[CP2K-user] [CP2K:17226] Advice for ELPA best use

Krack Matthias (PSI) matthias.krack at psi.ch
Fri Jun 24 10:42:50 UTC 2022


Hello Xavier

Could you provide input and output (including the final timing statistics) of your test run?
Which ELPA_KERNEL did you use? The CP2K default is “auto” and thus is depends on the build of the ELPA lib.

Best

Matthias

From: "cp2k at googlegroups.com" <cp2k at googlegroups.com> on behalf of Xavier Bidault <jazzquark at gmail.com>
Reply to: "cp2k at googlegroups.com" <cp2k at googlegroups.com>
Date: Thursday, 23 June 2022 at 22:27
To: "cp2k at googlegroups.com" <cp2k at googlegroups.com>
Subject: [CP2K:17225] Advice for ELPA best use


Hello,

I compiled ELPA with Intel compilers/MKL, and then I compiled CP2K 9.1 including ELPA, also with Intel compilers/MKL. So, everything was compiled with Intel compilers/MKL, and I have CP2K working with either ELPA or Scalapack Intel MKL.

I run a short example of DFTB optimization of a molecular crystal, of which I replicated the unit cell so that the system is ~30000 atoms. I used 576 MPI tasks (12 Intel nodes of 48 cores), in order to compare ELPA and Scalapack Intel MKL.

I find ELPA slighly slower (~5%) than Scalapack Intel MKL. So, not so good for ELPA whereas I was expecting better. When using ELPA with CP2K, it was with CP2K default settings for ELPA. I may have missed something, and I wonder if there is any best practice for ELPA I'm not aware of.

I'm all ears.

Thank you,
Xavier Bidault
--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com<mailto:cp2k+unsubscribe at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b02fec80-0eb8-4a8d-8b86-18c3ad0a6cdbn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/b02fec80-0eb8-4a8d-8b86-18c3ad0a6cdbn%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/36A8FD2F-5BFA-48C2-91B8-EE4B3B461ED0%40psi.ch.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20220624/15542062/attachment-0001.htm>


More information about the CP2K-user mailing list