Re: [HTCondor-users] Parallel Matlab in HTCondor

[Werner Hack <werner.hack@xxxxxxxxxx>]

Hi Gang,

thanks for this hint. I already have cgroups enabled.
But the problem is still the behaviour in matlab (and maybe also in other parallel applications):
matlab is not reducing the number of threads accordingly.
When matlab starts, it detects the available number of cores in the system and then spawns
that many threads, independantly what was assigned to the job by HTCondor.
If cgroups is limiting the number of cores to the job, all these threads have to share
a lesser number of cores now. This results in more theads per core and in increasing load average.

So I want to define in matlab maxNumCompThreads() accordingly to my assigned number of cores.
Does HTCondor provide any info to the Job to do this?

Best,
Werner

Am 12.12.2014 13:15, schrieb qing:

Hi， Werner:

      Condor cgroups can limit the usage of cpus at job level, but it works only when there is contention for the cpu.

   Cheers,Gang

On 12/12/2014 11:52, Werner Hack wrote:

Hi all,

we are running a lot of matlab jobs here.
Due to the fact matlab attempts to allocate all cpu cores on a node, we limit cpu usage
to a single core using the matlab command option "-singleCompThread" or
the function maxNumCompThreads(1) in matlabrc.
This is Ok for most cases.
But what can I do, if I want to use more cores/threads (request_cpus >1) in matlab?
Can HTCondor provide any info to the matlab job (environment variable? )
about the assigned request_cpus?
If I check all environment variables in a job I can not see such info.
Any ideas?

Regards,
Werner

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature