[gridengine users] set processor affinity inside gridengine

Reuti reuti at staff.uni-marburg.de
Wed May 1 15:09:07 UTC 2013


Am 30.04.2013 um 23:58 schrieb Dave Love:

> Reuti <reuti at staff.uni-marburg.de> writes:
> 
>>> Does anyone know a way to define this option in the gridengine directly ? I'm using ogs 2011.11p1 on Intel 6core processors.
>> 
>> With 6 cores per machine I would simply request full nodes.
> 
> [I guess they're two-socket; we don't insist on filling those, even for
> MPI jobs.]
> 
>> AFAIK the commercial GridEngine from Univa contains extensions to
>> monitor granted cores on a host level, so that each of such a job
>> could be given an unique list of cores. If any KMP_... variable would
>> allow a syntax to accept a list of cores, it should work that your
>> executable can make a proper allocation. But such a thing is not
>> available in the open source version.
> 
> How does it pass binding information to the individual MPI processes?

It's up to the application or the starting script to assign the ranks to the given cores after parsing the generated hostfile when the queues are named with a suffix of the core id.

-- Reuti


> The free software version works adequately for pure MPI or OpenMP jobs,
> just not mixed ones, assuming that's what the question is about.  With
> -binding linear:slots (and no over-subscription, of course) job tasks
> get a binding and can set GOMP_CPU_AFFINITY (or equivalent KMP_AFFINITY,
> presumably) from SGE_BINDING.  However, it's the same for all processes
> started by the one shepherd.
> 
> This just came up on the openmpi list.
> 
> -- 
> Community Grid Engine:  http://arc.liv.ac.uk/SGE/





More information about the users mailing list