LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] Running LAMMPS from PYTHON via python interface to Lammps
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] Running LAMMPS from PYTHON via python interface to Lammps

From: Axel Kohlmeyer <akohlmey@...24...>
Date: Wed, 18 Oct 2017 14:09:14 -0400

On Wed, Oct 18, 2017 at 1:43 PM, Dundar Yilmaz <duy42@...122...> wrote:

​first off. in the future, please consider this piece of mailing list etiquette:
if you want to write a post on a new topic. write a completely new e-mail. do *not* reply to a previous​ e-mail on a different topic and particularly do not reply to a digest e-mail leaving all the (pointless) other e-mails in your quoted body. remember, that you are e-mailing to hundreds of people that do not care about this quoted content (and neither do you). thanks for your consideration.

In a project I am trying to run lammps from python and between each simulation step I would like to do some analysis.
In a loop:
for i in range(10000):
lmp.command(“ run 1”)
analysis_function ( lmp)

Assuming that analysis function wont take much time is it a good approach? Is there any overhead running lammps step by step?

​it is a decent approach. another approach would be using fix python to have a python function/variable called regularly during ​a run.
yes, there is overhead, but if you only do analysis and not any modification, you can use lmp.command("run 1 pre no post no") to significantly reduce the overhead. with "pre yes", you essentially repeat step 0, even though the data from the previous step is still valid.
I tested this for a very small system:

lmp.command(“ run 10000)

for i in range(10000):
lmp.command(“run 1”)

Time spent for the latter case was three times of the first one.

To be more specific I would like to compare bonding information at each step to detect chemical reactions. If there is a way to access fix reax/bond data through lammps during the simulation
I could save it to a different variable and analyze it after the md simulation.

​that sounds like it would be simpler to do the whole analysis in post-processing.​



Thanks for your time.

Dundar Yilmaz, Ph. D.
Post Doctoral Researcher 
Department of Mechanical and Nuclear Engineering 
138 Research East Building 
Penn State University 
University Park, PA 16802

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