Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: [OMPI devel] 1.5.4 and 1.4.4 NEWS items
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2011-08-18 09:11:02


Please check through these; thanks!

1.5.4
-----

- Add support for the (as yet unreleased) Mellanox MXM transport.
- Add support for dynamic service levels (SLs) in the openib BTL.
- Fixed C++ bindings cosmetic/warnings issue with
  MPI::Comm::NULL_COPY_FN and MPI::Comm::NULL_DELETE_FN. Thanks to
  Júlio Hoffimann for identifying the issues.
- Also allow the word "slots" in rankfiles (i.e., not just "slot").
  (** also to appear in 1.4.4)
- Add Mellanox ConnectX 3 device IDs to the openib BTL defaults.
  (** also to appear in 1.4.4)
- Various FCA updates.
- Fix 32 bit SIGBUS errors on Solaris SPARC platforms.
- Add missing ARM assembly code files.
- Update to allow more than 128 entries in an appfile.
  (** also to appear in 1.4.4)
- Various VT updates and bug fixes.
- Update description of btl_openib_cq_size to be more accurate.
  (** also to appear in 1.4.4)
- Various assembly "clobber" fixes.
- Fix a hang in carto selection in obscure situations.
- Guard the inclusion of execinfo.h since not all platforms have it. Thanks
  to Aleksej Saushev for identifying this issue.
  (** also to appear in 1.4.4)
- Support Solaris legacy munmap prototype changes.
  (** also to appear in 1.4.4)
- Updated to Automake 1.11.1 per
  http://www.open-mpi.org/community/lists/devel/2011/07/9492.php.
- Fix compilation of LSF support.
- Update MPI_Comm_spawn_multiple.3 man page to reflect what it
  actually does.
- Fix for possible corruption of the environment. Thanks to Peter
  Thompson for the suggestion. (** also to appear in 1.4.4)
- Enable use of PSM on direct-launch SLURM jobs.
- Update paffinity hwloc to v1.2, and to fix minor bugs affinity
  assignment bugs on PPC64/Linux platforms.
- Let the openib BTL auto-detect its bandwidth.
- Support new MPI-2.2 datatypes.
- Updates to support more datatypes in MPI one-sided communication.
- Fix recursive locking bug when MPI-IO was used with
  MPI_THREAD_MULTIPLE. (** also to appear in 1.4.4)
- Fix mpirun handling of prefix conflicts.
- Ensure mpirun's --xterm options leaves sessions attached.
  (** also to appear in 1.4.4)
- Fixed type of sendcounts and displs in the "use mpi" F90 module.
  ABI is preserved, but applications may well be broken. See the
  README for more details. Thanks to Stanislav Sazykin for
  identifying the issue. (** also to appear in 1.4.4)
- Fix indexed datatype leaks. Thanks to Pascal Deveze for supplying
  the initial patch. (** also to appear in 1.4.4)
- Fix debugger mapping when mpirun's -npernode option is used.
- Fixed support for configure's --disable-dlopen option when used with
  "make distclean".
- Fix segv associated with MPI_Comm_create with MPI_GROUP_EMPTY.
  Thanks to Dominik Goeddeke for finding this.
  (** also to appear in 1.4.4)
- Improved LoadLeveler ORTE support.
- Add new WindVerbs BTL plugin, supporting native OpenFabrics verbs on
  Windows (the "wv" BTL).
- Add new btl_openib_gid_index MCA parameter to allow selecting which
  GID to use on an OpenFabrics device's GID table.
- Add support for PCI relaxed ordering in the OpenFabrics BTL (when
  available).
- Update rsh logic to allow correct SGE operation.
- Ensure that the mca_paffinity_alone MCA parameter only appears once
  in the ompi_info output. Thanks to Gus Correa for identifying the
  issue.
- Fixed return codes from MPI_PROBE and MPI_IPROBE.
  (** also to appear in 1.4.4)
- Remove --enable-progress-thread configure option; it doesn't work on
  the v1.5 branch. Rename --enable-mpi-threads to
  --enable-mpi-thread-multiple. Add new --enable-opal-multi-threads
  option.
- Updates for Intel Fortran compiler version 12.
- Remove bproc support. Farewell bproc!
- If something goes wrong during MPI_INIT, fix the error
  message to say that it's illegal to invoke MPI_INIT before
  MPI_INIT.

1.4.4
-----

- Ensure that MPI_Accumulate error return in 1.4 is consistent with
  1.5.x and trunk.
- Allow the word "slots" in rankfiles (i.e., not just "slot").
  (** also appeared in 1.5.4)
- Add Mellanox ConnectX 3 device IDs to the openib BTL defaults.
  (** also appeared in 1.5.4)
- Update description of btl_openib_cq_size to be more accurate.
- Ensure mpirun's --xterm options leaves sessions attached.
  (** also appeared in 1.5.4)
- Update to allow more than 128 entries in an appfile.
  (** also appeared in 1.5.4)
- Update description of btl_openib_cq_size to be more accurate.
  (** also appeared in 1.5.4)
- Fix for deadlock when handling recursive attribute keyval deletions
  (e.g., when using ROMIO with MPI_THREAD_MULTIPLE).
- Fix indexed datatype leaks. Thanks to Pascal Deveze for supplying
  the initial patch. (** also appeared in 1.5.4)
- Fixed the F90 types of the sendcounts and displs parameters to
  MPI_SCATTERV. Thanks to Stanislav Sazykin for identifying the issue.
  (** also appeared in 1.5.4)
- Exclude opal/libltdl from "make distclean" when --disable-dlopen is
  used. Thanks to David Gunter for reporting the issue.
- Fixed a segv in MPI_Comm_create when called with GROUP_EMPTY.
  Thanks to Dominik Goeddeke for finding this.
  (** also appeared in 1.5.4)
- Fixed return codes from MPI_PROBE and MPI_IPROBE.
  (** also appeared in 1.5.4)
- Fixed undefined symbol error when using the vtf90 profiling tool.
- Fix for referencing an uninitialized variable in DPM ORTE. Thanks
  to Avinash Malik for reporting the issue.
- Fix for correctly handling multi-token args when using debuggers.
- Eliminated the unneeded u_int*_t datatype definitions.
- Change in ORTE DPM to get around gcc 4.[45].x compiler wanrings
  about possibly calling free() on a non-heap variable, even though it
  will never happen because the refcount will never go to zero.
- Fixed incorrect text in MPI_File_set_view man page.
- Fix in MPI_Init_thread for checkpoint/restart.
- Fix for libtool issue when using pgcc to compile ompi in conjunction
  with the -tp option.
- Fixed a race condition in osc_rdma_sync. Thanks to Guillaume
  Thouvenin for finding this issue.
- Clarification of MPI_Init_thread man page.
- Fixed an indexing problem in precondition_transports.
- Fixed a problem in which duplicated libs were being specified for
  linking. Thanks to Hicham Mouline for noticing it.
- Various autogen.sh fixes.
- Fix for memchecking buffers during MPI_*INIT.
- Man page cleanups. Thanks to Jeremiah Willcock and Jed Brown.
- Fix for VT rpmbuild on RHEL5.
- Support Solaris legacy munmap prototype changes.
  (** also appeared in 1.5.4)
- Expands app_idx to int32_t to allow more than 127 app_contexts.
- Guard the inclusion of execinfo.h since not all platforms have it. Thanks
  to Aleksej Saushev for identifying this issue.
  (** also appeared in 1.5.4)
- Fix to avoid possible environment corruption. Thanks to Peter Thompson
  for identifying the issue and supplying a patch.
  (** also appeared in 1.5.4)
- Fixed paffinity base MCA duplicate registrations. Thanks to Gus
  Correa for bringing this to our attention.
- Fix recursive locking bug when MPI-IO was used with
  MPI_THREAD_MULTIPLE. (** also appeared in 1.5.4)
- F90 MPI API fixes.
- Fixed a misleading MPI_Bcast error message. Thanks to Jeremiah
  Willcock for reporting this.
- Added <sys/stat.h> to ptmalloc's hooks.c (it's not always included
  by default on some systems).
- Libtool patch to get around a build problem when using the IBM XL
  compilers.
- Fix to detect and avoid overlapping memcpy(). Thanks to
  Francis Pellegrini for identifying the issue.
- Fix to allow ompi to work on top of RoCE vLANs.
- Restored a missing debugger flag to support TotalView. Thanks to
  David Turner and the TV folks for supplying the fix.
- Updated SLURM support to 1.5.1.
- Removed an extraneous #include from the TCP BTL.
- When specifying OOB ports, fix to convert the ports into network
  byte order before binding.
- Fixed use of memory barriers in the SM BTL. This fixed segv's when
  compiling with Intel 10.0.025 or PGI 9.0-3.
- Fix to prevent the SM BTL from creating its mmap'd file in
  directories that are remotely mounted.

-- 
Jeff Squyres
jsquyres_at_[hidden]
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/