[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] CPU Core detection?
- Date: Mon, 1 Aug 2011 15:15:17 -0400
- From: Chi Chan <chichan2008@xxxxxxxxx>
- Subject: Re: [Condor-users] CPU Core detection?
Condor might use the hwloc (Portable Hardware Locality library) to
detect hardware threads, cores, sockets. Hwloc is also used by Open
Grid Scheduler (the open source version of Sun Grid Engine & Oracle
Grid Engine), Torque scheduler, and various MPI libraries (OpenMPI,
MVAPICH2 & MPICH2). IMO, it is better to let a special library deal
with all the hardware features than re-inventing the code everytime a
new CPU or CPU architecture is released.
On Mon, Aug 1, 2011 at 3:06 PM, Michael Di Domenico
> On Mon, Aug 1, 2011 at 2:42 PM, Matthew Farrellee <matt@xxxxxxxxxx> wrote:
>> You could try reading this code (mind the #ifdefs 8o)...
> Sometimes that's much easier said then done.... :)
> knowing where to look in the massive code base that is condor is helpful...
> so based on the code it looks like there's what i would consider a
> "bug" between condor's core detection and linux's core reporting
> interaction. It seems logical that adding node_id into /proc/cpuinfo
> and having condor pick that up would be a simple fix, but as the
> thread claims that patch to the kernel was rejected (i'll have to see
> why), so it's not available.
> Can anyone comment on whether the /proc/cpuinfo routines might ever
> get ported to /sys entries instead of /proc which would also alleviate
> this (the kernel thread claims is a better interface)?
>> $ _CONDOR_TOOL_DEBUG=D_ALL condor_config_val -debug
> thats a spiffy commandline trick, i'll have to note that one...
> Condor-users mailing list
> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> The archives can be found at: