Bug of Voronoi package for large number of atoms?

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS users,

I am using Voronoi package since quite some time already and it works fine, except in the following case.

I am working on bcc Fe. I first equilibrate the crystal and then I invoke the voronoi package so as to get the reference positions.

compute voronoi all voronoi/atom occupation
compute r0 all reduce sum c_voronoi[1]
compute r1 all reduce sum c_voronoi[2]
thermo_style custom c_r0 c_r1
run 0

This always works. However, I found it crashes for a large number of atoms. It was a relatively large box with 863 million atoms.
There was no error in the log. It just stopped. I tried several times, even taking more nodes, thinking it was a problem of memory. But nothing. It still crashes.

Any idea why this occurs?

Best regards,
Christophe

2

Where is it crashing? Inside the Voro++ library,

in compute_voronoi_atom.cpp, or elsewhere?

How many atoms/MPI task are you running with?

You can add some print statements to figure out

where it is crashing. It could be that

enough memory cannot be allocated, either

by LAMMPS or Voro++. If the latter doesn’t

check (LAMMPS generally does), then it

could crash.

Steve