1
1

24985 Коммитов

Автор SHA1 Сообщение Дата
Ralph Castain
e6ad1ad621 Up-port of change for 2.x: if user directs oversubscribe, then do not bind as we will otherwise overload resources 2016-04-28 13:21:10 -07:00
Howard Pritchard
f52dd511d4 Merge pull request from hppritcha/topic/pmix_fix_for_finalize
pmix/cray: set fence_nb to NULL
2016-04-28 13:50:15 -06:00
hppritcha
aa1d7b9c50 pmix/cray: set fence_nb to NULL
Rather than have a stub function for the pmix fence_nb
operation, just set to NULL.  Causes fewer problems.

Fixes 
Fixes 

Signed-off-by: hppritcha <howardp@lanl.gov>
2016-04-28 13:48:54 -05:00
Jeff Squyres
743f426f96 NEWS: Add bullets from 1.10 branch
These bullets were over in the v1.10 branch, but not here on master.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-28 11:15:21 -04:00
Jeff Squyres
d15c324d9d NEWS: A few formatting fixes
Also fix/update an Intel copyright (from the v2.x branch).

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-28 11:03:55 -04:00
Jeff Squyres
d7395798ec NEWS: sync some bullets from the v2.x branch
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-28 11:00:01 -04:00
Jeff Squyres
4564878173 NEWS: Add release dates for each version
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-28 10:49:55 -04:00
Nysal Jan K.A
18cf65dc24 Remove a stray print statement 2016-04-28 18:00:52 +05:30
Nathan Hjelm
a068b9483b Merge pull request from hjelmn/sparse_group_fix
ompi/group: fix sparse group proc reference counting
2016-04-27 17:06:01 -06:00
Nathan Hjelm
1783d94f91 ompi/group: fix sparse group proc reference counting
This commit fixes a bug when sparse groups are in use. Since sparse
group do not actually increment the reference counts of any procs
(they just retain the parent group) it is wrong to decrement the
reference counts of all procs in the group using
ompi_group_decrement_proc_count(). This commit makes the call to
ompi_group_decrement_proc_count() conditional on the group being
dense.

Fixes 

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-04-27 15:55:13 -06:00
rhc54
39137770ae Merge pull request from rhc54/topic/size
Correctly set the #procs in the job to "job_size", and the max_procs to "univ_size"
2016-04-27 13:01:12 -07:00
Ralph Castain
75dc4c305a Correctly set the #procs in the job to "job_size", and the max_procs to "univ_size" 2016-04-27 12:00:19 -07:00
Nathan Hjelm
936dfe5c26 Merge pull request from hjelmn/tcp_thread_fix
btl/tcp: fix add_procs race condition
2016-04-27 11:40:00 -06:00
Nathan Hjelm
03f4a854cb btl/tcp: fix add_procs race condition
This commit fixes a race between a thread calling the tcp btl's
add_procs and a thread processing an incomming connection. The race
occured because the add_procs thread adds a newly created proc object
to the hash table *before* the object is fully initialized. The
connection thread then attempts to use the object before the endpoints
array on the object has beeen allocation. The fix is to only add the
proc to the hash table after it has been completely initialized.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-04-27 10:24:39 -06:00
Gilles Gouaillardet
01c90d4e71 fortran/mpif-h: fix *_create_keyval_f
correctly handle out parameter *_keyval when OMPI_SIZEOF_FORTRAN_INTEGER > SIZEOF_INT
2016-04-27 13:34:32 +09:00
Gilles Gouaillardet
178dde6a20 fortran/mpif-h: fix MPI_Win_shared_query
correctly handle out parameter disp_unit when OMPI_SIZEOF_FORTRAN_INTEGER > SIZEOF_INT
2016-04-27 11:22:09 +09:00
Gilles Gouaillardet
529ab559d4 configury: consider double _Complex as a candidate for the C equivalent of Fortran COMPLEX
this is necessary when -d8 flag is passed to Fortran compiler
2016-04-27 10:55:35 +09:00
Gilles Gouaillardet
7f59d2a8c7 fortran/mpif-h: fix MPI_Win_free_keyval
initialize inout parameter when OMPI_SIZEOF_FORTRAN_INTEGER > SIZEOF_INT
2016-04-27 10:46:14 +09:00
Nathan Hjelm
8f93b15e90 Merge pull request from hjelmn/new_hooks_update
memory/patcher: cast away const in shmdt hook
2016-04-26 17:48:01 -06:00
Nathan Hjelm
df194087c7 Merge pull request from hjelmn/rcache_update
rcache: fix leave_pinned failure path
2016-04-26 16:50:06 -06:00
Nathan Hjelm
f0f3383006 Merge pull request from hjelmn/thread_multiple
osc/pt2pt: do not drop/reacquire the ompi_request_lock
2016-04-26 16:48:37 -06:00
Nathan Hjelm
25a97af695 rcache: fix leave_pinned failure path
This commit fixes an error in the failure path of leave_pinned. When
the rcache tries to enable leave_pinned but leave_pinned was not
specifically requested (opal_leave_pinned == -1) the code was
erroneously printing an error and returning NULL.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-04-26 14:39:23 -06:00
Nathan Hjelm
34ff6293bd osc/pt2pt: do not drop/reacquire the ompi_request_lock
This lock is now recursive so it is safe to call into the pml without
dropping the lock.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-04-26 14:19:38 -06:00
George Bosilca
bf190671e9 Make the request lock recursive.
If during the request completion callback we post another request that
completes right away (such a small send or a match for an unexpected
short message) we will try to complete the second request while holding
the lock for the completion of the first. For performance reasons
(mainly to avoid unlocking and locking the request mutex several times)
we have made the request lock recursive.
2016-04-26 16:16:07 -04:00
rhc54
163cb65bb2 Merge pull request from rhc54/topic/cleanup
Silence warning of zero-byte malloc
2016-04-26 13:12:22 -07:00
Nathan Hjelm
4cb13b0fdf Merge pull request from hjelmn/thread_multiple
mpi_init: move opal_set_using_threads() earlier in MPI_Init()
2016-04-26 14:04:49 -06:00
Nathan Hjelm
1e4daa2a0e mpi_init: move opal_set_using_threads() earlier in MPI_Init()
There is a potential race condition in MPI_Init() where an orte even
thread could be in a function that uses OPAL_THREAD_LOCK /
OPAL_THREAD_UNLOCK when ompi_mpi_init calls opal_set_using_threads().

Closes 

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-04-26 13:02:42 -06:00
Ralph Castain
02876564d4 Silence warning of zero-byte malloc 2016-04-26 11:55:59 -07:00
Nathan Hjelm
c16e639b2f Merge pull request from hjelmn/ompi_coverity
ompi coverity fixes
2016-04-26 09:17:48 -06:00
rhc54
9511e38691 Merge pull request from ggouaillardet/poc/ORTE_RML_SEND_COMPLETE
orte/rml: ORTE_RML_SEND_COMPLETE handles messages with both NULL iov …
2016-04-25 19:54:58 -07:00
Gilles Gouaillardet
6bf57c799f orte/rml: ORTE_RML_SEND_COMPLETE handles messages with both NULL iov and cbfunc.buffer 2016-04-26 09:19:31 +09:00
Nathan Hjelm
5612998d21 memory/patcher: cast away const in shmdt hook
The opal_mem_hooks_release_hook does not have const on the pointer
(though it probably should). This commit eliminates a warning by
casting away the const until opal_mem_hooks_release_hook is updated to
use const.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-04-25 15:32:11 -06:00
Jeff Squyres
8ab88f2051 ompi_mpi_finalize: add/update comments
This is a follow-on to open-mpi/ompi@7373111: add some comments
explaining why the code is the way it is.  Also update a previous
comment.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-25 13:42:30 -07:00
Jeff Squyres
78b367eb0d memory patcher: add some clarifying comments
This is complicated stuff: add some comments so that future
maintainers have some rationale to understand the way things have been
done.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-25 13:12:02 -07:00
Geoffrey Paulsen
55a15fb1d0 Missed one IBM Copyright message for contributions in memory patcher component 2016-04-25 15:47:15 -04:00
Geoffrey Paulsen
ed6f508735 Updated IBM Copyright message for contributions in memory patcher component. 2016-04-25 15:13:38 -04:00
Ralph Castain
abeb7b80fb Remove unneeded NEWS item 2016-04-25 11:51:38 -07:00
rhc54
da132f7e79 Merge pull request from rhc54/topic/finalize
Somehow, the logic for finalize got lost, so restore it here. If pmix…
2016-04-25 09:13:57 -07:00
Jeff Squyres
4a5fdd5083 README: fix spelling of "schizo" framework name
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-25 09:02:34 -07:00
Jeff Squyres
3cf9981f08 README: add one more missing framework to the list
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-25 08:37:56 -07:00
Jeff Squyres
73f7ce624a README: update list of frameworks
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-04-25 08:27:47 -07:00
Ralph Castain
7373111662 Somehow, the logic for finalize got lost, so restore it here. If pmix.fence_nb is available, then call it and cycle opal_progress until complete. If pmix.fence_nb is not available, then do an MPI_Barrier and call pmix.fence.
Needs to go over to 2.x
2016-04-25 08:04:35 -07:00
Karol Mroz
941f2c1e0b oshmem: fixup hostname max length usage
Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-04-25 07:08:23 +02:00
Karol Mroz
bb6ee605bf test: fixup hostname max length usage
Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-04-25 07:08:23 +02:00
Karol Mroz
5c11bdb251 orte: fixup hostname max length usage
Also removes orte specific max hostname value.

Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-04-25 07:08:23 +02:00
Karol Mroz
3322347da9 ompi: fixup hostname max length usage
Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-04-25 07:08:23 +02:00
Karol Mroz
e1c64e6e59 opal: standardize on max hostname length
Define OPAL_MAXHOSTNAMELEN to be either:
  (MAXHOSTNAMELEN + 1) or
  (limits.h:HOST_NAME_MAX + 1) or
  (255 + 1)

For pmix code, define above using PMIX_MAXHOSTNAMELEN.

Fixup opal layer to use the new max.

Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-04-24 08:19:47 +02:00
Ralph Castain
9d0e306fa0 Update 1.10.3 NEWS items 2016-04-23 19:49:04 -07:00
Nathan Hjelm
ae0ffbb67f Merge pull request from hjelmn/enable_thread_multiple
ompi: always enable MPI_THREAD_MULTIPLE support
2016-04-23 08:40:22 -06:00
Jeff Squyres
628e638a50 Merge pull request from jsquyres/pr/fix-usnic-resource-check
usnic: fix resource check
2016-04-22 20:08:00 -04:00