LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] Bad termination of processes
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] Bad termination of processes

From: Axel Kohlmeyer <akohlmey@...24...>
Date: Mon, 14 May 2018 07:19:14 -0400

On Mon, May 14, 2018 at 4:13 AM, Michael Papanikolaou <mich.papanikolaou@...24...> wrote:
Dear LAMMPS users,

I am running a LAMMPS script and I am getting the following message at the timestep no 17,000.

=   PID 14343 RUNNING AT dcn035
=   EXIT CODE: 6
   Intel(R) MPI Library troubleshooting guide:

This error  must be associated with the following 2 commands (when I do not compute the rdf and stop averaging, the problem stops occurring):

compute rdf1 all_atoms1 rdf 10 1 2 cutoff 4.5
fix avetime2 all_atoms1 ave/time 1000 1 1000 c_rdf1[*]  file rdf.profile mode vector

As this is about a grinding simulation, the error occurs specifically at the time when the abrasive approaches very closely to the workpiece (the number of 1-2 pairs is increasing sharply).

​unfortunately, none of the information from the job itself copied here or in the attachment provides any "smoking gun", that allows a clear identification ​of a possible error.
if your suspicion of the increase in the number of neighbors is correct, you can try increasing the maximum expected number of neighbors per atom using "neigh_modify one" (and for efficient memory use, also adapt the neigh_modify page options accordingly). however, this would only apply if your cutoff of 4.5 is much larger than the regular cutoff used. you may also try to reduce this cutoff, instead.

another test would be to run with a newer LAMMPS version in order to verify, that you are not running into a bug, that has already been fixed.

the final option would be to try and build and post a small test input, that reproduces the issue and can be run quickly in serial by LAMMPS developers, so they can apply their debugging tools to this issue (if indeed it is an issue in the LAMMPS code).


You can also find attached an error report. I am using LAMMPS/31Mar17.

Any advice on how to overcome this issue would more than useful.

Thank you in advance.

Kind Regards,

Check out the vibrant tech community on one of the world's most
engaging tech sites,!
lammps-users mailing list

Dr. Axel Kohlmeyer  akohlmey@...12...24...
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.