GitLab

[svn-r7416] Major improvement to RMA performance for long lists of operations, an immediate mode accumulate for single ints, store the MPID_Comm within the window, and added a basic performance instrumentation interface that was extensively used to improve the RMA performance (enabled with --enable-g=instr).  With these fixes, MPICH2 can run the one-sided version of the Graph500 benchmark at a respectable if not great rate

found.

[svn-r7341] replaced the runtime test that determine sizeof($MPI_OFFSET_TYPE) where MPI_OFFSET_TYPE is determined by ROMIO by AC_COMPUTE_INT of sizeof(MPI_Offset).  This fix is related to req#1121.

[svn-r7274] reverting r7261 so that we can fix the --enable-threads=multiple issue. Maybe we should also change the name to enable-threads=force_multiple or something

[svn-r7261] set --enable-threads=multiple to threads=runtime since it is more efficient. fixes #1100

[svn-r7197] Replaced warning by abort when only --disable-f77 without --disable-fc is used, the abort is issed after some length check but at AC_ARG_ENABLE level.  Also removed AC_HELP_STRING in the all AC_ARG_VAR, because AC_HELP_STRING which requires at least 2 arguments is meant to be used in AC_ARG_WITH and AC_ARG_ENABLE.

[svn-r7196] added check for --enable-f90/--disable-f90 and abort if the option is used, also moved *AC_PROG_CC* and the related after most of the AC_ARG_ENABLE/AC_ARG_WITH so the abort happens pretty much immediately (i.e. without any major configure processing).  This resolves the req# 1093.

Embarassingly, this has been broken for a while.  This commit also
adds a regression test: "cxx/coll/reduceboolx".

No reviewer.

use more recent compilers when available instead of relying on older
compilers by default: gfortran vs. g77 is one motivating example
(autoconf prefers g77 for Fortran 77). Also remove MPD specific
macros, allowing it to use confdb directly, so MPD can take advantage
of this change as well.

Reviewed by goodell.

Reviewed by balaji@.

as well.

Reviewed by goodell.

script. The compiler is needed for everything. The shared library
checks were failing without this. Also removed some comments that are
no longer relevant with the current state of the configure script, and
moved some code around to keep the section of the code that works on
the environment variables and other flags separate from the rest of
the code.

libtool, and --enable-sharedlibs=<type> for the MPICH2 core. To avoid
this confusion, if the user just provides --enable-shared alone,
assume --enable-sharedlibs=gcc as well.

Reviewed by goodell.

and Hydra can share it.

No reviewer.

to just link with libmpich.so, and the remaining libraries are
dynamically loaded. This support is only for shared libraries at this
point (static library support might require libtool). Also, this is
experimental, as it has a bunch of gotchas that we cannot handle
unless we reimplement libtool.

Reviewed by goodell.

Reviewed by goodell.

added a new test PAC_F2C_ATTR_ALIGNED_SIZE to replace PAC_C_ATTR_ALIGNED to determine the alignment of Fortran common block as seen in C.  Also added a switch, --enable-multi-aliases, to disable the feature just in case something goes wrong.

[svn-r7054] Add support for disable-rpath in configure to allow building shared libraries and scripts to support knowning where the shared libraries were installed (the best for most users) or, with --dsiable-rpath, requiring the use of LD_LIBRARY_PATH, which is best for building libraries to be installed in other locations

[svn-r6874] Removed the 2nd set of compilers in test/mpi/configure.in, i.e. BASExx, which are basically replaced by "mpixx -nativelinking".  This greatly simplifies the test/mpi/configure.in because there is no need to swap between 2 set of compilers and compiler flags, CC/CFLAGS vs MPICC/MPICC_CFLAGS, for every languages in every tests.  Also replaced all AC_TRY_* by AC_*_IFELSE and removed all references of AC_FD_CC, updated the configure.in to autoconf-2.6x style.  Removed AC_SUBST(autoconf precious variables)....

[svn-r6853] fixed the typo in MPICH2_BASE_* which should be MPICH2_BASE* in configure.in.  (These variables are exported to test/mpi/configure.in)  Fixed (temporary) the
incorrect setting of BASEF77 when FROM_MPICH2=1 in test/mpi/configure.in.

[svn-r6852] Initial deposit that removes the AC_LANG=[Fortran 90] and replaces it by AC_LANG=[Fortran] which is officially supported by autoconf.  The replaced PAC macros in aclocal_fc.m4 is also rewritten by replacing all use of AC_TRY_* by AC_*_IFELSE style as recommented by autoconf.  The change also removes some old autoconf macros date back to at least 2.13 as in aclocal_cross.m4.  The change that users see is to use FC/FCFLAGS instead of F90/F90FLAGS which will no longer be supported.  This also includes some misc fixes that allow mpif90 to access mpi.mod during build process....

[svn-r6817] Adding format specifier for Pints and removing instances of custom Aint format specifier defns - Codes that need Aint format spec should use MPI_AINT_FMT_*

Pass --with-atomic-primitives=auto_allow_emulation to the OPA configure
unless the user specifies a particular value.

This is a regression introduced in r6223, between 1.2.1p1 and 1.3a1.

Reviewed by balaji@.

A partial merge of r5954 from the bg-threads-1.1 branch.

No reviewer.

Reviewed by buntinas@.

This is actually several logical changes, but they are very difficult to
separate into individual commits because of how tangled the old macros
were.  Some of the changes are:

- general macro cleanup: do{}while(0) and whitespace issues
- add support for compiler-assisted thread-local storage (TLS)
- make MPID_Thread (e.g. pthread) TLS macros easier to understand and use
- drop non-portable parentheses from MPIU_ISTHREADED macro
- eliminate the MPID_CS_* family of macros which were simultaneously
  complementary to and in conflict with the newer MPIU_THREAD_CS_*
  macros

No reviewer.

This was a long-standing TODO.

No reviewer.

was broken in r6273, prevents CPPFLAGS from being honored
(but MPICH2LIB_CPPFLAGS is a workaround).

No reviewer.

is needed for suncc.

Reviewed by gropp.

appropriate variables are declared for it to work in emulation mode as
well. This showed up as an error with the suncc compiler on linux
which is not natively supported by OPA.

error out at configure time, instead of just throwing a warning and
failing at make time.