OMPI: Notified RMA ops Public API #9
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Matthew Whitlock <[email protected]>
Signed-off-by: Edgar Gabriel <[email protected]>
update VERSION file
Signed-off-by: Joseph Schuchart <[email protected]>
coll/ftagree: Memory bugfix
The switch from tree to bruck between 512 and 1023 processes leads to unexpected latency changes in benchmarks of other collectives. We should be consistent here. There is no good reason for why bruck would perform better in that range but not beyond. Signed-off-by: Joseph Schuchart <[email protected]>
…tead-of-bruck tuned: use tree instead of bruck at scale
Fix redundant declaration of MPI_Recv_c
if input index type is not fortran, only check for corresponding 'c' pair types. Related to open-mpi#13458 Checked against the test case in mpi4py that's activated when one declares one's MPI is 4.1 or higher compliant. Signed-off-by: Howard Pritchard <[email protected]>
…type_get_value_index make type_get_value_index language aware
Co-authored-by: Florent GERMAIN <[email protected]> Co-authored-by: Pierre LEMARINIER <[email protected]> Co-authored-by: Antoine CAPRA <[email protected]> Co-authored-by: Emmanuel BRELLE <[email protected]> Co-authored-by: Van Man NGUYEN <[email protected]> Co-authored-by: Julien DUPRAT <[email protected]> Co-authored-by: Tristan CALS <[email protected]> Co-authored-by: Anton DAUMEN <[email protected]> Co-authored-by: Alice CARIBONI <[email protected]> Co-authored-by: François WELLENREITER <[email protected]> Signed-off-by: Van Man NGUYEN <[email protected]>
This commit completely removes the mca/coll/hcoll component from the Open MPI source code. The hcoll component provided integration with Mellanox's Hierarchical Collectives library for collective operation offload. Changes include: - Removed ompi/mca/coll/hcoll component directory and all source files - Removed config/ompi_check_libhcoll.m4 configuration macro - Updated coll_tags.h to remove HCOLL tag space definitions - Updated coll_base_allgather.c and coll_base_allgatherv.c to use NEIGHBOR tag space instead of HCOLL tag space - Removed hcoll references from platform configuration files - Removed hcoll documentation and configuration options - Removed hcoll references from code comments Signed-off-by: Tomislav Janjusic <[email protected]>
coll: Remove hcoll component
related to open-mpi#13478 but without the controversial stuff. Signed-off-by: Howard Pritchard <[email protected]>
fix empty status fields
This reverts commit 9bd7757.
…ree-instead-of-bruck Revert "tuned: use tree instead of bruck at scale"
Signed-off-by: Jessie Yang <[email protected]>
ompi-prte: advance to sha 5ad79eb
Mac OS clang warns that sprintf is deprecated. Replace it with snprintf. Signed-off-by: Joseph Schuchart <[email protected]>
Signed-off-by: Joseph Schuchart <[email protected]>
Clang warns about possible uninitialized use. Signed-off-by: Joseph Schuchart <[email protected]>
Signed-off-by: Joseph Schuchart <[email protected]>
The expression `max_k < min(a, b)` may not produce what we think it does. Signed-off-by: Joseph Schuchart <[email protected]>
Replace sprintf with snprintf
UBCL: Add PML/UBCL and OSC/UBCL
The implementation of MPI_Group_from_session_pset was emitting a confusing error message when querying the PMIx server for members of a process group. The routine was using a method for creating an error message that was intended to be using during mpi initialization. Remove that statement and let an error code be returned. Related to open-mpi#13497 Signed-off-by: Howard Pritchard <[email protected]>
…_msg_for_group_from_session_pset group_from_session_pset - remove confusing error message
hdf5-2.0 was released, which was breaking our github workflow for hdf5. However, the new 2.0 release of hdf5 changed the build system to CMake instead of configure/make. Will investigate at a later stage how to adjust the CI workflow script for hdf5-2.0, for now just hard code the last release of hdf5-1.14.x Signed-off-by: Edgar Gabriel <[email protected]>
adjust hdf5 workflow script
joe-explr
commented
Dec 5, 2025
Signed-off-by: Brelle Emmanuel <[email protected]>
Signed-off-by: Brelle Emmanuel <[email protected]>
…ning and returns an error Signed-off-by: Brelle Emmanuel <[email protected]>
…d-returned-codes [osc/ubcl] Fixing bad returned codes
[UBCL] Fixed warnings to build with -Werror
Signed-off-by: Brelle Emmanuel <[email protected]>
Signed-off-by: Brelle Emmanuel <[email protected]>
Signed-off-by: Brelle Emmanuel <[email protected]>
update MPI I/O, network, section. Create a new directory for accelerator related stuff. Co-authored-by: Jeff Squyres <[email protected]> Signed-off-by: Edgar Gabriel <[email protected]>
…-for-6.0 Pass over the documentation tuning section
Signed-off-by: Roie Danino <[email protected]>
The sharedfp/lockedfile component had a temp. buffer of size 256 characters for generating a temp. file for testing the file locking feature during file open. For very long filenames, that was not sufficient. Switch to an asprintf based method instead of a fixed namelen to resolve the issue. Fixes issue open-mpi#13565 Signed-off-by: Edgar Gabriel <[email protected]>
…me-fix sharedfp/lockedfile: do not limit filename length
…fallback [UBCL] Ubcl endpoints fallback depending on UBCL library support
that were scattered around in the docs. Note that there are some places of historical interest where the ALPS nomenclature was retained. Signed-off-by: Howard Pritchard <[email protected]>
…yisms docs: cleanup some hpe/crayisms
Without this patch, one can successfully configure OMPI with fortran compiler flags that lead to compilation failures when building the mpi_f08 module. This can happen when the default integer size is the same as KIND=MPI_KIND_COUNT. Add a configure check to error out at that state with a hopefully useful error message: checking if Fortran compiler can compile interface containing MPI_KIND_COUNT arguments ... no configure: error: The MPI_F08 module cannot be compiled using configure Fortran commpiler options. Either change compiler options or use --enable-mpi-fortran=usempi Signed-off-by: Howard Pritchard <[email protected]>
release rather than hardwired to 5.0.(0,x). Signed-off-by: Howard Pritchard <[email protected]>
Signed-off-by: Howard Pritchard <[email protected]>
…_release_5 docs: generalize a couple of places for current
…not_mix MPI_F08: error out if fortran FCFLAGs do not work with big count
…es_for_6 docs: tweaks to developer docs for 6.0.x
related to open-mpi#13572 Signed-off-by: Howard Pritchard <[email protected]>
docs: add note about F08 big count and FCFLAGS
and squash some compiler warnings intel one-api cc was emitting Signed-off-by: Howard Pritchard <[email protected]>
ACCL/ZE: fix a missing symbols problem
OSHMEM/SHMEM/C: v1.5 - shmem_signal_fetch implementation
put_with_notify
get_with_notify
Signed-off-by: Joseph Antony <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
16 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary of Changes: