[OMPI devel] Incorrect minimum versions specified in Open-MPI's autogen.pl
Open MPI's autogen.pl has not had it's minimum tool version updated since sometime between v1.10 and v2.0.0. My colleague and I ran into this today when investigating incorrect CFLAGS being used. Jeff Squyres was able to root cause this as our use of old automake. Open MPI publishes it's minimum tool version needs here: https://www.open-mpi.org/source/building.php.To help give this issue more visibility (i.e. a fatal error at autogen.pl time, rather than incorrect CFLAGS (and who knows whatever other symptoms)), I've created PR 6837 (https://github.com/open-mpi/ompi/pull/6837/). All of the active release branches (v3.0.x, v3.1.x, and v4.0.x will also need this PR applied), if we take it.Since some of these tool's versions are still newer than what ships in common Linux distros, I wanted to communicate that this PR may cause some developers a level of pain. Please release the (very short) PR and discuss if this is the direction Open MPI Community would like to go. Thanks,Geoff Paulsen ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com ___ devel mailing list devel@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/devel
[OMPI devel] Anyone have any thoughts about cache-alignment issue in osc/sm?
Does anyone have any thoughts about the cache-alignment issue in osc/sm, reported in https://github.com/open-mpi/ompi/issues/6950? ___ devel mailing list devel@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/devel
[OMPI devel] Open MPI 4.0.2rc3 available for testing
The third (and possibly final) release candidate for the Open MPI v4.0.2 release is posted at https://www.open-mpi.org/software/ompi/v4.0/ Fixes since 4.0.2rc2 include: - Silent failure of OMPI over OFI with large messages sizes. - Conform MPIR_Breakpoint to MPIR standard. - btl/vader: when using single-copy emulation, fragment large rdma. - restore compilation of smpl/ikrit. Our goal is to release 4.0.2 in the next week. All testing and feedback is appreciated. Thanks, your Open MPI release team
[OMPI devel] Open MPI v4.0.3rc1 ready for testing.
Please test v4.0.3rc1: https://www.open-mpi.org/software/ompi/v4.0/ Changes since v4.0.2 include: 4.0.3 -- January, 2020 - Add support for Mellanox Connectx6. - Improve dimensions returned by MPI_Dims_create for certain cases. Thanks to @aw32 for reporting. - Fix an issue when sending messages larger than 4GB. Thanks to Philip Salzmann for reporting this issue. - Add ability to specify alternative module file path using Open MPI's RPM spec file. Thanks to @jschwartz-cray for reporting. - Clarify use of --with-hwloc configuration option in the README. Thanks to Marcin Mielniczuk for raising this documentation issue. - Fix an issue with shmem_atomic_set. Thanks to Sameh Sharkawi for reporting. - Fix a problem with MPI_Neighbor_alltoall(v,w) for cartesian communicators with cyclic boundary conditions. Thanks to Ralph Rabenseifner and Tony Skjellum for reporting. - Fix an issue using Open MPIO on 32 bit systems. Thanks to Orion Poplawski for reporting. - Fix an issue with NetCDF test deadlocking when using the vulcan Open MPIO component. Thanks to Orion Poplawski for reporting. - Fix an issue with the mpi_yield_when_idle parameter being ignored when set in the Open MPI MCA parameter configuration file. Thanks to @iassiour for reporting. - Address an issue with Open MPIO when writing/reading more than 2GB in an operation. Thanks to Richard Warren for reporting. ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com
[OMPI devel] v4.0.3rc3 ready for testing
Please test v4.0.3rc3: https://www.open-mpi.org/software/ompi/v4.0/ Changes since v4.0.2 include: 4.0.3 -- January, 2020 - Add support for Mellanox Connectx6.- Fix a problem with Fortran compiler wrappers ignoring use of disable-wrapper-runpath configure option. Thanks to David Shrader for reporting.- Fixed an issue with trying to use mpirun on systems where neither ssh nor rsh is installed.- Address some problems found when using XPMEM for intra-node message transport.- Improve dimensions returned by MPI_Dims_create for certain cases. Thanks to @aw32 for reporting.- Fix an issue when sending messages larger than 4GB. Thanks to Philip Salzmann for reporting this issue.- Add ability to specify alternative module file path using Open MPI's RPM spec file. Thanks to @jschwartz-cray for reporting.- Clarify use of --with-hwloc configuration option in the README. Thanks to Marcin Mielniczuk for raising this documentation issue.- Fix an issue with shmem_atomic_set. Thanks to Sameh Sharkawi for reporting.- Fix a problem with MPI_Neighbor_alltoall(v,w) for cartesian communicators with cyclic boundary conditions. Thanks to Ralph Rabenseifner and Tony Skjellum for reporting.- Fix an issue using Open MPIO on 32 bit systems. Thanks to Orion Poplawski for reporting.- Fix an issue with NetCDF test deadlocking when using the vulcan Open MPIO component. Thanks to Orion Poplawski for reporting.- Fix an issue with the mpi_yield_when_idle parameter being ignored when set in the Open MPI MCA parameter configuration file. Thanks to @iassiour for reporting.- Address an issue with Open MPIO when writing/reading more than 2GB in an operation. Thanks to Richard Warren for reporting. ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com
Re: [OMPI devel] v4.0.3rc3 ready for testing
Thanks so much for testing. If further testing reveals anything, please create an issue at https://github.com/open-mpi/ompi/. ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com - Original message -From: "Heinz, Michael William" To: Open MPI Developers Cc: Geoffrey Paulsen Subject: [EXTERNAL] RE: [OMPI devel] v4.0.3rc3 ready for testingDate: Fri, Jan 31, 2020 11:36 AM I’ve run the 3.1.6rc2 and 4.0.3rc3 src rpms through some smoke tests and they both built and ran properly on RHEL 8. From: devel On Behalf Of Geoffrey Paulsen via develSent: Wednesday, January 29, 2020 7:03 PMTo: devel@lists.open-mpi.orgCc: Geoffrey Paulsen Subject: [OMPI devel] v4.0.3rc3 ready for testing Please test v4.0.3rc3: https://www.open-mpi.org/software/ompi/v4.0/ Changes since v4.0.2 include: 4.0.3 -- January, 2020 - Add support for Mellanox Connectx6.- Fix a problem with Fortran compiler wrappers ignoring use of disable-wrapper-runpath configure option. Thanks to David Shrader for reporting.- Fixed an issue with trying to use mpirun on systems where neither ssh nor rsh is installed.- Address some problems found when using XPMEM for intra-node message transport.- Improve dimensions returned by MPI_Dims_create for certain cases. Thanks to @aw32 for reporting.- Fix an issue when sending messages larger than 4GB. Thanks to Philip Salzmann for reporting this issue.- Add ability to specify alternative module file path using Open MPI's RPM spec file. Thanks to @jschwartz-cray for reporting.- Clarify use of --with-hwloc configuration option in the README. Thanks to Marcin Mielniczuk for raising this documentation issue.- Fix an issue with shmem_atomic_set. Thanks to Sameh Sharkawi for reporting.- Fix a problem with MPI_Neighbor_alltoall(v,w) for cartesian communicators with cyclic boundary conditions. Thanks to Ralph Rabenseifner and Tony Skjellum for reporting.- Fix an issue using Open MPIO on 32 bit systems. Thanks to Orion Poplawski for reporting.- Fix an issue with NetCDF test deadlocking when using the vulcan Open MPIO component. Thanks to Orion Poplawski for reporting.- Fix an issue with the mpi_yield_when_idle parameter being ignored when set in the Open MPI MCA parameter configuration file. Thanks to @iassiour for reporting.- Address an issue with Open MPIO when writing/reading more than 2GB in an operation. Thanks to Richard Warren for reporting. ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com
[OMPI devel] Please test Open MPI v4.0.3rc4
Open MPI v4.0.3rc4 has been posted to https://www.open-mpi.org/software/ompi/v4.0/. Please test this on your systems, as it's likely to become v4.0.3. 4.0.3 -- March, 2020 --- - Update embedded PMIx to 3.1.5 - Add support for Mellanox ConnectX-6. - Fix an issue in OpenMPI IO when using shared file pointers. Thanks to Romain Hild for reporting. - Fix a problem with Open MPI using a previously installed Fortran mpi module during compilation. Thanks to Marcin Mielniczuk for reporting - Fix a problem with Fortran compiler wrappers ignoring use of disable-wrapper-runpath configure option. Thanks to David Shrader for reporting. - Fixed an issue with trying to use mpirun on systems where neither ssh nor rsh is installed. - Address some problems found when using XPMEM for intra-node message transport. - Improve dimensions returned by MPI_Dims_create for certain cases. Thanks to @aw32 for reporting. - Fix an issue when sending messages larger than 4GB. Thanks to Philip Salzmann for reporting this issue. - Add ability to specify alternative module file path using Open MPI's RPM spec file. Thanks to @jschwartz-cray for reporting. - Clarify use of --with-hwloc configuration option in the README. Thanks to Marcin Mielniczuk for raising this documentation issue. - Fix an issue with shmem_atomic_set. Thanks to Sameh Sharkawi for reporting. - Fix a problem with MPI_Neighbor_alltoall(v,w) for cartesian communicators with cyclic boundary conditions. Thanks to Ralph Rabenseifner and Tony Skjellum for reporting. - Fix an issue using Open MPIO on 32 bit systems. Thanks to Orion Poplawski for reporting. - Fix an issue with NetCDF test deadlocking when using the vulcan Open MPIO component. Thanks to O
[OMPI devel] GitHub v4.0.2 tag is broken
Ben, Oops, looks like I may have pushed a v4.0.2 branch around March 10th. Fortunately the v4.0.2 tag is fine and unaltered. I've deleted the v4.0.2 branch. Thanks for bringing this to our attention. Geoff Paulsen
[OMPI devel] Open-MPI v5.0 branch date pushed back to May 14th
Open MPI Developers, At today's web-ex we've decided to push back the date for branching Open-MPI v5.0 from master to May 14th. We're still targeting June 30th as the release date (see v5.0.0 milestone: https://github.com/open-mpi/ompi/milestone/37). If possible, we're still interested in having all New Feature Pull Requests posted to https://github.com/open-mpi/ompi by this Thursday, to allow for ample time to review, discuss and possibly iterate on. If you find yourself with a few free cycles, please pickup an open pull request, chime in and review it! If you find yourself with some extra computer cycles, please head over to https://github.com/open-mpi/ompi-collectives-tuning, and gather some collective tuning data for your system. Amazon AWS is collating that data to update our collectives tuning values, and your input is necessary. Thanks, Geoff Paulsen
[OMPI devel] Announcing Open MPI v4.0.4rc2
Open MPI v4.0.4rc2 is now available for download and test at: sso_last: https://www.open-mpi.org/software/ompi/v4.0/Changes from v4.0.4rc1 include: view commit • OPAL/UCX: enabling new API provided by UCX view commit • event/external: Fix typo in LDFLAGS vs LIBS var before check view commit • Updating README to include WARNING about ABI break view commit • Add checks for libevent.so conflict with LSF view commit • Move from legacy -levent to recommended -levent_core view commit • Correct typo in mapping-too-low* help messages view commit • A slightly stronger check for LSF's libevent view commit • Fix LSF configure check for libevent conflict view commit • VERSION -> v4.0.4rc2 view commit • update NEWS for 4.0.4rc2 view commit • sys limits: fixed soft limit setting if it is less than hard limit ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com
[OMPI devel] Open MPI v4.0.5rc1 available for test.
Announcing Open-MPI v4.0.5rc1 available for download and test at https://www.open-mpi.org/software/ompi/v4.0/ Please test and send feedback to devel@lists.open-mpi.org Fixed in v4.0.5: When launching under SLURM's srun, Open MPI will honor SLURM's binding policy even if that would leave the processes unbound. 4.0.5 -- July, 2020 Disable binding of MPI processes to system resources by Open MPI if an application is launched using SLURM's srun command.- Disable building of the Fortran mpi_f08 module when configuring Open MPI with default 8 byte Fortran integer size. Thanks to @ahcien for reporting.- Fix a problem with mpirun when the --map-by option is used. Thanks to Wenbin Lyu for reporting.- Fix some issues with MPI one-sided operations uncovered using Global Arrays regression test-suite. Thanks to @bjpalmer for reporting.- Fix a problem with make check when using the PGI compiler. Thanks to Carl Ponder for reporting.- Fix a problem with MPI_FILE_READ_AT_ALL that could lead to application hangs under certain circumstances. Thanks to Scot Breitenfeld for reporting.- Fix a problem building C++ applications with newer versions of GCC. Thanks to Constantine Khrulev for reporting.
[OMPI devel] Announcing Open-MPI v4.0.5rc2 available for testing
Open MPI v4.0.5rc2 is now available for download and test at: https://www.open-mpi.org/software/ompi/v4.0/ Please test and give feedback soon. Thanks!The Open-MPI Team
[OMPI devel] Open MPI v4.0.6 rc1 available for testing.
The first release candidate for Open-MPI v4.0.6 rc1 is now available for testing:https://www.open-mpi.org/software/ompi/v4.0/ Some fixes include: - Update embedded PMIx to 3.2.2. This update addresses several MPI_COMM_SPAWN problems. - Fix a symbol name collision when using the Cray compiler to build Open SHMEM. Thanks to Pak Lui for reporting and fixing. - Correct an issue encountered when building Open MPI under OSX Big Sur. Thanks to FX Coudert for reporting. - Various fixes to the OFI MTL. - Fix an issue with allocation of sufficient memory for parsing long environment variable values. Thanks to @zrss for reporting. - Improve reproducibility of builds to assist Open MPI packages. Thanks to Bernhard Wiedmann for bringing this to our attention. ---Geoffrey PaulsenSoftware Engineer, IBM Spectrum MPIEmail: gpaul...@us.ibm.com
[OMPI devel] Open MPI v5.0.x branch created
Open MPI developers, We've created the Open MPI v5.0.x branch today, and are receiving bugfixes. Please cherry-pick any master PRs to v5.0.x once they've been merged to master. We're targeting an aggressive but achievable release date of May 15th. If you're in charge of your organization's CI tests, please enable for v5.0.x PRs. It may be a few days until all of our CI is enabled on v5.0.x. Thanks everyone for your continued commitment to Open MPI's success. Josh Ladd, Austen Lauria, and Geoff Paulsen - v5.0 RMs
[OMPI devel] Help us find good times for design discussions.
Open MPI developers, We have not been able to meet together face to face for quite some time. We'd like to schedule a few 2-hour blocks for detailed discussions on topics of interest. Please fill out https://doodle.com/poll/rd7szze3agmyq4m5?utm_source=poll&utm_medium=link, include your name and time blocks that might work for you. Also please add any agenda items to discuss on our wiki page here: https://github.com/open-mpi/ompi/wiki/Meeting-2021-07 Thanks, Geoff Paulsen
[OMPI devel] OMPI -> PRRTE MCA parameter changes in Open MPI v5.0
Open MPI developers, The v5.0 RM managers would like to solicit opinions to come to consensusaround our three independent MCA frameworks in Open MPI v5.0.x. As you know, Open MPI v5.0.x has abstracted the runtime away to usethe Open PMIx Reference Runtime Environment (PRRTE) implementation. In doing so, many of MCA paremeters are now different as they'reread in by PRRTE. The problem is that this affects some vary common components, forexample `--oversubscribe`, which is now "--map-by :OVERSUBSCRIBE".prterun prints a nice warning message, but that is not true for manymca parameters that may never get read in at ALL in Open MPI v5.0.xwhere they were accepted and used in earlier versions. This means that users will need to use both ompi_info and prte_infoto understand how to translate their MCA parameters, and furthermorethere is no safety net if an old parameter is not read in with thenew Open MPI v5.0.x releases. If someone is interested in writing an aliasing system that mighthelp users adopt Open MPI, that would be FANTASTIC. However, itdoesn't help educate users for the long term. For this reason, the Open MPI v5.0 Release Managers are recommendingthat we document this loudly on the website for Open MPI v5.0.0. What are other's thoughts? Lets discuss on the devel mailing list: Let's try to come to consensus by Nov 15th.Thanks,The Open MPI v5.0 Release Managers,Tommy, Austen, Geoff
[OMPI devel] Open MPI v5.0.0rc3 available for testing
Open MPI v5.0.0rc3 is now available for testing (https://www.open-mpi.org/software/ompi/v5.0/). Pease test and send feedback either to users at lists dot open dash mpi dot org or create an issue at https://github.com/open-mpi/ompi/issues/. See https://github.com/open-mpi/ompi/blob/v5.0.x/NEWS, for changes in rc3. Thank you, v5.0 Release Managers
[OMPI devel] Find a time in next few days to discuss Han/Adapt.
I’ve created a doodle here: https://doodle.com/meeting/participate/id/axGpPgre Please add yourself, and select times you’re available. We’ll decide at Noon Eastern time on Thursday and reply to this email with final (I think we can also update the doodle at that time to reflect the best time, so can check back there as well. Description: Discuss performance results of various platforms for HAN and Adapt collective components, with the aim of deciding if we can make them default on and at what priority. Required Reading: - Discussion of Making coll HAN and Adapt Default for v5.0.0: Issue #10347 - Improvements for coll HAN: Issue #10438 - older large message perf numbers: Issue #9062 - Discussed in Weekly Telcon last month, see OMPI Wiki for WeeklyTelcon_20220614 and WeeklyTelcon_2022062 -- Geoff Paulsen IBM Spectrum MPI Engineer He/His
Re: [OMPI devel] Find a time in next few days to discuss Han/Adapt.
Resending new Doodle (https://doodle.com/meeting/participate/id/e76ENl1d) with times next week. Sorry for the inconvenience. Please add yourself, and select times you’re available. We’ll decide on Tuesday’s Web-ex and reply to this email with final time / dial-in info. Description: Discuss performance results of various platforms for HAN and Adapt collective components, with the aim of deciding if we can make them default on and at what priority. Required Reading: - Discussion of Making coll HAN and Adapt Default for v5.0.0: Issue #10347 - Improvements for coll HAN: Issue #10438 - older large message perf numbers: Issue #9062 - Discussed in Weekly Telcon last month, see OMPI Wiki for WeeklyTelcon_20220614 and WeeklyTelcon_2022062 -- Geoff Paulsen IBM Spectrum MPI Engineer He/His
Re: [OMPI devel] Find a time in next few days to discuss Han/Adapt.
August 18th, from 2-3pm Central US time is the winner. Here’s Webex-info: Open MPI - Discuss HAN / Adapt performance Hosted by Geoff Paulsen https://ibm.webex.com/ibm/j.php?MTID=m330b3d97ef8828d8b7dae0fa7105d47e Thursday, Aug 18, 2022 2:00 pm | 1 hour | (UTC-05:00) Central Time (US & Canada) Meeting number: 145 206 0471 Password: puZSuEmc768 (78978362 from phones and video systems) Agenda: Discuss performance results of various platforms for HAN and Adapt collective components, with the aim of deciding if we can make them default on and at what priority. Required Reading: - Discussion of Making coll HAN and Adapt Default for v5.0.0: Issue #10347 - Improvements for coll HAN: Issue #10438 - older large message perf numbers: Issue #9062 - Discussed in Weekly Telcon last month, see OMPI Wiki for WeeklyTelcon_20220614 and WeeklyTelcon_2022062 Join by video system Dial 1452060...@ibm.webex.com You can also dial 173.243.2.68 and enter your meeting number. Join by phone 1-844-531-0958 United States Toll Free 1-669-234-1178 United States Toll Access code: 145 206 0471 -- Geoff Paulsen IBM Spectrum MPI Engineer He/His
[OMPI devel] Discuss HAN / Adapt performance (part 2)
Good meeting today on HAN / Adapt performance. Joseph is going to go run some more numbers. We will finalize our discussion next week. Same time, different Webex Info: Agenda: o Discuss Joseph's new results of HAN run with newly-tuned coll_tune. o Come to consensus about plans for v5.0.0 release. Discuss HAN / Adapt performance (part 2) Hosted by Geoff Paulsen https://ibm.webex.com/ibm/j.php?MTID=m69e2f10a6293a0b3a6eb81084779294c Thursday, Aug 25, 2022 2:00 pm | 1 hour | (UTC-05:00) Central Time (US & Canada) Meeting number: 145 719 0018 Password: 7VEiuBMMM63 (78348266 from phones and video systems) Agenda: Agenda: o Discuss Joseph's new results of HAN run with newly-tuned coll_tune. o Come to consensus about plans for v5.0.0 release. Join by video system Dial 1457190...@ibm.webex.com You can also dial 173.243.2.68 and enter your meeting number. Join by phone 1-844-531-0958 United States Toll Free 1-669-234-1178 United States Toll Access code: 145 719 0018 -- Geoff Paulsen IBM Spectrum MPI Engineer He/His
[OMPI devel] Open MPI v5.0.0 release timeline delay
Open MPI developers, I’ve got some bad news from a OMPI v5.0.0 release timeframe. IBM has asked Austen and I (and our team) to focus 100% on another project for the next two full weeks. Open MPI v5.0.x still has a few remaining blocking items including documentation, PRRTE 3.0 release, some collective performance data/marketing/messaging, along with a few platform specific bugs (see: https://github.com/open-mpi/ompi/projects/3) Due to these reasons (along with Super Computing and the holidays) the Open MPI v5.0 RMs feel that a January of 2023 is a more realistic timeframe for release. Thank you for your understanding. The Open MPI v5.0.x Release Managers: - Tomislav Janjusic, nVidia - Austen Lauria, IBM - Geoff Paulsen, IBM
[OMPI devel] Tomorrow (7/4/2023) Open-MPI developer call cancelled.
Tomorrow’s Open MPI call (7/4/2023) is cancelled due to the U.S. Holiday.
