Add new PSM2 MTL.

[afriedle-intel]

This new MTL supports PSM2 for Omni Path. PSM2 is a descendant of PSM
with changes to support more processes and some MPI-3 features like mprobe.

PSM2 will only support Omni Path networks; PSM only supports True Scale.
Likewise, the existing PSM MTL will continue to be maintained for True
Scale, while the PSM2 MTL is developed and maintained for Omni Path.

[mellanox-github]

Refer to this link for build results (access rights to CI server needed):
http://bgate.mellanox.com/job/gh-ompi-master-pr/639/

Build Log
last 50 lines

[...truncated 48512 lines...]
640      1.538933
704      1.564410
768      1.579106
832      1.580462
896      1.613407
960      1.618928
1024     1.635517
+ '[' -n '' ']'
+ for exe in latency_th bw_th message_rate_th
+ exe_path=/var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/ompi_install1/thread_tests/thread-tests-1.1/latency_th
+ PATH=/var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/ompi_install1/bin:/hpc/local/bin::/usr/local/bin:/bin:/usr/bin:/usr/sbin:/hpc/local/bin:/hpc/local/bin/:/hpc/local/bin/:/sbin:/usr/sbin:/bin:/usr/bin:/usr/local/sbin:/opt/ibutils/bin
+ LD_LIBRARY_PATH=/var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/ompi_install1/lib:
+ mpi_runner 2 /var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/ompi_install1/thread_tests/thread-tests-1.1/latency_th 8
+ local np=2
+ local exe_path=/var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/ompi_install1/thread_tests/thread-tests-1.1/latency_th
+ local exe_args=8
+ local 'common_mca=-bind-to core'
+ local 'mca=-bind-to core'
+ '[' no == yes ']'
+ '[' yes == yes ']'
+ timeout -s SIGKILL 10m mpirun -np 2 -bind-to core -mca pml ob1 -mca btl self,sm /var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/ompi_install1/thread_tests/thread-tests-1.1/latency_th 8
Size (bytes)     Time (us)
0    12.584484
16   94.967711
32   117.122708
48   28.904879
64   272.144444
80   83.756944
96   171.889636
112      255.150686
128      75.062764
144      218.371365
160      403.257619
176      208.569274
[jenkins01:12471] *** An error occurred in MPI_Recv
[jenkins01:12471] *** reported by process [140735849824257,0]
[jenkins01:12471] *** on communicator MPI_COMM_WORLD
[jenkins01:12471] *** MPI_ERR_INTERN: internal error
[jenkins01:12471] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
[jenkins01:12471] ***    and potentially your MPI job)
Build step 'Execute shell' marked build as failure
[htmlpublisher] Archiving HTML reports...
[htmlpublisher] Archiving at BUILD level /var/lib/jenkins/jobs/gh-ompi-master-pr/workspace/cov_build to /var/lib/jenkins/jobs/gh-ompi-master-pr/builds/639/htmlreports/Coverity_Report
Setting commit status on GitHub for https://github.com/open-mpi/ompi/commit/f2aa01c684542af94da4b5005ce9ed0590964b4e
[BFA] Scanning build for known causes...
[BFA] No failure causes found
[BFA] Done. 0s
Setting status of 2c9be59b3726d27c21312ca12c485f83ccc67f58 to FAILURE with url http://bgate.mellanox.com:8888/jenkins/job/gh-ompi-master-pr/639/ and message: 'Build finished.'
Using conext: Mellanox

[rhc54]

@miked-mellanox We keep seeing this failure recently - is it bogus?

[mike-dubman]

the failure seems unrelated to your commit but real.

it seems that ob1 and sm have issues with threading support.

@hjelmn - could you please take a look?

[hppritcha]

I"m not sure for the 1.10.x that we are claiming support for mpi thread multiple for ob1.

[hjelmn]

ob1 is thread safe but not all of the btls are. I know vader is thread safe but have no idea if sm is. It might be close to time to put btl/sm out to pasture.

[mike-dubman]

cool, thanks
will remove this test

[mellanox-github]

Refer to this link for build results (access rights to CI server needed):
http://bgate.mellanox.com/job/gh-ompi-master-pr/641/

[mellanox-github]

Refer to this link for build results (access rights to CI server needed):
http://bgate.mellanox.com/job/gh-ompi-master-pr/642/

[afriedle-intel]

Is it OK for me to hit the merge button, or should one of the core developers do it? Anything else that I need to address?

[rhc54]

Go for it 😀

Sent from my iPhone

On Jun 22, 2015, at 12:41 PM, Andrew Friedley notifications@github.com wrote:

Is it OK for me to hit the merge button, or should one of the core developers do it? Anything else that I need to address?

—
Reply to this email directly or view it on GitHub.