I would like to ask what might be a possible cause for linearly
increasing virtual memory amount upon box volume change in a reax/c
simulation of a periodic system (I attach the input and output files),
which causes it to crash. I do observe that the memory remains constant
between the volume changes. The lammps version is 22 Aug 2012.
control.lmp (1.43 KB)
data.restart (318 KB)
ffield (21.2 KB)
in.combined (1.08 KB)
stretch_Al2O3_hole_h2o_xf.sh.o1200053 (2.81 MB)
Ray can probably look at this.
The problem only presents itself when qeq/reax is used with the volume
change; there seems to be a memory related issue or leak in qeq/reax.
Allow me some more time to fix it, thanks.
I have checked fix_qeq_reax throughly with both naked eye and
Valgrind, a memory leak checker, and it came out clean.
Try with the following command, with which the virtual memory does not
increase between runs and all allocated arrays are properly
deallocated/freed at the end.
this seems to have fixed the memory problem.