The code is provided by Steve Oyanagi and Bill Long from Cray.

No review since F08 binding is experimental now.

for one-sided communication, the completion of a local operation is when
the completion handler of the operation is activated. The datatype
associated with the operation can not be freed until the completion
handler is activated which is MPIDI_Win_DoneCB. In the current code, the
mapped datatype is unmapped immediately freed after the routine gets the
control from PAMI calls. The calls could be PAMI_Send, PAMI_Get, PAMI_Rget,
PAMI_Put and PAMI_Rget. Please note that the datatype is needed for PAMI
to process the data.

The fix is to unmap the datatype in MPIDI_Win_DoneCB() and remove all
"unmap" datatype operations from put and get operations.

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

MPID_Win_allocated_shared(), the routine could obtain memory space via either
shmget() or MPIU_Malloc() if one task on a node. In MPID_Win_free(), the
routine only frees shared memory space.

The fix is to free none-shared space if there is one.

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

Request based RMA communication operations require that each operation
allocates a communication request object and associates it with the request
handle (the argument request) that can be used to wait or test for
completion. Both the structures for the request handle and the window
structure are not being freed. The problem is when the request handle
is created, the ref_count of the object is set to 2. When
MPID_Request_releas_inline() is called by a request based RMA operation,
the ref_count of the handle is decremented to 1. The allocated memory spaces
are freed only if ref_count is equal to 0.

To fix the problem, set the ref_count to 1 when the request handle is
created. In MPID_Request_release_inline, the routine decrements the
ref_count to 0 which leads the memory space be freed.

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

In MPID_Win_accumulate, when processing non-contiguous datatype
for origin, the ref_count of datatype is incremented by one which
prevents the space allocated for contents and datatloop from being freed.

  MPID_Datatype_add_ref(req->origin.dt.pointer);

In the routine, the data is processed by MPIR_Localcopy and no need to
increment the reference count (ref_count) of the object by 1.

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

No reviewer

Tested using the mpich testsuite and the following command:

  make testing MPITEST_PROGRAM_WRAPPER=" --ranks-per-node=8 --block R00-M1-N10 --timeout 60 --envs BG_MAPCOMMONHEAP=1 : " MPIEXEC=runjob V=1

The configure script could accidentally override the user's selection
of Fortran compiler for mpif77. Move AC_PROG_F77 until after we've
potentially set F77 = FC.

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

Lisandro Dalcin authored Apr 23, 2014 and Rob Latham committed Apr 28, 2014

The standard says a user-defined error code without a string should
return "", not NULL.

Closes #2067

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

a successful test gives ' No Errors'

Lisandro Dalcin authored Apr 22, 2014 and Rob Latham committed Apr 28, 2014

The MPI-3 standard says (pp.354, lines 39-40):

"""
If MPI_ERROR_STRING is called when no string has been set, it will
return a empty
string (all spaces in Fortran, "" in C).
"""

RobL added bits to hook into the test infrastrucutre.

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

if there are a lot of mpi processes with memory errors, the output can
get horribly jumbled if the "memory at address" bit is emitted in one
fprintf call and the "at file/line" bit emitted in a second.  For each
memory region, Build up the address/file-name string in memory and make
one fprintf (instead of two) per memory region.

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

This has been essentially removed for a while now and has been
throwing an error when someone uses it.  This patch gets rid of the
code for good.

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

The --enable-f77 and --enable-fc options are deprecated in favor of
--enable-fortran.  However, instead of just breaking existing scripts
that use them, allow them to be used as well.

In the next major release (3.2), we should remove these options and
throw an error if someone uses them.

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

No ABI changes.  Just update the release count.

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

This test is behaving badly now as well. Given the overhaul of the ft system
that's coming, we're marking these tests as xfail now instead of putting a lot
of work into fixing them. They'll get cleaned up when the ULFM changes go in.

No reviewer

We moved away from --enable-f77 and --enable-fc to --enable-fortran.

No reviewer.

Pavan Balaji authored Apr 23, 2014 and Kenneth Raffenetti committed Apr 25, 2014

We need a better way to maintain consistency between environment
variables common to the MPICH library and mpiexec.  This patch doesn't
do that, but manually moves all of the variables controlling the port
range to mpiexec as well.

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

Some NTFS deployments don't like the optimization where one process
creates a file and all processes open.  From Mark Allen: "On windows
when multiple hosts were used, it ended up that in ad_ntfs_open.c rank 0
used CreateFile() with OPEN_ALWAYS and everybody else used
OPEN_EXISTING, and those ranks all saw ERROR_FILE_NOT_FOUND.  As a fix I
added a retry loop and eventually (within 4-5 sec) the remote ranks
would see the file."

Simply have all processes create the file, as was done long long ago.

Signed-off-by: Mark Allen <markalle@us.ibm.com>

Before returning a larger than size 1 communicator in comm_split type,
make sure that the platform gives us the right tools to support shared
memory.  For example, on FreeBSD, pthread_mutexes don't work
correctly, even though they are a part of POSIX.  In such cases, we
return a size 1 communicator in comm_split_type.

Signed-off-by: Xin Zhao <xinzhao3@illinois.edu>

No reviewer.

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

The user can now enable/disable fortran using the --enable-fortran
option, instead of separate --enable-f77 and --enable-fc.  Expert
users can control specific Fortran standards to use with
--enable-fortran=f77,f90 (and f08 in the future).

Also allow the user to specify only "FC".  If "F77" is not specified,
we use the same compiler as FC.

We now use mpifort as the primary Fortran wrapper, with mpif90 as a
symlink to it for backward compatibility.  If the user specifies a
different F77 compiler than the FC compiler, or if FFLAGS and FCFLAGS
are different, then a separate mpif77 is generated.  Otherwise, mpif77
is a symlink to mpifort.

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

Also squash bad white space changes in the move and replace f90
references with fc.

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

This was added when we were playing with inter-library dependencies.
Since we don't use it anymore, we can remove this code.

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

When we are using the embedded version of libmpl/opa, we can merge it
into libmpi or libpmpi.  It doesn't need to be maintained as a
separate library.

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

Now that we have decided on consistent library names, we don't need to
do any special setting for ABI compatibility.  Just set the
LD_LIBRARY_PATH and run.

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

Don't provide a default prefix path since that's updated at make
install time anyway, for example, if the user specifies a new prefix.

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

These were gratuitous variables that didn't serve any real purpose.

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

Pavan Balaji authored Apr 16, 2014 and Pavan Balaji committed Apr 23, 2014

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

The rerun of buildiface for rebuilding mpicxx.h wasn't use the
appropriate init file that reorders the virtual functions.  Also
remove an empty unused variable.

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

Pavan Balaji authored Apr 22, 2014 and Pavan Balaji committed Apr 23, 2014

The paths used in the extractcvars script were not VPATH build safe
earlier.  Since this is mostly called from autogen.sh, it didn't
matter earlier.  But in cases where the make logic tries to run it
again, it was failing.

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

Lisandro Dalcin authored Apr 22, 2014 and Rob Latham committed Apr 22, 2014

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

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

Paul Coffman authored Apr 21, 2014 and Rob Latham committed Apr 21, 2014

a bug with the balancecontig optimization could result in the ions
having non-contiguous file offset ranges. this was hurting overall
performance of the optimization plus the gpfs access range hints.

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

In the MPIC_Sendrecv functions, the status object should always be defined
since we use it internally. This won't have any impact on performance since
the default is always to have FT collectives turned on anyway, but it will
prevent a crash when someone overwrites that default.

Fixes #2026

Signed-off-by: Sangmin Seo <sseo@anl.gov>