parallel computing: matlab only uses half the cores available. Help?
8 vues (au cours des 30 derniers jours)
Afficher commentaires plus anciens
Luis Hernandez
le 2 Nov 2016
Modifié(e) : Dylan Gomes
le 6 Mar 2023
Hello,
I have a machine with 40 cores and I'm trying to use them all for a job. I have set the preferred number of workers to 40 in the Preferences, but I can see that matlab is only using 20 workers. Any suggestions about how to fix this? Thanks a lot
-L
2 commentaires
Adam
le 2 Nov 2016
Does it have 40 actual cores or is it 20 with hyper-threading? I know next to nothing about hyper-threading, but I know my machine has 28 physical cores, but 56 show up in my task manager due to hyper threading. Matlab only uses the 28 though.
Mujahid Sultan
le 13 Août 2017
Modifié(e) : Mujahid Sultan
le 14 Août 2017
I have the same issue, "nproc" shows 56 where as Matlab only uses 28 cpus, features('nucores') shows 28 cores. Interestingly the map-reduce function of Matlab uses all 56 cores and returns only results for 28. I am loosing half of my computations somewhere.
Réponse acceptée
Kojiro Saito
le 2 Nov 2016
On MATLAB,
- Go to Parallel Menu -> Manage Cluster Profiles
- Edit your cluster profile (name is 'local')
- Input 40 (the number of your cores) in "NumWorkers" column and click "Done"
Thanks
0 commentaires
Plus de réponses (1)
Luis Hernandez
le 2 Nov 2016
5 commentaires
Kojiro Saito
le 7 Nov 2016
@Luis, could you provide me the version of your MATLAB?
Also, if you can figure out the difference between your settings, please let us know.
Thanks,
Dylan Gomes
le 6 Mar 2023
Modifié(e) : Dylan Gomes
le 6 Mar 2023
I thought I was having the same issue with MATLAB R2021a, but it turned out that I only actually have 6 physical cores, when I thought I had 12.
>> feature('numCores')
%
% MATLAB detected: 6 physical cores.
% MATLAB detected: 12 logical cores.
% MATLAB was assigned: 12 logical cores by the OS.
% MATLAB is using: 6 logical cores.
% MATLAB is not using all logical cores because hyper-threading is enabled.
Voir également
Catégories
En savoir plus sur MATLAB Parallel Server dans Help Center et File Exchange
Produits
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!