120680 atom system - CPASSERT failed

[Pierre-André Cazade]

Dear CP2K users,

I am trying to run an energy calculation for a 120680 atom biological system. I run this on a node with almost 2TB of RAM. I tried with and without LS_SCF and I always get a CPASSERT failure despite all the tweeting of the threshold I may try. Could anyone point what is wrong in my system and what I need to address to make it work. Please find my input file at the following link:

https://drive.google.com/open?id=1ibpjZuupTjNBCUAoSDQKEbh-v-YxqSCR

Regards,

Pierre

[Tiziano Müller]

Hi Pierre,

I haven't looked at your input (yet), but it is always helpful to also
post the exact error message: CPASSERT usually shows at which source
code line the error occurred, which helps to pinpoint the problem.

Best regards,
Tiziano

> --
> 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+uns...@googlegroups.com
> <mailto:cp2k+uns...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/cp2k/eae745b6-ac09-437f-b96c-391a35152340%40googlegroups.com
> <https://groups.google.com/d/msgid/cp2k/eae745b6-ac09-437f-b96c-391a35152340%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Tiziano Müller
University of Zurich
Department of Chemistry
Winterthurerstrasse 190
CH-8057 Zürich

Tel: +41 44 63 54234
www.chem.uzh.ch
tiziano...@chem.uzh.ch

[Pierre Cazade]

Hi Tiziano,

My bad. Please find below the most recent error message I got:

 *******************************************************************************
 * ___ *
 *  / \ *
 * [ABORT] *
 *  \___/                             CPASSERT
failed                          *
 * | *
 * O/| *
 * /| | *
 * / \ pw/pw_grids.F:853 *
 *******************************************************************************


 ===== Routine Calling Stack =====

            8 pw_grid_distribute
            7 pw_grid_setup_internal
            6 pw_grid_setup
            5 pw_env_rebuild
            4 qs_env_rebuild_pw_env
            3 qs_env_setup
            2 qs_init_subsys
            1 CP2K

I tried to change the values of the CUTOFF and REL_CUTOFF, as well as
the values of  EPS_DEFAULT and EPS_PGF_ORB. I tried the SZV and DZVP
MOLOPT-SR basis sets. I also changed NGRIDS from 4 to 6. Nothing seems
to get me through. I am looking forward to you insight.

Regards,
Pierre

Dr Pierre Cazade, PhD
AD3-023, Bernal Institute,
University of Limerick,
Plassey Park Road,
Castletroy, co. Limerick,
Ireland

[Matt W]

Your large cutoff and huge cell have probably never been tested fully. Try running with just a single atom in the cell and see what happens, maybe scale the cell dimensions to see where it breaks. 

Matt

>> send an email to cp...@googlegroups.com
>> <mailto:cp...@googlegroups.com>.

[Pierre Cazade]

Dear Matt,

I can hardly scale the dimension of the cell as it is a crystal with 8 protein units (space group 96). The crystal has been tested and relaxed with a classical model to remove bad contacts with water molecules. As for the large cutoff, I don't know what you mean by that. Even for smaller systems (280 atoms), applying the methodology described in the howtos, I have a hard time identifying a satisfactory combination of cutoff and rel_cutoff providing good accuracy and convergence around these values (I mean by convergence that, by slightly changing one or the other cutoff, the accuracy remains the same). I usually find that very high values for the cutoff are necessary. In fact, in the present context I feel it is actually quite small but I need the system to run first before I can optimize them anyway.

Regards,
Pierre

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/93fb0100-2a7d-4a79-be6b-4dbef3682c83%40googlegroups.com.

[hut...@chem.uzh.ch]

Hi

what Matt was pointing to is
1) there might be not enough memory
2) the plane wave part of the code might be buggy for this large
size of code

As we have done calculations of similar size cells, my guess would be 1).
If I run your cell with one single water molecule and a cutoff of 100,
the job fits on my workstation with 256 GB memory. For a cutoff of
800 you will need approx. 8^1.5=22.6 times more memory (= 6 TB).
However, these are very rough estimates and my jobs stop with another
error message if there is not enough memory.

In any case, the CPASSERT points to a problem with the plane waves.

regards

Juerg

--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: cp...@googlegroups.com
From: "Pierre Cazade"
Sent by: cp...@googlegroups.com
Date: 01/21/2020 11:39AM
Subject: Re: [CP2K:12796] 120680 atom system - CPASSERT failed

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/138a652b-eddc-64b6-fcf4-671c6430a16b%40gmail.com.