[CP2K-user] [CP2K:16199] Re: memory leak

[Matt D]

Hi Daria,

I've recently encountered this problem elsewhere. The fix in this case was 
to recompile with OpenMPI using a GNU compiler rather than IntelMPI and an 
intel compiler. 

I'm not 100% sure about the technical details as this was done by the 
support services but it seemed to work. They also used the leak sanitizer 
tool to test things and this can be found in the cp2k documentation. 

Finally I'd check whether it is a leak or simply a high memory requirement 
job. If it's the former, using the above fix worked for us. However with 
the latter, the memory consumption with job time should plateau rather than 
be constantly increasing. You can check to see by first using only a few 
processors per node to increase the maximum memory capacity and allowing 
jobs to run for at least a few minutes. If you then set a short WALLTIME in 
cp2k of just a few minutes, it should then terminate properly and tell you 
the maximum memory consumption at the end. If you systematically increase 
the WALLTIME you can then evaluate how the max memory changes with 
WALLTIME. If it is constantly increasing then it is a leak, if not using 
less cores per node is the best solution.

Best wishes,
Matt

On Friday, 5 November 2021 at 13:51:35 UTC Daria wrote:

> Dear all,
>
> I’m trying to run cp2k on the new Dutch supercomputer Snellius. However I 
> experience huge memory leaks.
>
> The code has been compiled with Easybuild
>
>
> https://github.com/easybuilders/easybuild-easyconfigs/blob/develop/easybuild/easyconfigs/c/CP2K/CP2K-8.2-foss-2021a.eb
>
> Does anyone have suggestions on what I could try to fix the problem?
>
> Thank you
>
> Daria
>

-- 
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/42b9ea3a-e6e8-474a-a132-c4ba2ffc18f8n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20211108/ce9c3952/attachment.htm>