[CP2K:5505] large memory with nonlocal vdW

hut... at chem.uzh.ch hut... at chem.uzh.ch
Wed Jul 9 09:39:21 UTC 2014


for small systems the memory is dominated by the grid.
The size of the grid depends on the size of the computational box
and the cutoff. 
For the vdW functionals we need a relative large number of additional
grids. This explains the strong increase in memory for your system.


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

-----cp... at googlegroups.com wrote: -----
To: cp... at googlegroups.com
From: Gary Kedziora 
Sent by: cp... at googlegroups.com
Date: 07/09/2014 12:30AM
Subject: [CP2K:5505] large memory with nonlocal vdW


I'm trying a water dimer calculation with the XC_GGA_X_OPTB88_VDW functional and with the DZVP-MOLOPT-GTH basis set (input file attached).  On one Sandy Bridge processor node using 16 MPI processes, a system memory monitor is reporting mem=17572912kb, which is rather large for this small calculation.  Changing to the built-in CP2K implementation of revPBE, the maximum memory usage is mem=6117036kb, about 3 times smaller.  Is this a known problem?  Or perhaps it is a system-dependent problem?

When I try TRACE in the GLOBAL section of the input, I'm seeing near the end of the output

 000000:000016<<    1      1 CP2K     140.991 Hostmem: 1119 MB GPUmem: 0 MB

for the OptB88 case, and

 000000:000016<<    1      1 CP2K      29.834 Hostmem: 411 MB GPUmem: 0 MB

for the revPBE case. I'm not sure how to interpret that Hostmem figure.  Will someone provide guidance?  Is this helpful for debugging this situation?  I guess that it is the maximum amount of memory allocated by CP2K.

 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... at googlegroups.com.
 To post to this group, send email to cp... at googlegroups.com.
 Visit this group at http://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.

[attachment "water_dimer_1.inp" removed by Jürg Hutter/at/UZH]

More information about the CP2K-user mailing list