1. 23 Jul, 2013 1 commit
  2. 16 Jul, 2013 1 commit
  3. 08 Jul, 2013 1 commit
  4. 07 Jul, 2013 1 commit
  5. 11 Jun, 2013 1 commit
  6. 10 Jun, 2013 1 commit
    • Pavan Balaji's avatar
      Revert [6a4899dc]. · 40e4efb7
      Pavan Balaji authored
      The previous patch was meant to allow devices to automatically pick
      the default timer type they like.  However, it forces the device to
      pick one of the predefined timer types.
      40e4efb7
  7. 09 Jun, 2013 1 commit
  8. 08 Jun, 2013 2 commits
  9. 07 Jun, 2013 8 commits
  10. 22 May, 2013 1 commit
  11. 17 May, 2013 1 commit
  12. 06 May, 2013 7 commits
    • Michael Blocksome's avatar
      Allow f90 mod files to be installed in an arbitrary directory. · 1be61fbc
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      Instead of using 'includedir' to install the mod file headers the
      automake 'modincdir' is used instead. For example:
      
        ./configure --includedir=$some_dir modincdir=$some_dir/gnu
      
      This is needed because the fortran mod files are essentially
      pre-compiled headers and are therefore compiler-specific. This means
      that an application can not compile with xlf90 using a mpif90 script
      that was configured and installed using gfortran. By separating out the
      mod files generated by gfortran and xlf90 we can create compile scripts
      that use the appropriate mod files depending on the compiler used.
      
      Note: For some reason the normal configure processing does NOT check for
            a user-specified 'modincdir' variable and, if not set, default to
            use 'includedir'.
      
      (ibm) CPS 92BQ9Q
      (ibm) 52931bf950b389d49177a9aa94acf01c3583a8ce
      1be61fbc
    • Michael Blocksome's avatar
      Add '--with-mpl-prefix=DIR' configure option. · 3b8afffb
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      (ibm) CPS 92XMVJ
      (ibm) db1d16976caeaece627c157ad258f5c361ecf80e
      3b8afffb
    • Michael Blocksome's avatar
      Fix '--with-openpa-prefix=DIR' so it actually works. · c18320d3
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      (ibm) CPS 92XMVJ
      (ibm) 10855cb7c8e22e9018144ce14b986a3aa8f9d80f
      c18320d3
    • Michael Blocksome's avatar
      Allow pkgconfig files to be installed in an arbitrary directory. · 9164d3db
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      Instead of using 'libdir' to install the pkgconfig files the automake variable
      'pkgconfigdir' is used instead.  The configure processing checks for a
      user-specified 'pkgconfigdir' variable and, if not set, defaults to use
      'libdir' as before.
      
      (ibm) CPS 92XMVJ
      (ibm) 1a294c56db2e0d9405f14558d2943ea5ac8211a1
      9164d3db
    • Michael Blocksome's avatar
      Add ability to customize the MPL library name · 2bfe261f
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      To change the name from the default ("mpl") specify the environment
      variable "MPLLIBNAME=name" on configure.
      
      (ibm) CPS 92XMVJ
      (ibm) 27904f81cf1fbc5f3230adecd5317ddf5c29e800
      2bfe261f
    • Michael Blocksome's avatar
      Add ability to customize the OpenPA library name · 11702389
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      To change the name from the default ("opa") specify the environment
      variable "OPALIBNAME=name" on configure.
      
      (ibm) CPS 92XMVJ
      (ibm) cde8a03e9d2745a32901878b73b2b3047fa0f9d6
      11702389
    • Michael Blocksome's avatar
      Add ability to customize the mpich fortran 90 library name · 47ea4740
      Michael Blocksome authored and Pavan Balaji's avatar Pavan Balaji committed
      Previously, the f90 library name was forced to be:
      
        lib@MPILIBNAME@f90
      
      This change adds the '--with-fcwrapname' configure option so that the
      following is now allowed:
      
        ./configure --with-fcwrapname=mpichf90-foo
      
      .. which results in the f90 mpich library named:
      
        lib@FCWRAPNAME@
      
      (ibm) CPS 92XMVJ
      (ibm) 9346973f8d1dbf6a09b28446b2f7c3c7c5953ea8
      47ea4740
  13. 02 May, 2013 1 commit
    • William Gropp's avatar
      Fixes for #1804 and #1828, related to alignment issues with MPI_Status · 3d96864e
      William Gropp authored
      There are two alignment-related problems with MPI_STATUS arguments in
      Fortran.  The first is that while MPI_STATUS is an INTEGER array in
      Fortran, in C there is an element of type MPI_Count.  MPI_Count may
      be longer than an INTEGER and thus a C MPI_Status may have a stricter
      alignment requirement than a Fortran INTEGER array. The fixes to
      src/binding/f77/buildiface ensures that STATUS arguments that are
      provided to C by the Fortran wrappers are properly aligned, while avoiding
      copies when possible.  The second fix deals with padding that the
      introduction of MPI_Count as the count type into the C structure for
      status will often introduce.
      3d96864e
  14. 28 Apr, 2013 1 commit
    • William Gropp's avatar
      Fix for #1804 - Alignment issues in Fortran with MPI_Status · 81ad170b
      William Gropp authored
      Changed the Fortran wrapper generator to test for proper alignment of
      MPI_Status - because of the new MPI_Count field, some systems may need
      MPI_Status to be aligned on a multiple of sizeof(MPI_Count).  By
      default, new code enforces that, using a temporay status variable only
      when required.  Cost is a few additional tests.
      81ad170b
  15. 23 Apr, 2013 1 commit
    • Dave Goodell's avatar
      tt#1817: move configure subdir exports earlier · 0cca71f8
      Dave Goodell authored
      We already were exporting these variables for the "main" subdirs, such
      as the configured device/channel/netmod.  The bug was that we weren't
      setting these exports in time for OPA or MPL to pick them up.  When
      [1a1ff11e] came along and reordered the compiler search list, this
      caused MPICH to choose `cc` and OPA to choose `gcc`, which don't agree
      on the validity of certain warning flags.  So `--enable-strict=c99`
      causes OPA's configure to fail because of a bad compiler flag exported
      by the top level.
      
      An alternative "fix" to this commit is to ensure that every single
      subdir package uses exactly the same logic in order to select compilers
      and compiler options.  This seems much more difficult to ensure.
      
      Reviewed-by: balaji
      0cca71f8
  16. 12 Mar, 2013 1 commit
    • Dave Goodell's avatar
      `--enable-g=all` now means _all_, add `=most` · 4b6df5cd
      Dave Goodell authored
      Before this commit `--enable-g` meant `--enable-g=all`, which actually
      turned on all debugging options *except* `memarena` and `handle`.
      
      As of this commit:
       * `most` is a new valid debug option having the old effect of
         `all,handle`
       * `all` really means everything now, including `memarena` and `handle`
       * `--enable-g` with no `=BLAH` means `most`
      
      Reviewed by balaji@.
      4b6df5cd
  17. 07 Mar, 2013 1 commit
  18. 06 Mar, 2013 1 commit
  19. 05 Mar, 2013 3 commits
    • Dave Goodell's avatar
      `long double _Complex` depends on `long double` · a9305613
      Dave Goodell authored
      It was possible to have a working `long double _Complex` even if the MPI
      library was pretending that `long double` should not work.
      
      Ditto for the C++ version.
      
      No reviewer.
      a9305613
    • Dave Goodell's avatar
      configure: add `--disable-long-double` · afe8c4a4
      Dave Goodell authored
      This helps Cray out (and conceivably other vendors), since they ship
      several compilers which have different interpretations of what `long
      double` should mean (80-bit vs. 128-bit, being most common).
      
      Note that exporting `MPID_NO_LONG_DOUBLE=yes` in the environment before
      running `configure` does work, but may not work as intended in the face
      of an automake-initiated reconfigure.  Autoconf does not treat this
      variable as precious (and we don't ask it to, since it's for
      inside-configure-use only), so its value will not be preserved for the
      reconfigure.  The same holds for any `MPID_` variables mentioned at the
      top of `configure.ac`.
      
      No reviewer.
      afe8c4a4
    • Dave Goodell's avatar
      fix compile error when `long double` is disabled · 3f1fcead
      Dave Goodell authored
      Otherwise `oputil.h` barfs on a bad preprocessor comparison when
      `MPID_NO_LONG_DOUBLE` is set.
      
      Fix reported by and suggested by Steve Oyanagi from Cray.
      
      No reviewer.
      3f1fcead
  20. 07 Feb, 2013 1 commit
  21. 25 Jan, 2013 2 commits
    • Dave Goodell's avatar
      fixup for new pvar build system code · 7f528a33
      Dave Goodell authored
      Reviewed-by: rgunter
      7f528a33
    • Ralf Gunter's avatar
      Add configure args to selectively enable MPIT pvars. · a4d103b7
      Ralf Gunter authored
      Users can now pass --enable-mpit-pvars to configure to statically
      enable/disable MPIT performance variables so as to limit performance
      impacts only to those variables they need.
      
      Current valid arguments are as follows:
          none     - No performance info recorded
          recvq    - All message queue-related variables
          nem      - All nemesis-related variables
          all      - All variables above
      
      A macro framework is added to facilitate the addition of new variables
      (see MPIR_T_GATE on mpid_common_statistics.h).
      
      Also add a check to the MPIT nemesis test to ensure that the variables
      being tested have been registered with the runtime.
      a4d103b7
  22. 10 Jan, 2013 1 commit
    • Dave Goodell's avatar
      tt#1776: fix platforms missing LLONG_MAX · 465ca32c
      Dave Goodell authored
      Provide an LLONG_MAX (and similar) definition in mpiimpl.h if the
      current platform/environment does not provide one.  Extremely useful for
      overflow checking, especially for the "large count" functionality.
      
      Reviewed-by: buntinas
      465ca32c
  23. 09 Jan, 2013 1 commit
    • Dave Goodell's avatar
      fix automake-1.13+ recursive var warning · 3f56885a
      Dave Goodell authored
      Starting with automake 1.13, the `silent-rules` option no longer implies
      the `-Wno-portability-recursive` option.  (The relevant automake commit
      that changed this behavior is 14141f258, in case we ever need to chase
      this down again.)
      
      No reviewer.
      3f56885a