Flag an error and exit when dynamic tasking APIs are called on
jobs that are not launched in dynamic mode in PE.

closes ticket #1961

(ibm) D192324

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

Fixes a crash in finalize when --enable-debuginfo is set

(ibm) D193788

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

(ibm) D192947

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

  MPI Standard states:
  MPI_PROC_NULL is a valid target rank in all MPI RMA communication calls. The effect
  is the same as for MPI_PROC_NULL in MPI point-to-point communication. After any RMA
  operation with rank MPI_PROC_NULL, it is still necessary to finish the RMA epoch with the
  synchronization method that started the epoch.

  In the current design, both MPI_Win_lock and MPI_Win_unlock do not comply with the standard, e.g.
  if MPI_PROC_NULL is passed in as the target_rank of MPI_Win_lock or MPI_Win_unlock, without the
  fix, the job will be terminated with segfault.

  (ibm) D194295

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

The following functions are added:
- MPI_Win_allocate
- MPI_Win_allocate_dynamic
- MPI_Win_attach
- MPI_Win_deatch
- MPI_Win_lock_all
- MPI_Win_unlock_all
- MPI_Win_sync
- MPI_Win_flush

(ibm) F189033

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

The following changes apply only to the PAMId layer
 - provided MPID_Win_set_info and MPID_Win_get_info functions
 - redesigned the window creation.The change is to support new window creation
   functions introduced in MPI 3.0.
 - updated pscw, MPIDI_WinCtrlSend etc. to reflect the change made in
   MPI_Group_incl().

The following changes were made after a code review
 - added a comment for zero_req in mpidi_onesided.h
 - replaced "pami_task_t peer" "int rank"  in MPIDI_WinCtrlSen()

(ibm) F189041

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

Wesley Bland authored Dec 03, 2013 and Kenneth Raffenetti committed Dec 04, 2013

Fixes #1897. This adds a simple man page for hydra_nameserver, hydra_persist,
and hydra_pmi_proxy. None of these man pages have any real information in them
as users shouldn't be using these executables directly.

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

Wesley Bland authored Dec 03, 2013 and Kenneth Raffenetti committed Dec 04, 2013

This adds the framework to add doctext docs to hydra. To add a .txt file, just
add it to the doc1_src_txt autogen variable.

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

Thanks to Masamichi Takagi @ NEC, Japan for reporting the error and
suggesting the patch.  The cached PUTs array was not reinitialized to
NULL values after the first time they were sent upstream and cleared.
So the next time we push them upstream, they contained garbage data
causing segfaults.  This patch looks for the number of arguments
rather than a NULL argument as a termination detector.

Signed-off-by: Masamichi Takagi <masamichi.takagi@gmail.com>

Wesley Bland authored Dec 03, 2013 and Kenneth Raffenetti committed Dec 03, 2013

There are some files being generated during the build process that aren't
properly ignored. This adds them to the test ignore list.

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

Masamichi Takagi authored Dec 02, 2013 and Pavan Balaji committed Dec 03, 2013

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

Signed-off-by: Masamichi Takagi <masamichi.takagi@gmail.com>

Signed-off-by: Masamichi Takagi <masamichi.takagi@gmail.com>

Signed-off-by: Masamichi Takagi <masamichi.takagi@gmail.com>

Masamichi Takagi authored Nov 29, 2013 and Pavan Balaji committed Dec 03, 2013

dcfa_poll.c and dcfa_send.c are modified.

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

Masamichi Takagi authored Nov 29, 2013 and Pavan Balaji committed Dec 03, 2013

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

Masamichi Takagi authored Nov 29, 2013 and Pavan Balaji committed Dec 03, 2013

Replace manipulating req->status.count with MPIR_STATUS_SET_COUNT.

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

Masamichi Takagi authored Nov 27, 2013 and Pavan Balaji committed Dec 03, 2013

DCFA is IB drivers for Xeon Phi running on McKernel. McKernel is an OS
developed by the University of Tokyo.

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

Some devices may not require ICOMM_WORLD and could cause false
'context not freed' messages in debug mode.

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

James Dinan authored Nov 20, 2013 and Kenneth Raffenetti committed Nov 27, 2013

Update MPI_Comm_create to use sparse, rather than dense, context_id
allocation. This fixes incorrect context ID exhaustion errors caused by
including processes that are not in the group of the new communicator in
the allocation operation. This bug is exercised by
test/mpi/cerrors/comm/too_many_comms3.c.

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

James Dinan authored Nov 20, 2013 and Kenneth Raffenetti committed Nov 27, 2013

Adds a check to determine if context ID allocaiton failed because of
exhaustion or fragmentation and improves error reporting.

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

Adds the -Werror flag to this test because in cases like icc/osx
the compiler only throws a warning that it will ignore function
__attribute__ defs, but in this case we need to know if they actually
work.

Fixes #1980

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

This reverts commit 9fd89d37.

Fixes #1980

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

Patch contributed by IBM. The ordering of this test was causing an
issue in their builds.

Configures testsuite to only run certain large type/count tests on
platforms that can supported larger memory allocations.
Fixes #1975

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

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

Move error checking from critical paths. Fixes #1978

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

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

There were no ABI changes during this RC release cycle. Therefore, we only
update the revision (middle) number. The new string is 11:1:0.

3.1rc1 was released a few weeks ago, but we forgot to update the version
number.

This was causing errors when printing the error descriptions in the manpages.

Fixes #1979

Signed-off-by: Antonio J. Pena <apenya@mcs.anl.gov>

The commented-out variable in 53640e26 is in fact hiddenly used by the macro
MPIU_ERR_SET1, only if HAVE_ERROR_CHECKING is defined. That macro assumes the
variable is declared and does not take it from its arguments. This hotfix
declares the FCNAME[] variable if HAVE_ERROR_CHECKING is defined.

Commented out in case we need it in the future. All other related functions
have it, but they actually use it. This avoids a compiler warning.

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

Moved a couple of variable declarations inside their block to avoid compilers
complaining when those blocks are not compiled because of an #ifdef evaluated
to false.

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