In MPID_Win_allocate and MPID_Win_allocate_shared, baseptr are defined
as void * and void ** separately, while in MPIDI_Win_fns, both
MPID_Win_allocate and MPID_Win_allocate_shared are registered as
MPIDI_CH3U_Win_allocate, where baseptr is defined as void *.

Fixes #1995

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

Insert a barrier for every 16 calls to async collectives

(ibm) D192641

Signed-off-by: Michael Blocksome <blocksom@us.ibm.com>

In the unlikley event that someone checks error codes, let them know
they might want to pass a non-null filename to MPI-IO

Signed-off-by: Wesley Bland <wbland@mcs.anl.gov>

Thanks be to Charles :)

Signed-off-by: Su Huang <suhuang@us.ibm.com>

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

I have the PVFS/Orangefs headers on my system now, so problems like this
will stop happening.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

ROMIO is rarely built standalone, and anyway in 2014 one should need no
prompting to "type make".

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Drive down the differences between OpenMPI's ROMIO and "upstream" ROMIO
(us) by teaching ROMIO what to do if it might be part of OpenMPI.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Was converting to and from the same buffer in some cases.

OpenMPI SVN r30319

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

This should help OpenMPI next time they want to re-sync.

Sort of based on OpenMPI SVN r27501, but now actually checks.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Basic rule for C macros: wrap the macro arguments with parens to prevent
unexpected order of operations.  From OpenMPI SVN r24356

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Seems unlikely that anything *doesn't* have statfs these days, but
OpenMPI SVN r22547 added this check for Catamount.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

OpenMPI defines this constant in mpi.h

See OpenMPI svn revsion r21307

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

from OpenMPI SVN, revision r17543

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Huiwei Lu authored Mar 09, 2014 and Pavan Balaji committed Mar 09, 2014

Last fix corrected malloc(0) problem but forgot the free part.

Fixes #2054

Signed-off-by: Pavan Balaji <balaji@mcs.anl.gov>

electric-fence tool detects malloc(0) in mpid_nem_init.c

Fixes #2054

Signed-off-by: Wesley Bland <wbland@mcs.anl.gov>

Bob Cernohous authored Oct 15, 2013 and Rob Latham committed Mar 04, 2014

Also, do not reserve the internal-only copy of comm world context id when
it isn't being used (MPID_NEEDS_ICOMM_WORLD)

See mpich.org ticket #1947

closes #1947

(ibm) 984acc33d6c8149987016b62da5afb36450be94e

Signed-off-by: Michael Blocksome <blocksom@us.ibm.com>
Signed-off-by: Rob Latham <robl@mcs.anl.gov>

Haizhu Liu authored Oct 11, 2013 and Rob Latham committed Mar 04, 2014

Signed-off-by: Michael Blocksome <blocksom@us.ibm.com>
Signed-off-by: Rob Latham <robl@mcs.anl.gov>

See OpenMPI revision 30932:
https://svn.open-mpi.org/trac/ompi/changeset/30932

ifdef test on the need for a prototype declaration was using the
wrong name.  This may have been caused by a change in the name
of the CPP value without correcting the use of that name in this file.

We were disabling VA_ARGS whenever the compiler throws warnings.
However, in some cases, the compiler is throwing warnings in general
(because of bad user-specified flags) and not specifically because of
VA_ARGS.  See discussion on the discuss@mpich.org mailing list:
http://lists.mpich.org/pipermail/discuss/2014-February/002246.html



This patch works around this issue by checking the difference in the
warning flags with and without VA_ARGS.

Fixes #2047.

Signed-off-by: Huiwei Lu <huiweilu@mcs.anl.gov>

MPICH datatypes can be used in a switch statement, but OpenMPI uses
pointers.

The OpenMPI guys carried this patch for far too long. see
https://svn.open-mpi.org/trac/ompi/ticket/2808

Create index.html files for the docs so they are easier to navigate.
This restores functionality that was lost during the build system
overhaul. Closes #1800

Signed-off-by: Pavan Balaji <balaji@mcs.anl.gov>

Simply ran the new ./maint/check_copyright.bash script.

Fixes #2032.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

Simple script to check the git log and mark the first date the file
was created on as the copyright date.

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

The functions searched generally live in libc, but some platforms (e.g.
Solaris) have them in special libraries. Use AC_SEARCH_LIBS to include
the extra libraries when needed. Fixes #2034

Signed-off-by: Pavan Balaji <balaji@mcs.anl.gov>

The comment in the test program was referring to ARMCI, since it was
carried over from the ARMCI-MPI test suite.

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

ARMCI-MPI is released separately.  There's no reason to package it in
MPICH again.

Fixes #2037.

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

if aio not supported or doesn't work or is temporarily unavailable we
fall back to blocking versions, but forgot to clean up allocated memory.

Closes #2018

Signed-off-by: Ken Raffenetti <raffenet@mcs.anl.gov>

see http://lists.mpich.org/pipermail/discuss/2014-February/002224.html


for the details

Fixes #2038

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

PETSc users are confused by error messages because they can't be
troubled to understand them.

Barry Smith requested that MPICH not indicate that it is aborting.

this is part 1 of N patches to address this.

Signed-off-by: Junchao Zhang <jczhang@mcs.anl.gov>

These tests were adapted from the mpich-1 test suite but they are their
own creatures now.

Original-author: William Gropp <wgropp@illinois.edu>
Signed-off-by: Rob Latham <robl@mcs.anl.gov>

William Gropp authored Feb 02, 2014 and Rob Latham committed Feb 24, 2014

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

William Gropp authored Feb 02, 2014 and Rob Latham committed Feb 24, 2014

The xfail additions to the runtests code had an incorrect test on xfail
that caused runtests to stop recording a test that could not be built
as failed - thus, the test script would report success for all tests when
some tests were not run.

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

William Gropp authored Jan 29, 2014 and Rob Latham committed Feb 24, 2014

This adds dataalign, sendrecvt2, and sendrecvt4 to the tests of MPI
datatypes in the MPICH test suite.

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

William Gropp authored Jan 29, 2014 and Rob Latham committed Feb 24, 2014

This test moves the sendrecv2 and sendrecv4 test from MPICH-1 to the new
MPICH tests.  It has been updated to use the MTest framework.  This test
uses the datatypes created by the dtypes.c file (also moved from
MPICH-1).

These tests involve creating a structure type with an explicit offset of
a variety of types, and then communicated with MPI_BOTTOM as the buffer
location.  Several of these tests failed in a recent update to the Dataloop
code for optimization of performance.

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

William Gropp authored Jan 29, 2014 and Rob Latham committed Feb 24, 2014

This commit add the dataalign test from MPICH1.  It has been updated to
use the MTest framework.  This tests a specific struct datatype, and is
a test that caused one recent dataloop optimization to fail.

Signed-off-by: Rob Latham <robl@mcs.anl.gov>

William Gropp authored Jan 29, 2014 and Rob Latham committed Feb 24, 2014

This adds the routines originally defined in the MPICH test suite that
created a range of datatype tests, including the input and output data.
This commit moves them to the new MPICH and puts them into the MTest
framework, as an optional file.

Signed-off-by: Rob Latham <robl@mcs.anl.gov>