1
1
Граф коммитов

25069 Коммитов

Автор SHA1 Сообщение Дата
Gilles Gouaillardet
5ec1eedbae Merge pull request #1682 from ggouaillardet/topic/fix-ethtool-again
opal/util/ethtool: use system ethtool_cmd_speed when available
2016-05-20 10:30:43 +09:00
Gilles Gouaillardet
cbbdce05b1 pmix/pmix114: silence a warning 2016-05-20 09:35:26 +09:00
Gilles Gouaillardet
ed3fd1775f rcache/grdma: silence a warning 2016-05-20 09:30:29 +09:00
Gilles Gouaillardet
a01a5487a8 opal/util/ethtool: use system ethtool_cmd_speed when available
Refs: open-mpi/ompi#1679
2016-05-20 09:05:09 +09:00
rhc54
99d3c283f5 Merge pull request #1681 from rhc54/topic/pmixupdate
Update PMIx 114 to current release candidate
2016-05-19 13:50:16 -07:00
Ralph Castain
6f743f81b6 Update PMIx 114 to current release candidate 2016-05-19 12:55:05 -07:00
Jeff Squyres
110755be37 Merge pull request #1679 from jsquyres/pr/fix-ethtool
ethtool: better handle portability
2016-05-19 15:09:51 -04:00
rhc54
fcf62a8708 Merge pull request #1678 from rhc54/topic/mutex
Unlock the mutex prior to destructing it.
2016-05-19 11:48:26 -07:00
Jeff Squyres
87233aae49 ethtool: better handle portability
Be sure to handle the case where we don't have ethtool support at all.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-05-19 10:57:14 -07:00
Ralph Castain
a35bb8453a Unlock the mutex prior to destructing it.
Thanks to Nicolas Joly for the report
2016-05-19 10:36:58 -07:00
George Bosilca
50b37758d4 Don't overwrite the function argument.
In a MPMD setup the app in the jdata can be NULL, so make sure we
don't leave the main argument to an inconsistent value.
2016-05-19 10:35:23 -04:00
Gilles Gouaillardet
fd93d236b1 opal/util/ethtool: fix compilation on older Linux when struct ethtool_cmd has no speed_hi field
Refs: open-mpi/ompi#1628
2016-05-19 11:58:04 +09:00
Jeff Squyres
66f53ec29a Merge pull request #1628 from kmroz/wip-btl-tcp-ethtool-speed
btl/tcp: autodetect bandwidth and latency if unset by the user
2016-05-18 12:12:55 -04:00
Nathan Hjelm
9371a6a52d Merge pull request #1673 from hjelmn/fix_rcache_deadlock
rcache: fix deadlock in multi-threaded environments
2016-05-18 08:32:21 -07:00
Karol Mroz
ca6ddf3270 btl/tcp: autodetect bandwidth and latency if unset
Fixes open-mpi/ompi#120

Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-05-18 16:25:52 +02:00
Karol Mroz
b9c6c43c6b btl/tcp: add default defines for bandwidth and latency
Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-05-18 16:25:52 +02:00
Karol Mroz
31e33a64f9 opal/util: add function to obtain interface speed
If kernel ethtool_cmd_speed() is not available, use copies if possible.

Signed-off-by: Karol Mroz <mroz.karol@gmail.com>
2016-05-18 16:25:51 +02:00
Nathan Hjelm
ab8ed177f5 rcache: fix deadlock in multi-threaded environments
This commit fixes several bugs in the registration cache code:

 - Fix a programming error in the grdma invalidation function that can
   cause an infinite loop if more than 100 registrations are
   associated with a munmapped region. This happens because the
   mca_rcache_base_vma_find_all function returns the same 100
   registrations on each call. This has been fixed by adding an
   iterate function to the vma tree interface.

 - Always obtain the vma lock when needed. This is required because
   there may be other threads in the system even if
   opal_using_threads() is false. Additionally, since it is safe to do
   so (the vma lock is recursive) the vma interface has been made
   thread safe.

 - Avoid calling free() while holding a lock. This avoids race
   conditions with locks held outside the Open MPI code.

Fixes open-mpi/ompi#1654.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2016-05-17 09:02:40 -06:00
Nathan Hjelm
f6938868bd Merge pull request #1659 from hjelmn/sync_64
sync_builtin: check for 64-bit atomic support
2016-05-17 05:40:04 -07:00
rhc54
8b534e9897 Merge pull request #1668 from rhc54/topic/slurm
When direct launching applications, we must allow the MPI layer to pr…
2016-05-16 12:23:19 -07:00
Howard Pritchard
9066d26c67 Merge pull request #1672 from hppritcha/topic/fix_cray_pmix_bust
pmix/cray: fix some breakage
2016-05-16 20:06:51 +01:00
Jeff Squyres
daa26f4578 Merge pull request #1671 from jsquyres/pr/fix-bml-opal-output-func-names
bml_r2: use __func__ to identify function names
2016-05-16 14:17:04 -04:00
Howard Pritchard
1a676e5b35 pmix/cray: fix some breakage
Signed-off-by: Howard Pritchard <howardp@lanl.gov>
2016-05-16 12:45:05 -05:00
Jeff Squyres
5275e5e2a1 bml_r2: use __func__ to identify function names
There were some old/stale function names in some debugging/verbose
opal_output calls.  Use __func__ instead, so that they won't become
stale in the future.

Thanks to Durga Choudhury for pointing out the issue.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-05-16 11:06:47 -04:00
Gilles Gouaillardet
4e21933a74 memory/patcher: declare __curbrk as extern in order not to generate an (unitialized) common symbol 2016-05-16 09:30:11 +09:00
rhc54
e0fef390a8 Merge pull request #1670 from rhc54/topic/dashhost
In MPMD case, add slots given to each of the executables instead of overwriting
2016-05-15 13:51:54 -07:00
Ralph Castain
ca69403cc8 In MPMD case, add slots given to each of the executables instead of overwriting 2016-05-15 08:55:43 -07:00
Ralph Castain
01ba861f2a When direct launching applications, we must allow the MPI layer to progress during RTE-level barriers. Neither SLURM nor Cray provide non-blocking fence functions, so push those calls into a separate event thread (use the OPAL async thread for this purpose so we don't create another one) and let the MPI thread sping in wait_for_completion. This also restores the "lazy" completion during MPI_Finalize to minimize cpu utilization.
Update external as well

Revise the change: we still need the MPI_Barrier in MPI_Finalize when we use a blocking fence, but do use the "lazy" wait for completion. Replace the direct logic in MPI_Init with a cleaner macro
2016-05-14 16:37:00 -07:00
Aurélien Bouteiller
7f65c2b18e forgot to update copyright in commits 627a89b 4899c89 2016-05-13 11:34:59 -04:00
George Bosilca
37e03e3e5b Don't update req_bytes_received if no bytes were received. 2016-05-12 23:39:32 -04:00
Gilles Gouaillardet
456b73da69 btl/openib: fix error path in init_one_device()
do not explicitly release ib verbs components since they will
be released in the object destructor

Thanks Durga for the report
2016-05-13 09:03:48 +09:00
rhc54
4d026e223c Merge pull request #1661 from matcabral/master
PSM and PSM2 MTLs to detect drivers and link
2016-05-11 17:43:17 -07:00
Matias A Cabral
40717198c6 add glob.h check fro psm and psm2 mtls 2016-05-11 13:05:43 -07:00
George Bosilca
f8facb177d atomically update the refcount on the datatype args. 2016-05-11 12:40:18 -04:00
rhc54
055a3add73 Merge pull request #1667 from rhc54/topic/pmixext
Ensure we fail to configure if external PMIx was requested and is not found.
2016-05-11 08:56:16 -07:00
Ralph Castain
9a5ef60602 Ensure we fail to configure if external PMIx was requested and is not found. 2016-05-11 07:59:05 -07:00
Matias A Cabral
528abff6ae Merge remote-tracking branch 'upstream/master' 2016-05-10 15:42:08 -07:00
Jeff Squyres
30f913f217 Merge pull request #1652 from jsquyres/pr/remove-aix-timer
timer/aix: remove stale code
2016-05-10 15:47:02 -04:00
Jeff Squyres
5aa38cc300 Merge pull request #1644 from ggouaillardet/topic/two_proc_algo_error
coll/tuned: two_proc algo errors when forced on comm size > 2
2016-05-10 11:54:15 -04:00
Jeff Squyres
3706c955cc Merge pull request #1663 from jsquyres/pr/hwloc-wrapper-extra-flags-fix
hwloc/external: set WRAPPER_EXTRA_* vars in proper location
2016-05-10 11:51:33 -04:00
Jeff Squyres
eccf0ff4cd hwloc/external: set WRAPPER_EXTRA_* vars in proper location
WRAPPER_EXTRA flags are checked *before* the POST_CONFIG macro is
invoked.  So set them in the main CONFIG macro.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2016-05-10 07:34:56 -07:00
Nathan Hjelm
b2f33bc076 opal/asm: fall back on inline asm atomics in some cases
This commit changes the asm configure logic to fall back on inline asm
atomics on systems that 1) have __sync atomics, 2) do not have 64-bit
__sync atomics, and 3) support 64-bit asm.

Signed-off-by: Nathan Hjelm <hjelmn@me.com>
2016-05-10 04:27:58 -06:00
Gilles Gouaillardet
ef3ee027b0 autogen: patch configure in order to correctly detect Solaris Studio 12.5 beta compilers
previously: f90: Sun Fortran 95 8.7 Linux_i386 2014/10/20
now : Studio 12.5 Fortran 95 8.8 Linux_i386 Beta 2015/11/17
note the "Sun" branding has gone

Thanks Paul Hargrove for the report and the patch
2016-05-10 16:28:33 +09:00
Matias A Cabral
d28ee62a96 Update in PSM and PSM2 MTLs to detect entries created by drivers for
Intel TrueScale and Intel OmniPath, and detect a link in ACTIVE state.
This fix addresses the scenario reported in the below OMPI users email,
including formerly named Qlogic IB, now Intel True scale. Given the
nature of the PSM/PSM2 mtls this fix applies to OmniPath:
https://www.open-mpi.org/community/lists/users/2016/04/29018.php
2016-05-09 12:08:44 -07:00
Josh Hursey
44d95cb610 Merge pull request #1657 from bgoglin/hwloc-for-2.0
configure: check the actual may_alias syntax that we use
2016-05-09 13:37:08 -05:00
Ralph Castain
7767882346 Per user request, add some missing data and definitions:
OPAL_PMIX_UNIV_RANK - synonym for OPAL_PMIX_GLOBAL_RANK
OPAL_PMIX_APP_SIZE - #ranks in the application of this proc
2016-05-09 08:39:01 -07:00
Nathan Hjelm
d99a9786b6 sync_builtin: check for 64-bit atomic support
This commit adds an additional check for 64-bit atomic support for __sync
builtins. If 64-bit support is not available the opal_atomic_*_64 atomics
are disabled.

Signed-off-by: Nathan Hjelm <hjelmn@me.com>
2016-05-09 03:17:51 -06:00
Gilles Gouaillardet
0a19337371 coll/base: return MPI_ERR_UNSUPPORTED_OPERATION when coll_base_*_two_procs algo is used on a communicator that has no two tasks
Thanks Dave Love for the report
2016-05-09 14:18:40 +09:00
Brice Goglin
6839d928c2 configure: check the actual may_alias syntax that we use
xlc 13.1.0 crashes because of our may_alias attributes in nolibxml.c
on Power7. libxml.c and nolibxml.c are the only may_alias users for now,
so change our configure check to match the actual code using it.

Thanks to Paul Hargrove for reporting and debugging the issue,
and providing the patch.

https://www.open-mpi.org/community/lists/devel/2016/05/18918.php

(cherry picked from open-mpi/hwloc@0ab7af5e90)
2016-05-08 22:22:30 +02:00
Ralph Castain
1911d74095 Prevent segfault when -debug given to mpirun 2016-05-08 10:19:05 -07:00