWTH Aachen University Seffenter Weg 23 52074 Aachen Phone: +49 241 80-24381 hermanns(a)itc.rwth-aachen.de www.itc.rwth-aachen.de Social Media Kanäle des IT Centers: https://blog.rwth-aachen.de/itc/ https://www.facebook.com/itcenterrwth https://www.linkedin.com/company/itcenterrwth https://twitter.com/ITCenterRWTH https://www.youtube.com/channel/UCKKDJJukeRwO0LP-ac8x8rQ

1 0

Re: [Mpi-forum] MPI Virtual Meeting / EuroMPI / MPI-Forum September Meeting
by Wesley Bland 30 Aug '21

30 Aug '21

Hi all, I want to remind everyone to register for next week’s meeting and/or conference. As of right now, we still don’t have enough orgs registered to meet quorum (need 25 IMOVE orgs and currently have 21). If you’re not sure if you registered, you can find the list of everyone I have here: https://www.mpi-forum.org/meetings/2021/09/attendance <https://www.mpi-forum.org/meetings/2021/09/attendance>. Thanks, Wes > On Aug 24, 2021, at 4:37 PM, Martin Schulz via mpi-forum <mpi-forum(a)lists.mpi-forum.org> wrote: > > Hi all, > > Just a few reminders: > > - We have our next MPI Forum Virtual Meeting tomorrow (Wednesday 8/25) at the usual time of 10am US CT (link on github). The agenda is to continue discussions on changes for MPI 4.1. In particular, Bill is planning to discuss a few items from his editorial pass(es). > > - EuroMPI registration is open and registration is free. As last year, the conference is virtual. I think we have an exciting program over the two days, September 7th and 8th (from 9am US CT to 1pm US CT), which will also bleed over to the MPI forum with RMA and Sessions discussions on Day 2. More information on https://www.eurompi21.lrz.de/. Please help spread the word. > > - The MPI-Forum on September 9th and 10th (also virtual, also 9am US CT to 1pm US CT) is getting close and this Thursday is the two week deadline for plenary announcements (readings, erratas, regular ballots). I already saw some going around - if there are any others, please post them to the email list. If there are any other items you would like to see on the agenda, please let me know as well. > > - We have a combined registration for both events (thanks to Wesley!), which is at https://forms.gle/vJ32yJqNxe2DCUxC7 - if you plan on attending either one, please register as soon as possible. > > Thanks! > > Martin > > > -- > Prof. Dr. Martin Schulz, Chair of Computer Architecture and Parallel Systems > Department of Informatics, TU-Munich, Boltzmannstraße 3, D-85748 Garching > Member of the Board of Directors at the Leibniz Supercomputing Centre (LRZ) > Email: schulzm(a)in.tum.de > > > > > > _______________________________________________ > mpi-forum mailing list > mpi-forum(a)lists.mpi-forum.org > https://lists.mpi-forum.org/mailman/listinfo/mpi-forum

1 0

Reading Announcement for "Synchronization on Partitioned Communication for Accelerator Optimization"
by Grant, Ryan Eric 25 Aug '21

25 Aug '21

Dear Forum Members, I would like to announce a reading at the upcoming MPI forum for Synchronization on Partitioned Communication for Accelerator Optimization, aka MPI_Pbuf_prepare: Ticket: https://github.com/mpi-forum/mpi-issues/issues/302 Pull request: https://github.com/mpi-forum/mpi-standard/pull/264 Semantics table addendum pull request: https://github.com/regrant/function-table/pull/1 Best Regards, Ryan

1 0

Errata reading announcement for Sept. 2021 Meeting
by Guillaume Mercier 25 Aug '21

25 Aug '21

Dear fellow Forum members, I'm announcing the reading of an errata, in order to correct a couple of examples in the partitioned communications chapter. Here is the issue: https://github.com/mpi-forum/mpi-issues/issues/509 And here is the related PR: https://github.com/mpi-forum/mpi-standard/pull/607 Thanks, Guillaume

1 0

Reading announcements for September 2021 MPI Forum
by Ruefenacht, Martin 25 Aug '21

25 Aug '21

Hi all, I am announcing two readings for the September 2021 MPI Forum. First, the errata to add the forgotten MPI_Status_set_elements_c: Issue: https://github.com/mpi-forum/mpi-issues/issues/517 PR: https://github.com/mpi-forum/mpi-standard/pull/620 Second the deprecation of _x procedures: Issue: https://github.com/mpi-forum/mpi-issues/issues/518 PR: https://github.com/mpi-forum/mpi-standard/pull/623 Thank you, Martin Ruefenacht

1 0

Reading/Vote and Plenary Announcements for the Fault Tolerance WG
by Wesley Bland 25 Aug '21

25 Aug '21

Hi all, I'd like to make a couple of announcements on behalf of the Fault Tolerance Working Group for the September meeting. First, I've made a bunch of edits based on the feedback at the last meeting for #283: Add MPI_Delete_error_class, MPI_Delete_error_code, MPI_Delete_error_string <https://github.com/mpi-forum/mpi-issues/issues/283> (PR #166 <https://github.com/mpi-forum/mpi-standard/pull/166>). The changes here could be considered no-no changes, so I would like to present it that way and continue with a first vote afterward. If there is pushback on that, I'm also ok with re-reading the full proposal (since it's pretty short) and calling it a reading with a first vote in December. Secondly, the Fault Tolerance Working Group would like to give an updated plenary about the state of the working group and what we've been working on and our future plans. There won't be any reading associated with this, but it will lead in to some readings we plan to do at future meetings. Thanks, Wes

1 0

Reading announcement for September 2021 MPI Forum
by Raffenetti, Kenneth J. 25 Aug '21

25 Aug '21

Hi all, Announcing the reading of issue #510 (errata): C Binding for MPI_CREATE_INFO_ENV Issue link: https://github.com/mpi-forum/mpi-issues/issues/510 PR link: https://github.com/mpi-forum/mpi-standard/pull/609 Thanks, Ken

1 0

Re: [Mpi-forum] [EXTERNAL] Re: An f2c question
by Pritchard Jr., Howard 24 Aug '21

24 Aug '21

Thanks Rolf! On 8/24/21, 3:26 PM, "Rolf Rabenseifner" <rabenseifner(a)hlrs.de> wrote: %IMPI_VAL was a typo. %MPI_VAL is correct, see MPI-4.0 page 18, line 20. ----- Original Message ----- > From: "Pritchard" <howardp(a)lanl.gov> > To: "Rolf Rabenseifner" <rabenseifner(a)hlrs.de>, "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> > Sent: Tuesday, August 24, 2021 10:55:51 PM > Subject: Re: [EXTERNAL] Re: [Mpi-forum] An f2c question > Hi Ralph, > > Okay, not quite right, looks like both mpich and open mpi use %MPI_VAL, so I'll > go with that. > > Thanks, > > Howard > >On 8/24/21, 2:28 PM, "Rolf Rabenseifner" <rabenseifner(a)hlrs.de> wrote: > > Dear Howard, > > SUBROUTINE MPI_TYPE_COMMIT(DATATYPE, IERR) > INTEGER :: IERR > ! some module USE to include the definition of TYPE(MPI_Datatype) > TYPE(MPI_Datatype) :: DATATYPE > CALL MPI_X_TYPE_COMMIT(DATATYPE%IMPI_VAL, IERR) > RETURN > END > > I do not know how to handle the optional ierr, but here is the way you should > handle the datatype handle. > > Best regards > Rolf > > ----- Original Message ----- > > From: "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> > > To: "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> > > Cc: "Pritchard" <howardp(a)lanl.gov> > > Sent: Tuesday, August 24, 2021 10:19:27 PM > > Subject: [Mpi-forum] An f2c question > > > Hi All, > > > > > > > > In porting an MPI application written mainly in Fortran and a little c to using > > mpi_f08 rather than mpi, I’ve hit an issue. > > > > > > > > Namely, how would I write example 19.11 in the MPI 4 standard if the app uses > > type(MPI_Type) rather than integer? > > > > Note I’m working with an application, so the MPI_Type structure is opaque, or at > > least I thought we’re supposed to treat it as opaque. > > > > > > > > Here’s the example: > > > > > > > > Example 19.11 The example below illustrates how the Fortran MPI function > > > > MPI_TYPE_COMMIT can be implemented by wrapping the C MPI function > > > > MPI_Type_commit with a C wrapper to do handle conversions. In this example a > > Fortran-C > > > > interface is assumed where a Fortran function is all upper case when referred to > > from C and > > > > arguments are passed by addresses. > > > > ! FORTRAN PROCEDURE > > > > SUBROUTINE MPI_TYPE_COMMIT(DATATYPE, IERR) > > > > INTEGER :: DATATYPE, IERR > > > > CALL MPI_X_TYPE_COMMIT(DATATYPE, IERR) > > > > RETURN > > > > END > > > > /* C wrapper */ > > > > void MPI_X_TYPE_COMMIT(MPI_Fint *f_handle, MPI_Fint *ierr) > > > > { > > > > MPI_Datatype datatype; > > > > datatype = MPI_Type_f2c(*f_handle); > > > > *ierr = (MPI_Fint)MPI_Type_commit(&datatype); > > > > *f_handle = MPI_Type_c2f(datatype); > > > > return; > > > > } > > > > > > > > Now I suspect that most of the time if one has some old ‘c’ code assuming a > > pointer to a 4 byte fortran integer argument for the datatype is just going to > > work, but I’d rather not just try to be lucky. > > > > Anyway, there are fortran gurus on this project and they’d be sure to want a > > better solution. > > > > > > > > Thanks for any help/ideas, > > > > > > > > Howard > > > > > > > > > > > > — > > > > > > > > > > > > > > Howard Pritchard > > > > Research Scientist > > > > HPC-ENV > > > > > > > > Los Alamos National Laboratory > > > > howardp(a)lanl.gov > > > > > > > > [ https://www.instagram.com/losalamosnatlab/ ] [ > > https://twitter.com/LosAlamosNatLab ] [ > > https://www.linkedin.com/company/los-alamos-national-laboratory/ ] [ > > https://www.facebook.com/LosAlamosNationalLab/ ] > > > > > > > > > > > > > > _______________________________________________ > > mpi-forum mailing list > > mpi-forum(a)lists.mpi-forum.org > > https://lists.mpi-forum.org/mailman/listinfo/mpi-forum > > -- > 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) . -- 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

MPI Virtual Meeting / EuroMPI / MPI-Forum September Meeting
by Martin Schulz 24 Aug '21

24 Aug '21

Hi all, Just a few reminders: - We have our next MPI Forum Virtual Meeting tomorrow (Wednesday 8/25) at the usual time of 10am US CT (link on github). The agenda is to continue discussions on changes for MPI 4.1. In particular, Bill is planning to discuss a few items from his editorial pass(es). - EuroMPI registration is open and registration is free. As last year, the conference is virtual. I think we have an exciting program over the two days, September 7th and 8th (from 9am US CT to 1pm US CT), which will also bleed over to the MPI forum with RMA and Sessions discussions on Day 2. More information on https://www.eurompi21.lrz.de/. Please help spread the word. - The MPI-Forum on September 9th and 10th (also virtual, also 9am US CT to 1pm US CT) is getting close and this Thursday is the two week deadline for plenary announcements (readings, erratas, regular ballots). I already saw some going around - if there are any others, please post them to the email list. If there are any other items you would like to see on the agenda, please let me know as well. - We have a combined registration for both events (thanks to Wesley!), which is at https://forms.gle/vJ32yJqNxe2DCUxC7 - if you plan on attending either one, please register as soon as possible. Thanks! Martin -- Prof. Dr. Martin Schulz, Chair of Computer Architecture and Parallel Systems Department of Informatics, TU-Munich, Boltzmannstraße 3, D-85748 Garching Member of the Board of Directors at the Leibniz Supercomputing Centre (LRZ) Email: schulzm(a)in.tum.de

1 0

Re: [Mpi-forum] [EXTERNAL] Re: An f2c question
by Rolf Rabenseifner 24 Aug '21

24 Aug '21

%IMPI_VAL was a typo. %MPI_VAL is correct, see MPI-4.0 page 18, line 20. ----- Original Message ----- > From: "Pritchard" <howardp(a)lanl.gov> > To: "Rolf Rabenseifner" <rabenseifner(a)hlrs.de>, "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> > Sent: Tuesday, August 24, 2021 10:55:51 PM > Subject: Re: [EXTERNAL] Re: [Mpi-forum] An f2c question > Hi Ralph, > > Okay, not quite right, looks like both mpich and open mpi use %MPI_VAL, so I'll > go with that. > > Thanks, > > Howard > >On 8/24/21, 2:28 PM, "Rolf Rabenseifner" <rabenseifner(a)hlrs.de> wrote: > > Dear Howard, > > SUBROUTINE MPI_TYPE_COMMIT(DATATYPE, IERR) > INTEGER :: IERR > ! some module USE to include the definition of TYPE(MPI_Datatype) > TYPE(MPI_Datatype) :: DATATYPE > CALL MPI_X_TYPE_COMMIT(DATATYPE%IMPI_VAL, IERR) > RETURN > END > > I do not know how to handle the optional ierr, but here is the way you should > handle the datatype handle. > > Best regards > Rolf > > ----- Original Message ----- > > From: "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> > > To: "Main MPI Forum mailing list" <mpi-forum(a)lists.mpi-forum.org> > > Cc: "Pritchard" <howardp(a)lanl.gov> > > Sent: Tuesday, August 24, 2021 10:19:27 PM > > Subject: [Mpi-forum] An f2c question > > > Hi All, > > > > > > > > In porting an MPI application written mainly in Fortran and a little c to using > > mpi_f08 rather than mpi, I’ve hit an issue. > > > > > > > > Namely, how would I write example 19.11 in the MPI 4 standard if the app uses > > type(MPI_Type) rather than integer? > > > > Note I’m working with an application, so the MPI_Type structure is opaque, or at > > least I thought we’re supposed to treat it as opaque. > > > > > > > > Here’s the example: > > > > > > > > Example 19.11 The example below illustrates how the Fortran MPI function > > > > MPI_TYPE_COMMIT can be implemented by wrapping the C MPI function > > > > MPI_Type_commit with a C wrapper to do handle conversions. In this example a > > Fortran-C > > > > interface is assumed where a Fortran function is all upper case when referred to > > from C and > > > > arguments are passed by addresses. > > > > ! FORTRAN PROCEDURE > > > > SUBROUTINE MPI_TYPE_COMMIT(DATATYPE, IERR) > > > > INTEGER :: DATATYPE, IERR > > > > CALL MPI_X_TYPE_COMMIT(DATATYPE, IERR) > > > > RETURN > > > > END > > > > /* C wrapper */ > > > > void MPI_X_TYPE_COMMIT(MPI_Fint *f_handle, MPI_Fint *ierr) > > > > { > > > > MPI_Datatype datatype; > > > > datatype = MPI_Type_f2c(*f_handle); > > > > *ierr = (MPI_Fint)MPI_Type_commit(&datatype); > > > > *f_handle = MPI_Type_c2f(datatype); > > > > return; > > > > } > > > > > > > > Now I suspect that most of the time if one has some old ‘c’ code assuming a > > pointer to a 4 byte fortran integer argument for the datatype is just going to > > work, but I’d rather not just try to be lucky. > > > > Anyway, there are fortran gurus on this project and they’d be sure to want a > > better solution. > > > > > > > > Thanks for any help/ideas, > > > > > > > > Howard > > > > > > > > > > > > — > > > > > > > > > > > > > > Howard Pritchard > > > > Research Scientist > > > > HPC-ENV > > > > > > > > Los Alamos National Laboratory > > > > howardp(a)lanl.gov > > > > > > > > [ https://www.instagram.com/losalamosnatlab/ ] [ > > https://twitter.com/LosAlamosNatLab ] [ > > https://www.linkedin.com/company/los-alamos-national-laboratory/ ] [ > > https://www.facebook.com/LosAlamosNationalLab/ ] > > > > > > > > > > > > > > _______________________________________________ > > mpi-forum mailing list > > mpi-forum(a)lists.mpi-forum.org > > https://lists.mpi-forum.org/mailman/listinfo/mpi-forum > > -- > 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) . -- 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