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

Reading Announcement
by Jim Dinan 19 Oct '23

19 Oct '23

Hi All, If time permits, the Hybrid WG would like to do an informal reading of the memory allocation kinds side document at the upcoming meeting. https://github.com/mpi-forum/mpi-issues/issues/809 Cheers, ~Jim.

1 0

Reading announcement for Reinit (slice 1)
by Ignacio Laguna 17 Oct '23

17 Oct '23

The MPI FT group would like to read the Coarse-grained fault-tolerance interface (Reinit, slice 1). This is for MPI v5: Issue: https://github.com/mpi-forum/mpi-issues/issues/578 PR: https://github.com/mpi-forum/mpi-standard/pull/711 I recommend reading the PDF chapter posted at the top of the issue: https://github.com/mpi-forum/mpi-issues/files/12703802/reinit.pdf Right now it's a separate chapter, but once approved we plan to integrate it in the already voted-in FT chapter (from ULFM slice 1). -- Ignacio Laguna Center for Applied Scientific Computing (CASC) Lawrence Livermore National Laboratory Phone: 925-422-7308, Fax: 925-422-6287

1 0

MPI 4.1 RC2 Document and Upcoming Meeting
by Wes Bland 17 Oct '23

17 Oct '23

Hi folks, Sorry about the delay on getting the new version of the 4.1 RC document out for the upcoming meeting. It was very helpful to have the last virtual meeting to make a better document that is more likely to be adopted at the next meeting. The newest version of the document is now posted on the docs page: https://www.mpi-forum.org/docs/ Direct link: https://www.mpi-forum.org/docs/mpi-4.1/mpi-41-rc2.pdf Please take a look and let us (Bill, Martin, and/or myself) know if you see any showstoppers. Anything that isn’t a complete blocker at this point will be deferred to either an errata or a future version of the standard if the proposed document is adopted at the upcoming meeting. Speaking of the upcoming meeting, the full agenda with all of the votes and readings is now posted on the website: https://www.mpi-forum.org/meetings/2023/10/agenda. If you feel that I missed anything, please let me know. The only thing that may change at this point is that some of the readings might be marked as getting an errata vote, but need to be changed into a no-no vote. Remember that everything that was on Martin’s list (https://docs.google.com/spreadsheets/d/1ZpsyVTnfNHv85K9Mzi_3niNXdmQ67yV2iwS…) will be voted on with errata rules and anything that isn’t will be voted on with no-no rules. I’ll go through the list soon to make sure I have them categorized correctly, but feel free to let me know if you see something wrong. The only things that are exempted from being voted on are the “editor changes” from Bill to fix whitespace, formatting, etc. You can find those by looking for any of Bill’s recent pull requests, but voting on those would be difficult as they are often needed after merging any other PR after it reflows the text. Today is also the deadline for officer nominations and any other readings for the October meeting. For anyone who is interested in being an officer, please have a friend nominate you and you can still be on the ballot. Lastly, please make sure you’ve registered for the upcoming meeting. The current list of registrants is here: https://www.mpi-forum.org/meetings/2023/10/attendance. Please double check that you’re on there and sign up here if not: https://forms.gle/tPvZNmT7ob7rNtqJ6 Thanks, Wes

1 0

Re: [Mpi-forum] MPI Forum Virtual Meeting Wednesday on MPI_Psend_init_c & MPI_Precv_init_c
by Martin Schulz 11 Oct '23

11 Oct '23

Hi all, Quick reminder – talk to you all in 8 minutes. 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 From: Martin Schulz <schulzm(a)in.tum.de> Date: Tuesday, 10. October 2023 at 02:05 To: Benson Muite via mpi-forum <mpi-forum(a)lists.mpi-forum.org> Subject: MPI Forum Virtual Meeting Wednesday on MPI_Psend_init_c & MPI_Precv_init_c Hi all, We are getting close to the final version of MPI 4.1, but there is still the ongoing discussion on what to do with the mistake/decision that we took on the count parameters in MPI_Psend_init_c & MPI_Precv_init_c. For those who have not been following this, here is the link to the github thread: https://github.com/mpi-forum/mpi-issues/issues/765 We need to settle this one way or the other, but a growing email thread does not seem to be the most effective way. Wesley and I are therefore suggesting to use our traditional meeting slot on Wednesday to hold a virtual meeting on this (Wednesday Oct. 11th, 10am US CT). Call-in information will be, as usual, on the forum website. I hope this works for most of you and hopefully we’ll find an easy way to resolve this. Thanks! Martin (and Wesley) -- 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

MPI Forum Virtual Meeting Wednesday on MPI_Psend_init_c & MPI_Precv_init_c
by Martin Schulz 09 Oct '23

09 Oct '23

Hi all, We are getting close to the final version of MPI 4.1, but there is still the ongoing discussion on what to do with the mistake/decision that we took on the count parameters in MPI_Psend_init_c & MPI_Precv_init_c. For those who have not been following this, here is the link to the github thread: https://github.com/mpi-forum/mpi-issues/issues/765 We need to settle this one way or the other, but a growing email thread does not seem to be the most effective way. Wesley and I are therefore suggesting to use our traditional meeting slot on Wednesday to hold a virtual meeting on this (Wednesday Oct. 11th, 10am US CT). Call-in information will be, as usual, on the forum website. I hope this works for most of you and hopefully we’ll find an easy way to resolve this. Thanks! Martin (and Wesley) -- 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

Reading announcement Dec '23 meeting
by Benson Muite 06 Oct '23

06 Oct '23

Have done an initial conversion of all figures to use TikZ. Would like to discuss these changes in the December meeting. Can also do chapter meetings as needed for chapters 2 (Terms), 6 (Collectives), 7 (Groups), 8 (Topologies), 9 (Environments), 11 (Processes), 12 (One-Sided), 14 (I/O) and 19 (Language Bindings). Issue: https://github.com/mpi-forum/mpi-issues/issues/712 Draft: https://github.com/mpi-forum/mpi-standard/pull/861 PDF with updated figures: https://github.com/mpi-forum/mpi-standard/files/12834910/mpi41-report.pdf Benson

1 0

October 2023 MPI Forum Meeting Registration
by Wes Bland 27 Sep '23

27 Sep '23

Hi all, Registration for the October (Final Ratification) meeting is now open. Please register ASAP before you forget about it. https://forms.gle/DYMyFaknKrTXszjB8 September 2023 MPI Forum Meeting Registration forms.gle This meeting will be fully remote and the meeting information will be in the usual place on the website under the logistics page: https://www.mpi-forum.org/meetings/2023/10/logistics Thanks, Wes

1 0

MPI Forum Officer Nominations & October Reading Announcements
by Wes Bland 26 Sep '23

26 Sep '23

Hi all, Another reminder that nominations for MPI Forum officers opened at the end of the last meeting and will stay open until October 17th. I’ll post them on the agenda page for the upcoming meeting as I get them. https://www.mpi-forum.org/meetings/2023/10/agenda Please send nominations to this mailing list. I’ll go ahead and point to the nominations that Jeff Hammond posted on the Slack channel for the current slate of officers to continue in their positions. Those have already been put on the website. If you’d like to familiarize yourself with the job descriptions, you can find them in the procedures doc (https://www.mpi-forum.org/docs/other/procedures-34.pdf) or reach out to the current office holders for more details. In addition, I’m going to help out a bit with reading announcements for the October meeting. Because there are going to be a large number of relatively small issues/PRs for this meeting and we’re dealing with them a little differently by merging them early, I’m going ahead and adding them to the agenda as they are merged. Therefore, if you have an issue for the MPI 4.1 RC, you don’t need to announce it as long as you see it on the agenda page above (and everyone should consider anything on that page as having been announced). If anyone has readings that aren’t based on MPI 4.1 fixes (things for 4.2 or 5.0), please announce them as usual. I would imagine that we’ll have time to do additional readings after we finish our MPI 4.1 business. Thanks, Wes

1 0

Re: [Mpi-forum] Complete RC open issues list from MPI Forum RCM
by Joseph Schuchart 21 Sep '23

21 Sep '23

Martin, all, I went through the issues grouped under the RMA chapter (Chapter 12) and found that some of them are issues in the I/O chapter: rows 228-259 (except for row 239) seem relevant for the I/O chapter and not the RMA chapter. I guess there has been some confusion about document and PDF numbers. I don't want to mess with the excel sheet but wanted to raise this here so that the I/O CC can take a look at these points. Cheers Joseph On 9/17/23 10:30, Martin Schulz via mpi-forum wrote: > > Hi all, > > First of all, thank all for a very productive and focused meeting last > week on our path to MPI 4.1 – both to those attending in person in > Bristol and to those online at often quite unfortunate hours! > > As discussed, I compiled a list of all open points and consolidated > this into a single spreadsheet, which is here: > > https://docs.google.com/spreadsheets/d/1ZpsyVTnfNHv85K9Mzi_3niNXdmQ67yV2iwS… > > There is a total of 380 points that we collected that we think should > be still addressed somehow. I placed them into five categories: > > * MPI 4.1: This is more or less the default category and should at > the end contain all small changes that we still want to apply the > standard (and which will be voted on). For all items in these > categories, we will need Pull requests and Issue created. > * Editor changes: Small changes and formatting changes that Bill > will apply directly – no separate PR or Issue is necessary. Once > an item is in this category, please do not touch it (unless Bill > pushes it back to one of the other categories) > * Flow: Items covering bad page breaks, strange spacing, etc., which > do not change the contents of the document and which need to be > addressed at the very end during layout – Bill will go over them > at the end and these items are mainly meant as hints for him to > pay attention to > * MPI 4.2: Items we think are too big and which should be pushed to > the next version of the standard or separate erratas. For those, > we will eventually also need PRs and Issues and we could > technically start reading them in the next meeting if there is time. > * Reject: items we decided to drop > > Note: this list all points individually as we recorded them – multiple > of them can of course (and should) be grouped into a single PR or Issue. > > ACTION ITEMS: > > * If you reviewed parts of the standard, please check the list and > make sure comments are properly recorded (if you noticed mistakes > in the page numbers, please let me know, as this means updates in > hidden parts of the spreadsheet) > * CC chairs: please take a look at all comments for your chapter and > ensure that they are consolidated correctly. By default you are > responsible for completing or shepherding all items in your chapter. > * If you have created PRs and Issues (or if you do so from now on), > please place them into the appropriate columns, so we can track it > and place yourself into the owner category. > * If the PR/Issue is complete and ready for review, please mark this > in the column “DONE”. > * If you think an MPI 4.1 issue should be moved to one of the other > categories, please move the checkmark and note the reason for the > decision in the “Comments” column at the end (together with your > name). > > As Wes noted on Friday, we have a short time line here – all items we > want for 4.1 must be in the DONE column by September 26. > > I hope this all makes sense and covers all steps forward – if you have > any questions, comments or concerns, please let me know. > > Thanks! > > Martin > > PS: This is a shared spreadsheet, so please be careful when editing > and please do not change anything structural. Also, mark any > decisions, observations or comments in the last column together with > your name, so we have a track record for the discussions next time. > Thanks! > > -- > > 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

Complete RC open issues list from MPI Forum RCM
by Martin Schulz 17 Sep '23

17 Sep '23

Hi all, First of all, thank all for a very productive and focused meeting last week on our path to MPI 4.1 – both to those attending in person in Bristol and to those online at often quite unfortunate hours! As discussed, I compiled a list of all open points and consolidated this into a single spreadsheet, which is here: https://docs.google.com/spreadsheets/d/1ZpsyVTnfNHv85K9Mzi_3niNXdmQ67yV2iwS… There is a total of 380 points that we collected that we think should be still addressed somehow. I placed them into five categories: MPI 4.1: This is more or less the default category and should at the end contain all small changes that we still want to apply the standard (and which will be voted on). For all items in these categories, we will need Pull requests and Issue created. Editor changes: Small changes and formatting changes that Bill will apply directly – no separate PR or Issue is necessary. Once an item is in this category, please do not touch it (unless Bill pushes it back to one of the other categories) Flow: Items covering bad page breaks, strange spacing, etc., which do not change the contents of the document and which need to be addressed at the very end during layout – Bill will go over them at the end and these items are mainly meant as hints for him to pay attention to MPI 4.2: Items we think are too big and which should be pushed to the next version of the standard or separate erratas. For those, we will eventually also need PRs and Issues and we could technically start reading them in the next meeting if there is time. Reject: items we decided to drop Note: this list all points individually as we recorded them – multiple of them can of course (and should) be grouped into a single PR or Issue. ACTION ITEMS: If you reviewed parts of the standard, please check the list and make sure comments are properly recorded (if you noticed mistakes in the page numbers, please let me know, as this means updates in hidden parts of the spreadsheet) CC chairs: please take a look at all comments for your chapter and ensure that they are consolidated correctly. By default you are responsible for completing or shepherding all items in your chapter. If you have created PRs and Issues (or if you do so from now on), please place them into the appropriate columns, so we can track it and place yourself into the owner category. If the PR/Issue is complete and ready for review, please mark this in the column “DONE”. If you think an MPI 4.1 issue should be moved to one of the other categories, please move the checkmark and note the reason for the decision in the “Comments” column at the end (together with your name). As Wes noted on Friday, we have a short time line here – all items we want for 4.1 must be in the DONE column by September 26. I hope this all makes sense and covers all steps forward – if you have any questions, comments or concerns, please let me know. Thanks! Martin PS: This is a shared spreadsheet, so please be careful when editing and please do not change anything structural. Also, mark any decisions, observations or comments in the last column together with your name, so we have a track record for the discussions next time. Thanks! -- 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