ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
29 147 коммитов 3 Ветки 0 Теги
Граф коммитов

167 Коммитов

Автор SHA1 Сообщение Дата
Gilles Gouaillardet
44c1ff60f1 Merge pull request #2672 from ggouaillardet/topic/misc_memory_leaks 
Plug misc memory leaks
 2017-01-10 13:16:04 +09:00
Ralph Castain
e25e69dc2f Resolve Coverity issues 
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2017-01-07 10:45:52 -08:00
Gilles Gouaillardet
6b90b03c28 orted/pmix: plug a memoy leak in pmix_server_fencenb_fn() 
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
 2017-01-06 15:38:45 +09:00
Ralph Castain
6509f60929 Complete the memprobe support. This provides a new scaling tool called "mpi_memprobe" that samples the memory footprint of the local daemon and the client procs, and then reports the results. The output contains the footprint of the daemon on each node, plus the average footprint of the client procs on that node. 
Samples are taken after MPI_Init, and then again after MPI_Barrier. This allows the user to see memory consumption caused by add_procs, as well as any modex contribution from forming connections if pmix_base_async_modex is given.

Using the probe simply involves executing it via mpirun, with however many copies you want per node. Example:

$ mpirun -npernode 2 ./mpi_memprobe
Sampling memory usage after MPI_Init
Data for node rhc001
	Daemon: 12.483398
	Client: 6.514648

Data for node rhc002
	Daemon: 11.865234
	Client: 4.643555

Sampling memory usage after MPI_Barrier
Data for node rhc001
	Daemon: 12.520508
	Client: 6.576660

Data for node rhc002
	Daemon: 11.879883
	Client: 4.703125

Note that the client value on node rhc001 is larger - this is where rank=0 is housed, and apparently it gets a larger footprint for some reason.

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2017-01-05 10:32:17 -08:00
Ralph Castain
91d714fe93 Add flags to direct PMIx to only use one listener, but without directing which one (tcp or usock) to use. This allows the user to set PMIX_MCA_ptl in their environment to select the transport method. 
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2017-01-04 09:16:44 -08:00
Ralph Castain
9eab9a1ed3 Remove stale global variables 
Revamp the event notification integration to rely on the PMIx event chaining and remove the duplicate chaining in OPAL. This ensures we get system-level events that target non-default handlers.

Restore the hostname entries for MPI-level error messages, but provide an MCA param (orte_hostname_cutoff) to remove them for large clusters where the memory footprint is problematic. Set the default at 1000 nodes in the job (not the allocation).

Begin first cut at memory profiler

Some minor cleanups of memprobe

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2017-01-02 14:04:24 -08:00
Ralph Castain
fe68f23099 Only instantiate the HWLOC topology in an MPI process if it actually will be used. 
There are only five places in the non-daemon code paths where opal_hwloc_topology is currently referenced:

* shared memory BTLs (sm, smcuda). I have added a code path to those components that uses the location string
  instead of the topology itself, if available, thus avoiding instantiating the topology

* openib BTL. This uses the distance matrix. At present, I haven't developed a method
  for replacing that reference. Thus, this component will instantiate the topology

* usnic BTL. Uses the distance matrix.

* treematch TOPO component. Does some complex tree-based algorithm, so it will instantiate
  the topology

* ess base functions. If a process is direct launched and not bound at launch, this
  code attempts to bind it. Thus, procs in this scenario will instantiate the
  topology

Note that instantiating the topology on complex chips such as KNL can consume
megabytes of memory.

Fix pernode binding policy

Properly handle the unbound case

Correct pointer usage

Do not free static error messages!

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-29 10:33:29 -08:00
Ralph Castain
52533f755e Remove debug 
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-28 13:24:39 -08:00
Ralph Castain
3a2d6a5ab6 Begin to reduce reliance of application procs on the topology tree itself by having the daemon provide more detailed info. In this case, provide the topology description string so that procs can readily determine the number of types of objects on the node, and a "locality" string that describes which objects this process is executing upon. The latter allows a process to compute the objects of overlap between itself and another proc without consulting the topology tree. 
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-28 09:14:26 -08:00
Ralph Castain
ea133206ec Sync the internal OMPI component to PMIx master 
Update external PMIx v2.x component
Add missing Makefile

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-19 19:14:16 -08:00
Ralph Castain
c6f6f40529 Transfer debugger support changes 
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-17 18:14:46 -08:00
Ralph Castain
269753f5c1 Transfer back changes from debugger attach work 
Silence warning

Remove debug

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-17 10:00:52 -08:00
Ralph Castain
884fb7fcf2 Update the PMIx2 support to include the latest shared memory optimizations 
Update ORTE support for dynamic PMIx operations e.g., PMIx_Spawn
Update to track master
Ensure that --disable-pmix-dstore actually disables the dstore. Sync to a few debugger updates

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-14 15:00:10 -08:00
rhc54
341ab683de Merge pull request #2532 from rhc54/topic/pmixptl 
Update to latest PMIx master + PTL branch
 2016-12-07 17:28:22 -08:00
Ralph Castain
e1aa7939ef Correctly cleanup the local children and node map info on remote orteds upon job completion. Ensure that register_nspace only includes procs from that job in the proc map 
Thanks to Ashley Pittman for the report

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-07 13:53:00 -08:00
Ralph Castain
fbed2d794a Update to latest PMIx master + PTL branch 
Update the usock component to disable it

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-06 20:47:44 -08:00
Ralph Castain
79cde184ad Allow a PMIx tool to spawn a job 
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
 2016-12-03 16:00:47 -08:00
Ralph Castain
649301a3a2 Revise the routed framework to be multi-select so it can support the new conduit system. Update all calls to rml.send* to the new syntax. Define an orte_mgmt_conduit for admin and IOF messages, and an orte_coll_conduit for all collective operations (e.g., xcast, modex, and barrier). 
Still not completely done as we need a better way of tracking the routed module being used down in the OOB - e.g., when a peer drops connection, we want to remove that route from all conduits that (a) use the OOB and (b) are routed, but we don't want to remove it from an OFI conduit.
 2016-10-23 21:52:39 -07:00
Ralph Castain
5b1484a836 Implement the backend support for process-generated event notification 2016-10-08 09:24:28 -07:00
Ralph Castain
e773c17cf3 Put show_help thru the PMIx "log" API. This pushes the show_help output from apps into the pmix thread, thus avoiding conflicts in the RML thread, which should help with thread lock situations. 2016-10-02 16:02:23 -07:00
Artem Polyakov
81195ab724 Several fixes related to session directories: 
* enable OMPI to retrieve paths from RM through PMIx
* cleanups related to tempdirs.
 2016-09-05 07:48:44 +03:00
Ralph Castain
0ea1cff733 Implement notification of completion on comm_spawn'd child jobs. Add a configure flag to enable PMIx 3's shared memory datastore, and set it disable by default so that comm_spawn functions again. Will reverse the default once that feature is fully functional 2016-09-01 13:10:10 -07:00
Ralph Castain
2f6e0fec90 Provide the number of nodes in the job 2016-08-26 14:50:41 -07:00
Ralph Castain
ae2af61ee3 Update the session dir structure. Restore the creation of a top-level dir based on userid so that everything is contained under the user's top-level dir. Make the next level down (the "job family" level) be either the pid (indicated by a name of "pid.N") or the job family if not launched by mpirun. This allows for proper rendezvous by direct-launched procs. 2016-08-15 22:46:46 -05:00
Ralph Castain
be8424b691 Provide backward compatible keys so that the non-PMIx components in the opal/pmix framework don't have to adjust as we continue to work on finalizing the PMIx reference scheme. Activate and utilize the new PMIx show_help capability to provide more meaningful error output when the server cannot start. 
Add a contrib script to cleanup permissions incorrectly modified due to things like smb mounts

dd
 2016-08-13 12:13:04 -07:00
Ralph Castain
08a0644df5 Fix shared memory rendezvous 2016-08-13 08:14:50 -07:00
Ralph Castain
d4327fd973 The node index isn't normally passed with the packed node object, so we need to set it on the remote end as the orted needs to pass it down to the procs. Refactor the registration code to better package proc-level info - we will separate out the node and app levels in a subsequent change. 2016-08-12 12:06:23 -07:00
Ralph Castain
527b5c692a Update to include extended tool support, new datatypes 2016-08-08 13:39:46 -07:00
Ralph Castain
16fccd4964 Establish a way for ORTE to tell PMIx the base tmpdir to use, and update PMIx to understand such directives 2016-07-29 09:52:36 -07:00
Ralph Castain
b748afceb1 Fix copy/paste error 2016-07-29 06:41:30 -07:00
Gilles Gouaillardet
e67c3d0a14 orted/pmix: protect against NULL node in orte_pmix_server_register_nspace() 2016-07-29 16:20:31 +09:00
Ralph Castain
cacb582ecd Support timeout values when performing connect/accept operations. Bump default timeout to 10 minutes so folks have time to start the partnering application 2016-07-28 14:09:06 -07:00
Ralph Castain
9ab20cafe3 Pass the nodeid for each proc in the job. Fix a mistaken error output message 2016-07-25 15:41:15 -07:00
Ralph Castain
ee56d9dc1a Shorten the session directory name as some OS's are now providing unusually long temp directory names, causing us to overflow the sockaddr field 2016-07-05 14:59:50 -07:00
Ralph Castain
c9ada8e095 Silence Coverity warnings 2016-07-03 20:45:08 -07:00
Ralph Castain
6e434d6785 Add support for PMIx tool connections and queries. Initially only support a request to list all known namespaces (jobids) from ORTE, but other folks will extend that support to include additional information 
Update to match PMIx RFC

Fix configury to point to correct libevent and hwloc locations
 2016-06-29 19:19:19 -07:00
Gilles Gouaillardet
5d32282230 orted/pmix_server_pub: fix packing type in pmix_server_lookup_fn() 
and make it match the one used when unpacking in orte_data_server()
 2016-06-27 14:37:08 +09:00
Ralph Castain
e3e4d73986 Need to be a little more careful when checking the range on a publish/lookup operation. If the range was constrained at publish, then we need to check that the lookup fits within that constraint. Otherwise, we should provide the data. More detailed constraint checking will be provided later. 2016-06-24 17:01:49 -07:00
Ralph Castain
0ba02821e6 Add requested key and job-level info 2016-06-19 18:22:31 -07:00
Ralph Castain
5d330d5220 Enable the PMIx event notification capability and use that for all error notifications, including debugger release. This capability requires use of PMIx 2.0 or above as the features are not available with earlier PMIx releases. When OMPI master is built against an earlier external version, it will fallback to the prior behavior - i.e., debugger will be released via RML and all notifications will go strictly to the default error handler. 
Add PMIx 2.0

Remove PMIx 1.1.4

Cleanup copying of component

Add missing file

Touchup a typo in the Makefile.am

Update the pmix ext114 component

Minor cleanups and resync to master

Update to latest PMIx 2.x

Update to the PMIx event notification branch latest changes
 2016-06-14 13:08:41 -07:00
Ralph Castain
dd0f843843 Fix rare hangs observed on OS-X by properly thread-shifting upcalls from the PMIx server into ORTE 2016-06-05 21:39:44 -07:00
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
rhc54
2fa8b6c6ac Merge pull request #1525 from rhc54/topic/schizo 
Extend the schizo framework
 2016-05-01 15:09:08 -07:00
Ralph Castain
6ac7929bd0 Extend the schizo framework to allow definition of CLI options by environment. Refactor orterun to mesh with the orted_submit code, thus improving code reuse. Eliminate the orte-submit tool as orterun can now meet that need. 
Cleanups per @jjhursey review
 2016-05-01 11:30:25 -07:00
Ralph Castain
0f05893952 Ensure consistency between max_procs and univ_size values - since orte wants max_procs, have the proc get that value instead of univ_size 
Make the singleton module consistent as well
 2016-05-01 11:13:33 -07:00
Rainer Keller
52080a5736 As per the pull request to pmix/master: 
https://github.com/pmix/master/pull/71

Have OMPI's current version of pmix120 nicely fail in case of
too long sun_path (longer than 108 or in case of OSX 103 chars).
And have OMPI return proper error messages with hints how to
amend.
 2016-04-07 22:12:53 +02:00
Gilles Gouaillardet
2ede47c462 pmix: fix misc missing conversion and type issues 2016-04-04 10:12:34 +09:00
Ralph Castain
60a7bc2e50 Enable the PMIx notification callback system. This currently is only supported by the pmix120 component, which is not selected by default. All other components will ignore error registration requests, and thus do not support debugger attach when launched via mpirun. Note that direct launched applications will support such attachment, but may not do so in a scalable fashion. 
Fixes ##1225
 2016-02-18 09:29:12 -08:00
Ralph Castain
8f9508cace Further enhance the support for Singularity containers. Extend the "personality" command-line option to allow specifying both model (e.g., "ompi") and container (e.g., "singularity"), and add the necessary logic to support multiple options. Add a new pmix "isolated" component to handle singletons where no HNP is available since containers cannot launch the HNP. 2016-02-17 13:33:06 -08:00
Ralph Castain
fc6b260146 Protect against PMIx-based requests that don't come thru the MPI comm_spawn interface 2016-01-16 13:36:06 -08:00