[CP2K-user] [CP2K:11534] Approaching performance issues within CP2K

[Krack Matthias (PSI)]

Dear Alejandro

You can use the TRACE keywords in the &GLOBAL section for a detailed (debug) output. Alternatively, you can post the final timing report of your run(s) to this forum and hope that someone has a clue.

Best

Matthias

From: cp... at googlegroups.com <cp... at googlegroups.com> On Behalf Of ar26... at gmail.com
Sent: Donnerstag, 4. April 2019 12:20
To: cp2k <cp... at googlegroups.com>
Subject: Re: [CP2K:11534] Approaching performance issues within CP2K

Dear Matthias,

thank you very much for your reply!! And I'm sorry for the delayed response. I tried to do that, but it is not so obvious to me how can I make the comparison. Also this info is globally for the complete job and I have time steps that run fast and slow. Is there a way to output this information every step?? That would be ideal!!

Best regards,

Alejandro.

El lunes, 1 de abril de 2019, 10:50:45 (UTC+2), Matthias Krack escribió:
Hi Alejandro

You can compare the timing reports at the end of the CP2K output file for the same input before and after the maintenance. Maybe that gives a hint which routine(s) have slowed down. On the other hand, your admin should know what has been changed or updated during the maintenance which might give further hints.

Matthias

From: c... at googlegroups.com<javascript:> <c... at googlegroups.com<javascript:>> On Behalf Of ar... at gmail.com<javascript:>
Sent: Montag, 1. April 2019 10:31
To: cp2k <c... at googlegroups.com<javascript:>>
Subject: [CP2K:11505] Approaching performance issues within CP2K

Dear all,

I'm having some performance issues with CP2K after a maintenance performed in the cluster I'm using. I'm performing DFTB calculations and I'm using one single node. Before the maintenance I obtained a time per frame profile like before_maintenance.png, with an average time per frame of 0.066 s; and after the maintenance, a profile like after_maintenance.png, with an average of 0.5 s (10 times slower!) and pretty unstable.  Other weird things happens like weird_job.png, where you have normal slow frames, super slow frames and a window where the calculations went fast as they should. Everything for the same job, node, input file and slurm script. Actually, this window was replicated in all my other running jobs at the same time and for the same time length, so something happened globally in the cluster that allowed calculations to run fast for a period of time.

I'm working with the administrator of the cluster to debug the problem, but my questions is: How can I approach this problem as a CP2K user? I want to discriminate if is a problem of mpi, hard drive or memory access, or other thing, so I can give the administrator more info of the problem.

Thanks in advance,

Alejandro.

--
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 c... at googlegroups.com<javascript:>.
To post to this group, send email to c... at googlegroups.com<javascript:>.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.
--
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 cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
To post to this group, send email to cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190404/8c6b847f/attachment.htm>