gCP Error

37 views

Skip to first unread message

FS

unread,

May 8, 2025, 7:16:36 AMMay 8

to cp2k

Hi all,

I'm trying to run some calculations with r2scan-D4 with TZVP-MOLOPT basis sets and run into crashes when attempting to add Grimmes gCP correction. It appears as if fortran runs into issues with overlap in attribution of variables (already used) when the gCP block is worked on.

The output stops at this section:

Pair potential geometrical counterpoise (gCP) calculation

Gloabal Parameters: sigma = 1.0000
alpha = 0.9410
beta = 1.4636
eta = 1.3249

kind nvirt Emiss asto
1 14.0 0.0800 2.8404
2 26.5 0.3000 2.7679
3 29.5 0.3000 3.1207
4 5.5 0.0270 1.5899

SLURM returns many errors, including many that sound similar to:

At line 247 of file /hpc/gpfs2/sw/cp2k/2025.1-ompi5.0-gcc13.2/src/subsys/atomic_kind_types.F
Fortran runtime error: Attempting to allocate already allocated variable 'atom_of_kind'

The test input files, and slurm error output is attached.

Can someone give pointers how to fix this or use gCP properly with cp2k?

All help would be appreciated,
FS

gCP error.zip

Frederick Stein

unread,

May 8, 2025, 7:30:57 AMMay 8

to cp2k

Dear Falk,

This is indeed a bug. The feature is not covered by our regtest suite. If you compiled CP2K yourself, it might help to remove or comment out (Prepend `!`) the allocate statement in line 150 in src/qs_gcp_method.F . I am currently investigating whether that actually solves the problem.

Best,

Frederick

Frederick Stein

unread,

May 9, 2025, 4:41:05 AMMay 9

to cp2k

Dear Falk,

I fixed the technical issues in the master branch. If compiled CP2K yourself, for instance with checkout the current master branch and recompile CP2K.

HTH,

Frederick

Reply all

Reply to author

Forward

0 new messages