Hi all, is the binding to cores automatic with SLURM? The command --cpu-bind=cores is not recognized. I tried to force the binding by using #SBATCH -B 2:24:1 but it says that there is no such nodes available. However #SBATCH -B 2:12:1 works unexpectedly fine! Thanks! Bests, Julien -- ************************************************** Dr. Julien GUÉNOLÉ Research Group Head ---------------------- Institute of Physical Metallurgy and Metal Physics RWTH Aachen University Kopernikusstrasse 14 52074 Aachen, GERMANY ---------------------- Room 202 Phone [office] +49 241 80 26866 Email guenole@imm.rwth-aachen.de Web http://www.julien-guenole.fr Twitter @nanouayeur **************************************************
Hi Julien, SLURM will attempt sane binding under the premise that you give it the right information in your batch script. The reason it does not behave like you expect it to is the virtual sub-NUMA clustering of the Skylake CPUs which splits a single die(/socket) into two distinct NUMA nodes. Thus, you can even ask for #SBATCH -B 4:12[:1] without any problem but as soon as you try to request more than 12 cores per (virtual) socket - or more than 4 sockets for that matter -, SLURM will reject the request. Best, Sven On 3/29/19 4:40 PM, Julien Guénolé wrote:
Hi all,
is the binding to cores automatic with SLURM?
The command --cpu-bind=cores is not recognized.
I tried to force the binding by using #SBATCH -B 2:24:1 but it says that there is no such nodes available. However #SBATCH -B 2:12:1 works unexpectedly fine!
Thanks!
Bests, Julien
-- Sven Hansen IT Center Abteilung: Systeme und Betrieb RWTH Aachen University Seffenter Weg 23 52074 Aachen (Germany) Tel.: + 49 241 80-29114 s.hansen@itc.rwth-aachen.de www.itc.rwth-aachen.de
Hi Sven, thanks for the details! Best, Julien On 29/03/2019 17:21, Sven Hansen wrote:
Hi Julien,
SLURM will attempt sane binding under the premise that you give it the right information in your batch script. The reason it does not behave like you expect it to is the virtual sub-NUMA clustering of the Skylake CPUs which splits a single die(/socket) into two distinct NUMA nodes. Thus, you can even ask for
#SBATCH -B 4:12[:1]
without any problem but as soon as you try to request more than 12 cores per (virtual) socket - or more than 4 sockets for that matter -, SLURM will reject the request.
Best, Sven
On 3/29/19 4:40 PM, Julien Guénolé wrote:
Hi all,
is the binding to cores automatic with SLURM?
The command --cpu-bind=cores is not recognized.
I tried to force the binding by using #SBATCH -B 2:24:1 but it says that there is no such nodes available. However #SBATCH -B 2:12:1 works unexpectedly fine!
Thanks!
Bests, Julien
--
Dr. Julien GUÉNOLÉ Research Group Head ---------------------- Institute of Physical Metallurgy and Metal Physics RWTH Aachen University Kopernikusstrasse 14 52074 Aachen, GERMANY ---------------------- Room 202 Phone [office] +49 241 80 26866 Email guenole@imm.rwth-aachen.de Web http://www.julien-guenole.fr Twitter @nanouayeur **************************************************
participants (2)
-
Julien Guénolé -
Sven Hansen