1
1
Граф коммитов

20127 Коммитов

Автор SHA1 Сообщение Дата
Ralph Castain
e05b88fd18 Take another stab at resolving the "called-abort" requirement without getting stuck. Return to "drop a turd" mode, perhaps with a little more intelligence behind it. Don't worry about catching it if session dirs weren't created
cmr=v1.8.2:reviewer=jsquyres:subject=cleanup MPI_Abort hangs

This commit was SVN r31543.
2014-04-29 17:29:46 +00:00
Jeff Squyres
908e178d08 Ensure that VPATH builds work with the ORTE version.h file.
This commit was SVN r31542.
2014-04-29 17:21:02 +00:00
MPI Team
ff7814cdd0 Update git/hg ignore files
This commit was SVN r31541.
2014-04-29 05:00:26 +00:00
Ralph Castain
2c6234698e Fix the tarball build - need to include the orte_config.h header
This commit was SVN r31540.
2014-04-29 00:05:19 +00:00
Nathan Hjelm
2f5b1ca4cf osc/rdma: do not leak the receive request
This commit fixes a bug that can cause request and communicator leaks
when cleaning up an OSC window. The should prevent a hang seen with
IMB-EXT.

cmr=v1.8.2:reviewer=jsquyres

This commit was SVN r31539.
2014-04-28 19:55:18 +00:00
Nathan Hjelm
e410401523 comm: detect if we run out of communicator ids (cids)
Due to a leak in the osc/rdma component we were running out of cids on
a one-sided tests. This resulted in a hang instead of an error. This
commit causes the nextcid algorithm to return an error if we run out
of cids.

cmr=v1.8.2:reviewer=jsquyres

This commit was SVN r31538.
2014-04-28 19:55:09 +00:00
Ralph Castain
3723b39f30 Ensure we don't silently fail when unable to make a connection - bark pleasantly first.
Refs trac:4571

This commit was SVN r31537.

The following Trac tickets were found above:
  Ticket 4571 --> https://svn.open-mpi.org/trac/ompi/ticket/4571
2014-04-28 19:16:32 +00:00
Ralph Castain
d642babff6 Derived from patch provided by Artem, cleanup the "abnormal" code path for selecting TCP OOB modules to connect to a remote process. If we can't find a direct interface-to-address match, then assign all the provided addresses to the first available TCP module and let the normal failure process determine if the remote proc is truly reachable.
cmr=v1.8.2:reviewer=artpol:subject=fix abnormal code connection path in tcp oob

This commit was SVN r31536.
2014-04-28 19:05:14 +00:00
Nathan Hjelm
626b521e9c pml/ob1: fix heterogeneous support when using the send_inline optimization
We will track #4568 from the 1.8 CMR.

Closes trac:4568

cmr=v1.8.2:reviewer=jsquyres

This commit was SVN r31535.

The following Trac tickets were found above:
  Ticket 4568 --> https://svn.open-mpi.org/trac/ompi/ticket/4568
2014-04-28 17:36:26 +00:00
Ralph Castain
fb61a94804 Follow the lead set by Jeff: no need to run AC_CONFIG_HEADERS on orte_config.h. However, unlike the MPI layer, we don't run that macro on another file in orte/include, so ensure we add that -I path back!
This commit was SVN r31534.
2014-04-28 17:12:15 +00:00
Ralph Castain
def965cbd4 Ensure we don't include the globals twice at the beginning of the file in non-svn environments as the .<foo>ignore_global file already contains them. Remove the extra check for svn repo in the hgignore script.
This commit was SVN r31533.
2014-04-27 14:20:18 +00:00
Ralph Castain
f8164bbd80 Update the automated build ignore scripts so they work properly in non-svn environments by simply copying the global and any local files into the respective .<foo>ignore file
This commit was SVN r31532.
2014-04-27 04:26:24 +00:00
MPI Team
108469a2c5 Update git/hg ignore files
This commit was SVN r31531.
2014-04-26 05:00:26 +00:00
Ralph Castain
4399d1157e Update the minimum required versions to match that on the web site
This commit was SVN r31530.
2014-04-25 21:10:05 +00:00
Mike Dubman
1f8b4776e6 update ignore list
This commit was SVN r31529.
2014-04-25 09:17:18 +00:00
Jeff Squyres
64c1228b55 Roll back r31519 and r31521: George convinced us that these approaches
weren't right.

This commit was SVN r31528.

The following SVN revision numbers were found above:
  r31519 --> open-mpi/ompi@b449c750b7
  r31521 --> open-mpi/ompi@e243805ed8
2014-04-24 20:27:03 +00:00
Nathan Hjelm
c9a257f1a0 btl/ugni: always buffer sendi fragments
This commit will improve the message rate when using the sendi function
by not waiting for the send to get to the remote process.

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31526.
2014-04-24 18:50:29 +00:00
Jeff Squyres
871e20cd4b MPI_Alltoallv.3in: fix typo
Fix minor typo reported by Xuankang Lin.

cmr=v1.8.2:reviewer=dgoodell

This commit was SVN r31525.
2014-04-24 18:14:42 +00:00
George Bosilca
17b3c7e906 Fix the issue reported by Gilles Gouaillardet regarding the
MPI_PROC_NULL persistent requests.

This commit was SVN r31524.
2014-04-24 18:07:09 +00:00
Nathan Hjelm
52f519dacb Allow MPI_MODE_NOPRECEDE | MPI_MODE_NOSUCCEED for MPI_Win_fence
This combination does not make sense but is not explicitly forbidden by
the standard so remove the argument check for this combination.

cmr=v1.8.2:reviewer=jsquyres

This commit was SVN r31523.
2014-04-24 17:36:10 +00:00
Nathan Hjelm
0849d61e38 btl/vader: improve performance under heavy load and eliminate a racy
feature

This commit should fix a hang seen when running some of the one-sided
tests. The downside of this fix is it reduces the maximum size of the
messages that use the fast boxes. I will fix this in a later commit.

To improve performance under a heavy load I introduced sequencing to
ensure messages are given to the pml in order. I have seen little-no
impact on the message rate or latency with this change and there is a
clear improvement to the heavy message rate case.

Lets let this sit in the trunk for a couple of days to ensure that
everything is working correctly.

cmr=v1.8.2:reviewer=jsquyres

This commit was SVN r31522.
2014-04-24 17:36:03 +00:00
Jeff Squyres
e243805ed8 coll tuned alltoallv: correctly handle 0-sized messages with MPI_IN_PLACE
Patch from Gilles Gouaillardet on #4517 to fix handling 0-sized
messages in coll tuned with MPI_ALLTOALLV and MPI_IN_PLACE.

Reviewed by Jeff Squyres.

Fixes trac:4517

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31521.

The following Trac tickets were found above:
  Ticket 4517 --> https://svn.open-mpi.org/trac/ompi/ticket/4517
2014-04-24 16:55:53 +00:00
Jeff Squyres
b449c750b7 coll basic: correctly handle alltoall[vw] 0-sized messages
Patch from Gilles Gouaillardet on #4506 to correctly handle 0-sized
messages in coll/basic MPI_Alltoallv and MPI_Alltoallw.

Reviewed by Jeff Squyres.

Fixes trac:4506.

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31519.

The following Trac tickets were found above:
  Ticket 4506 --> https://svn.open-mpi.org/trac/ompi/ticket/4506
2014-04-24 16:25:43 +00:00
Jeff Squyres
ca80c7a9bd cart_sub.c: allow remain_dims==NULL if there is no topology on comm
Patch submitted by Gilles Gouaillardet on #4518.  Reviewed by Jeff.

Fixes trac:4518

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31517.

The following Trac tickets were found above:
  Ticket 4518 --> https://svn.open-mpi.org/trac/ompi/ticket/4518
2014-04-24 16:10:44 +00:00
Jeff Squyres
e9b694f1d8 coll_base_comm_unselect.c: fix memory leaks
Ensure to also OBJ_RELEASE the neightbor and ineighbor modules.

Fixes trac:4444 (this patch is from that ticket).

This commit was SVN r31516.

The following Trac tickets were found above:
  Ticket 4444 --> https://svn.open-mpi.org/trac/ompi/ticket/4444
2014-04-24 15:53:06 +00:00
Jeff Squyres
d8715f1e3a Close 3 more fd's that were leaking into child processes.
Child processes now look clean; I can't find any more fd's that are
leaking from the parent to children.

Refs trac:4550

This commit was SVN r31515.

The following Trac tickets were found above:
  Ticket 4550 --> https://svn.open-mpi.org/trac/ompi/ticket/4550
2014-04-24 15:36:24 +00:00
Jeff Squyres
790cdb5cc7 Sigh. It helps when you commit the right version of the finished
code.

This commit fixes minor errors in the incorrectly-committed r31513
(new fd close-on-exec convenience function).

Refs trac:4550

This commit was SVN r31514.

The following SVN revision numbers were found above:
  r31513 --> open-mpi/ompi@e1655ae68d

The following Trac tickets were found above:
  Ticket 4550 --> https://svn.open-mpi.org/trac/ompi/ticket/4550
2014-04-24 13:20:32 +00:00
Jeff Squyres
e1655ae68d opal/util/fd.c: add new convenience function for setting FD_CLOEXEC
Paul Hargrove pointed out that Stevens tells us that we should
FD_GETFL before FD_SETFL.  And so we shall.

Make a new convenience function to do this (opal_fd_set_cloexec()),
just so that we don't have to litter this 2-step process throughout
the code.

Refs trac:4550

This commit was SVN r31513.

The following Trac tickets were found above:
  Ticket 4550 --> https://svn.open-mpi.org/trac/ompi/ticket/4550
2014-04-24 13:04:49 +00:00
Jeff Squyres
410f5bfb91 oob_tcp_listener.c: set both ends of this thread to be close-on-exec
This pipe is used to communicate between threads in this process.
Mark both fd as close-on-exec so that children don't inherit this
pipe.

Refs trac:4550

This commit was SVN r31512.

The following Trac tickets were found above:
  Ticket 4550 --> https://svn.open-mpi.org/trac/ompi/ticket/4550
2014-04-23 21:46:41 +00:00
Jeff Squyres
67bb0c261a hwloc: ensure that an internal fd is marked as close-on-exec
Make sure that an internal, long-lived hwloc fd is marked as
close-on-exec so that children don't inherit it.  This patch is
committed upstream in the hwloc master and v1.9 branches as 7489287
and b654e19, respectively.  The patch applied here is the exact same
logic, but the surrounding code changed slightly since the hwloc v1.7
series, so the patch doesn't apply cleanly.

Refs trac:4550

This commit was SVN r31511.

The following Trac tickets were found above:
  Ticket 4550 --> https://svn.open-mpi.org/trac/ompi/ticket/4550
2014-04-23 21:36:38 +00:00
Mike Dubman
0cba0b95b8 OSHMEM: fortran profiling API added
for spec compliance 

fixed by Roman, reviewed by Igor/Miked

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31510.
2014-04-23 18:56:37 +00:00
George Bosilca
024221f469 Initialize some fields (prevent valgrind complaints).
This commit was SVN r31503.
2014-04-23 13:38:30 +00:00
Mike Dubman
7c47aab927 OSHMEM: fix segv with new OFED
new OFED changed struct layout and static assignment caused segv.
detect struct new layout and use dynamic assignment

fixed by AlexM, reviewed by Miked

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31502.
2014-04-23 11:38:44 +00:00
Mike Dubman
8e4cea59da OSHMEM: bug fix for hca selection from list
fixed by AlexM, reviewed by Miked

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31501.
2014-04-23 11:32:52 +00:00
Jeff Squyres
87e6232e67 orterun.c: set an fd to be close-on-exec
Make sure the debugger attach fifo is marked as close-on-exec so that
children procs don't inherit it.  For example, if you salloc a SLURM
allocation and run "mpirun ..." in there (i.e., mpirun is running on
the head node, and launching on to back-end nodes), the forked srun's
will inherit this fd if it is still open.

Refs trac:4550

This commit was SVN r31499.

The following Trac tickets were found above:
  Ticket 4550 --> https://svn.open-mpi.org/trac/ompi/ticket/4550
2014-04-22 21:55:09 +00:00
Jeff Squyres
b65e7caa77 op_commutative.c: use ompi_op_is_commute(), not the expanded test
One more commit for this ticket... as pointed out by Giles, we have
ompi_op_is_commute().  We should use that instead of replicating the
logic for the test.

Refs trac:4548

This commit was SVN r31497.

The following Trac tickets were found above:
  Ticket 4548 --> https://svn.open-mpi.org/trac/ompi/ticket/4548
2014-04-22 20:08:40 +00:00
Jeff Squyres
b0c4b1aa20 op_commutative.c: ...but we should check for MPI_OP_NULL
Refs trac:4548

This commit was SVN r31492.

The following Trac tickets were found above:
  Ticket 4548 --> https://svn.open-mpi.org/trac/ompi/ticket/4548
2014-04-22 19:30:55 +00:00
Jeff Squyres
dba256cd6c op_commutative.c: don't check if the op is predefined or not
MPI_OP_COMMUTATIVE should work on all MPI_Op's -- regardless of
whether they are predefined or not.

Refs trac:4548.

This commit was SVN r31491.

The following Trac tickets were found above:
  Ticket 4548 --> https://svn.open-mpi.org/trac/ompi/ticket/4548
2014-04-22 19:26:15 +00:00
Jeff Squyres
5d17628823 Add in an opal_output_verbose() so that we'll see the case where there
are no usNICs found.

Refs trac:4549

This commit was SVN r31489.

The following Trac tickets were found above:
  Ticket 4549 --> https://svn.open-mpi.org/trac/ompi/ticket/4549
2014-04-22 18:59:10 +00:00
Mike Dubman
a4990de055 mca: track external lib version (runtime/compiletime) for mca component
based on thread: http://www.open-mpi.org/community/lists/devel/2014/04/14505.php

Create mca parameter to track runtime/compiletime ext lib version for component.

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31487.
2014-04-22 18:02:26 +00:00
Mike Dubman
18f9a282d5 OSHMEM: fix clang warnings
fixed by Roman, reviewed by Miked

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31486.
2014-04-22 16:49:20 +00:00
Jeff Squyres
2e191d2e21 Sync with 1.8.1/1.8.x NEWS.
This commit was SVN r31485.
2014-04-22 15:19:44 +00:00
Jeff Squyres
63b7ef4103 orterun.1in: Document --allow-run-as-root option
Add some verbiage about how mpirun now defaults to disallowing running
as root, but you can use the --allow-run-as-root option to override
this default behavior.

Refs trac:4536

This commit was SVN r31477.

The following Trac tickets were found above:
  Ticket 4536 --> https://svn.open-mpi.org/trac/ompi/ticket/4536
2014-04-22 14:34:32 +00:00
George Bosilca
75cf79c783 Ahem ... Correctly implement most of the 3 arguments
operator in Open MPI. Creepy that it was not discovered earlier.

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31473.
2014-04-21 23:31:23 +00:00
George Bosilca
8d92231de3 Deprecated comment.
This commit was SVN r31472.
2014-04-21 23:30:05 +00:00
George Bosilca
6a65d27bcc Print the 3rd buffer for the MPI_Op.
This commit was SVN r31471.
2014-04-21 23:29:30 +00:00
Oscar Vega-Gisbert
296bc56241 Java: apply a patch which solves the ticket #4313
This commit was SVN r31470.
2014-04-21 21:44:25 +00:00
Jeff Squyres
ea4c916096 plm_slurm_module.c: don't leave the extra fd to /dev/null open
Prior to r29058, this same logic was in place (i.e., ensure that the
extra fd to /dev/null is closed).  It looks like it was accidentally
removed in the ORTE conversion to the state machine in r29058.

This ''might'' have something to do with many hangs that we're seeing
in Cisco MTT with jobs that exhibit failure (e.g., call MPI_ABORT)...?

cmr=v1.8.2:reviewer=rhc

This commit was SVN r31469.

The following SVN revision numbers were found above:
  r29058 --> open-mpi/ompi@a200e4f865
2014-04-21 20:09:15 +00:00
Jeff Squyres
a3acc49688 usnic_component.c: don't complain if there are no usNIC devices
cmr=v1.8.2:reviewer=dgoodell

This commit was SVN r31468.
2014-04-21 19:28:48 +00:00
George Bosilca
399c8b18d2 Fix a parameter check. Thanks Lisandro Dalcin for the bug report.
More info @ http://www.open-mpi.org/community/lists/devel/2014/04/14561.php

cmr=v1.8.2:reviewer=ompi-rm1.8

This commit was SVN r31467.
2014-04-21 17:00:56 +00:00