[Openmp-dev] Recommendation for building with -DLIBOMP_USE_HWLOC=ON

[Churbanov, Andrey via Openmp-dev]

Hi Tom,

The hwloc topology discovery code is proven to be slower, but can give some NUMA awareness features.

For example on Intel Xeon Phi (KNL) it allows to bind threads to tiles (pairs of cores with shared L2 cache); and it allows to limit program affinity mask to particular NUMA node(s) inside package using KMP_HW_SUBSET environment variable.

If you don't need mentioned features than legacy topology code will work similarly but with faster initialization time.

Note also that the code base does not have hwloc library, so you will need to provide it for the build in order to use it.

Regards,
Andrey

-----Original Message-----
From: Openmp-dev [mailto:openmp-dev-bounces at lists.llvm.org] On Behalf Of Tom Stellard via Openmp-dev
Sent: Friday, March 30, 2018 7:31 AM
To: openmp-dev (openmp-dev at lists.llvm.org) <openmp-dev at lists.llvm.org>
Subject: [Openmp-dev] Recommendation for building with -DLIBOMP_USE_HWLOC=ON

Hi,

What's the general recommendation for building openmp with -DLIBOMP_USE_HWLOC=ON ?  Is hwloc better than the default topology code?

Thanks,
Tom
_______________________________________________
Openmp-dev mailing list
Openmp-dev at lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev

--------------------------------------------------------------------
Joint Stock Company Intel A/O
Registered legal address: Krylatsky Hills Business Park,
17 Krylatskaya Str., Bldg 4, Moscow 121614,
Russian Federation

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.