[mpich-commits] [mpich] MPICH primary repository annotated tag, v3.1.1, created. v3.1.1

Service Account noreply at mpich.org
Wed Jun 4 16:34:39 CDT 2014


This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "MPICH primary repository".

The annotated tag, v3.1.1 has been created
        at  113a33db7868b1c4ca9bcdb7c9f162eb2f427a99 (tag)
   tagging  d39b7f534d72ca862fea1ec80894f5633bc8d926 (commit)
  replaces  v3.1
 tagged by  Huiwei Lu
        on  Wed Jun 4 15:01:20 2014 -0500

- Log -----------------------------------------------------------------
tagging 'v3.1.1'

Antonio J. Pena (8):
      Fixed configure when disabling f77/fc
      Fix compilation warning in bcast.c
      Linker warning multiple common of 'HYD_pmcd_pmip'
      Fixed obsolete string format conversion
      Fix incorrect declaration of string pointer
      Fix Intel compiler warnings in fdebug.c
      Compiler complains about uninitialized variable
      Fix compiler warning about missing fncn prototype

Bob Cernohous (1):
      Free the builtin communicators (world/self/iworld)

Haizhu Liu (1):
      Fix Memory leak due to dynamic task change

Huiwei Lu (10):
      Fixes failing electric-fence check
      Fixes freeing uninitialized memory
      Fixes inconsistent definition of parameters
      Adds configuration check for lock-free option
      Fixes test scripts in test/mpi
      Fixes default optimization flags
      Fixes MPI_Comm_idup
      Adds two tests to MPI_Comm_idup
      Fixes building MPI_Comm_idup for single-thread case
      Update CHANGES for 3.1.1 release

James Dinan (1):
      Add RMA accumulate test with min/maxloc operator

Jeff Hammond (3):
      add CVAR to disable Abort message
      change default for abort msg suppression to false
      add Jed's cleanup w.r.t. silent abort

Junchao Zhang (19):
      Fix weak symbol redefinition error with Pathscale
      Put weak symbols in #ifndef MPICH_MPI_FROM_PMPI
      Rename Fortran binding directories.
      Add const& to Datatype of MPI::Op::Reduce_local
      Add an example f08 binding
      Remove support subarray in mpi_f08_constants.f90
      Add an almost complete F08 binding with dir re-arranged
      Add a script use_mpi_f08/buildiface
      Add a script use_mpi_f08/wrappers_c/buildiface
      Update the build system to include F08 binding
      Update .gitignore to skip autogenerated files
      Code cleanup
      Revise impl of MPI_IN_PLACE and MPI_BOTTOM
      Fix whole array access errs on assumed-size arrays
      Comment out c_funptr code due to Cray ftn bugs
      Add the missing mpi_c_interface.F90 in makefile
      Add basic (non)contiguous subarray support
      Add a bunch of F08 subarray tests
      Fix a perl script problem met with perl-5.8.8

Ken Raffenetti (22):
      use AC_SEARCH_LIBS for additional libraries
      add back script to create html indexes for docs
      fix warning for ambiguous usage of perl localtime
      fix $wl setting for Solaris Studio compilers
      macro fix in datatype test
      add missing consts to MPI::Op methods
      add test for Op::Reduce_local
      avoid duplicate data in MPIR_proctable
      correct validation of MPI_Info handle
      fix incorrect usage of MPI_Abort in f90 tests
      move the F77 compiler tests later in configure
      cleanup .gitignore for src/env directory
      setup framework for Fortran 2008 tests
      convert f90 tests to f08
      update CHANGES for F08
      isolate F08 binding configure tests
      add missing header file to sendrect2 test
      nonblocking collective test fixes
      type matching issues in pt2pt tests
      moved shared memory configure option to m4 macro
      remove check for strtoll
      use --enable-new-dtags in compiler scripts

Lisandro Dalcin (3):
      Human readable names for MPI-3 C++ datatypes
      testcase for  MPI_Error_string()
      Deal with empty error class/code strings

Michael Blocksome (26):
      Uninitialized variable in pamid 'MPID_Win_test' function
      warning cleanup: pamid
      Do not force -Werror when -Wall is specified
      fix comments in pamid subconfigure.m4 file
      add support for powerpcle
      pamid: mpid_win_* fixes
      pamid: support downlevel pami 'is_local_task' extension
      Document the 'MPIFCLIBNAME' configure environment variable
      pamid: 'gracefully degrade' instead of error out when node info is not available
      pamid: fix bgq 'is_local_task' extension bug.
      pamid: implement comm_split_type hook.
      pamid: detect smp by default on bgq
      pamid: force context post on when using 'per-object' locks
      pamid: use shared memory configure option m4 macro
      pamid: remove unnecessary bgq specialization.
      pamid: change macro to not reference a local variable
      pamid: move pe-specific code to sysv function
      pamid: add a posix (mmap) shared memory impl. for bgq
      pamid: fix win_allocate_shared error handling
      pamid: fix win_shared_query
      pamid: fix request completion race condition
      pamid: use MPIDI_Win_shm_ctrl_t structure consistently
      pamid: calculate correct shared memory base address
      pamid: shared window fixes for mpi_get
      pamid: shared window fixes for mpi_get_accumulate
      Update mpich 3.1.1 RELEASE_NOTES for Blue Gene/Q

Nysal Jan K.A (2):
      Fix vecblklen validation failure on power arch
      pamid: Allow message sizes greater than 4GB

Paul Coffman (18):
      subset peer-to-peer two phase
      balancecontig: topology-aware aggregator seleciton
      move ad_bg to ad_gpfs files and directories
      makefile and autoconf changes
      ad_bg to ad_gpfs major reorganization
      bg to gpfs new files
      remove ad_bgl and ad_bglockless directories
      pushed bluegene platform code to bg subdir and removed ad_bg_getsh.c and ad_bg_setsh.c
      further lockless removal/doc fixes
      fixed configuration but so gpfs-only build does not include bluegene files
      Replace ADIOI_GPFS_assert with ADIOI_Assert
      fix hintfile pocessing to initialize flag
      Balance-contig fix: wrong initial index
      ROMIO: Balance-contig bugfix
      Assert on non-zero gpfs_fcntl
      gpfs_find_access_for_ion is BGQ-specific
      fix to support gpfs hints and deferred open
      Blue Gene: one-hop aggregator placement

Pavan Balaji (53):
      Remove MPIX_ usage for program internal function names.
      Added a new test for Reduce_scatter_block in C++.
      Replace comm_free with comm_disconnect.
      Remove armci-mpi.
      Fix incorrect comments.
      Added copyright check script.
      Update Copyright version dates.
      Better checks for VA_ARGS.
      Remove more windows files.
      Revert aclocal_cc.m4 commits that tweak VA_ARGS.
      Updated the CHANGES file with Reduce_local fix.
      Incorrect placement of the MPIR_Comm_split_type_impl symbol.
      Rename mpich libraries.
      Use CPPFLAGS for preprocessor macros.
      Make binding automake conditional names consistent.
      Move symbols to correct libraries.
      Reorder library names in pkg-config.
      Use library dependencies.
      Update pkg-config with inter-library dependency changes.
      Updated the CHANGES file with library management changes.
      Merge libmpif77 and libmpifort.
      Add backward compatibility symlinks.
      Bug-fix: MPIR_CVAR_CH3_ODD_EVEN_CLIQUES was not working.
      Reorder libmpi and libpmpi builds.
      Added more information on mpiexec to the user guide.
      Add an explicit application dependency on the profiling library.
      Cleanup the symbol management to go into the right library.
      Create a dummy function that calls all ROMIO public symbols.
      Make extractcvars VPATH build safe.
      Cleanup mpicxx.h.in rebuild logic.
      White space cleanup.
      Remove MPICXX_NAME and MPICPP_NAME.
      Make install time prefix setting less hacky.
      Clean up some bad writeup in the README.
      Correct README for the library name changes in MPICH.
      Allow libmpl/opa to be merged into libmpi (or libpmpi).
      Added backward compatible links for mpl and opa.
      Remove unnecessary static linking code.
      Move mpif90 to mpifort and manf90.txt to manfort.txt.
      Add a symlink for mpif90 to maintain backward compatibility.
      Make "F77" an expert-level setting.
      Update documentation to reflect the change to --disable-fortran.
      Check for shared memory support capability in Comm_split_type.
      Update environment variables controlling the mpiexec port range.
      Fix the release script.
      Update version in preparation for the 3.1.1 release.
      Reenable --enable-f77 and --enable-fc options.
      Remove the --enable-f90 option.
      build system fixes for f08
      Fix ibsend failures with the pathscale compilers.
      Remove mpd.
      Added a CHANGES note about mpd's death.
      Revert "Update mpich 3.1.1 RELEASE_NOTES for Blue Gene/Q"

Rob Latham (93):
      corrects old text in the introductory comments.
      Clean up resources in error cases
      ROMIO portablity: cannot "switch" on datatypes
      Add missing pvfs2 headers to tarball
      panasas: some layout types not handled
      avoid ompi redefinition of MPI_MAX_DATAREP_STRING.
      Ompi re-sync: check for statfs function
      OpenMPI-resync: housekeeping: clean up MIN macro
      Romio portability: check for extended greqs
      OpenMPI resync: converted wrong buffer in external32
      OpenMPI resync: teach ROMIO about OpenMPI option
      OpenMPI re-sync: remove unneeded end-of-script stuff
      incorrect header guards
      fix PVFS2 compile error with hint processing
      request newer features for pread/pwrite
      better error handling for NULL filename
      Feature check for scalable resize
      some lustre parameters larger than signed int
      introduce missing function prototypes
      warning cleanup: clarify = and == in conditional
      warning fixup: clean up casting
      warning cleanup: non-void function explicit return
      warning fix: shadowed declarations
      warning fix: provide proper prototype
      fix fs detection when multiple fs exist
      dust off old Blue Gene timing infrastrucutre
      bg-timing: DO NOT MERGE WITH MASTER: time lockless
      use pwrite/pread instead of seek+write/read
      bluegene timing: condense into one set of timers
      remove uneeded barrier
      Allocate two-phase buffer outside write path
      remove extraneous locks in bluegene driver
      coll_perf buffer size too small
      Two-phase I/O with threaded write
      romio-timing: even more finer grained timing
      deferred open fixup: broadcast from correct root
      additional logging information
      option to read/write to /dev/null
      Promote blocksize to ADIOI_FileD struct
      additional broadcast in open for blocksize
      Rework "scalable flush" logic
      Significant simplification of ad_bg_open
      romio configure: if/fi unbalanced
      disable aio on BlueGene
      blksize_t not available under enable-strict
      clean up P2Pcontig warnings
      feature request for pread/pwrite
      unused variable cleanup
      include pthread.h for pthread routines
      romio pvfs2: bad use of memcpy
      romio pvfs2: const-ify interface
      romio pvfs2 warning fixes
      Missed a call site when renamed this function
      remove unused variables
      bring in missing header files for lseek and close
      A few items describing the reworked BlueGene driver
      Remove preprocessor defines no longer used
      Bug: Inverted logic for scalable resize check
      clean up stray declrations of GPFS_SUPER_MAGIC
      more comments for crazy ROMIO multi-platform stat
      Bugfix: detected any fs as gpfs
      Disable "create on one, open on all" for NTFS
      modify MPL_trdump to make a single fprintf call
      update errstring2 for MPICH test harness
      Revert "Mark indexperf as failing perf test"
      clean up unused variables in panasas hints
      tiny warning fix for panasas debugging routine
      update panasas routines for const-ification
      usleep available but not prototyped in strict mode
      slightly better checks for Panasas SDK
      Make concurrent write the default for Panasas
      turn datatype test into a function
      ROMIO dtype flattening: ignore zero-length counts
      test case for 'resized of resized'
      Deal with more errno values
      ROMIO: Consolidate errno processing
      Actually initialize debug event logging
      Test case for darray
      update darray_read for test suite conventions
      Got a bit carried away zapping zero-length blocks
      exercise a few more datatypes
      rename indexed_zeros
      zero-length blocklens fix for structs
      typo in ROMIO runtests
      romio+gpfs: experiment with gpfs_fcntl hints
      romio+gpfs: experiement with per-access hints
      an IO-node aware mode for gpfs hints
      Debugging code caused segfault
      make event elements array a bit bigger
      Free duplicated communicator in error path
      Don't build GPFS hints if headers not found
      revert old "removezeros" optimization
      fix "used uninitilized" compiler warning

Sameer Kumar (1):
      pamid: fixes for MPI_Win_allocate_shared on BG/Q.

Sameh Sharkawi (2):
      OOM running broadcast stress test w/ COLLECTIVE_SELECTION=yes
      PAMID: Initializing mpierrno before calling MPIDO_Barrier in MPIDI_Coll_comm_create

Sangmin Seo (1):
      Fix of integer variable size

Su Huang (8):
      MPICH test case linked_list_lockall hang in MPI_Win_flush
      pamid: datatype is not freed in MPID_Win_accumulate
      pamid: memory leak on handles for request based RMA operations
      pamid: fix a bug in MPID_Win_free
      pamid: mapped datatypes needed to be unmapped
      pamid: task 0 hang in MPI_Init() if MP_PRINTENV=yes
      pamid: mutiple fixes for RMA shared window
      pamid: fixed various RMA problems with shared window

Wei-keng Liao (1):
      ROMIO test case demonstrating indexed type bugs

Wesley Bland (6):
      Remove the use of MPIDI_TAG_UB
      Adds universe check to C spawn tests
      Add the universe check to the fortran spawn tests
      Makes sure that the status object is always defined
      Mark ft/isendalive as xfail
      Make handling of request cleanup more uniform

William Gropp (21):
      Add datatype test support from MPICH1
      Add dataalign test from MPICH1
      Add sendrecv2 and sendrecv4 test from MPICH-1
      Add some MPICH-1 tests to MPICH datatype tests
      Fix bug in marking unbuilt tests as failed
      Fix typo in comment in dataalign.c
      Correct CPP name
      Add style commands to generated files
      Use MPIR_MIN/MPIR_MAX instead of MIN/MAX
      Add style comments for coding style checker
      Added summary.junit.xml to clean list
      Mark indexperf as failing perf test
      Minor typo fix for datatype/dataloop code
      Fix incorrect use of status and buffer mgmt
      Add detail to error messages in darray-pack
      Add more datatype tests
      Address many of the perf problems in #1788
      Remove incorrect use of status in test
      Fix location of debugging statement
      Correct error in description
      Remove incomplete dataloop optimizations

-----------------------------------------------------------------------


hooks/post-receive
-- 
MPICH primary repository


More information about the commits mailing list