[lammps-users] GPU LAMMPS Pair energy getting stuck at zero

LAMMPS LAMMPS Mailing List Mirror
1

Energy decreases - and oscillates about a minimum (negative value) - as it should - the simulation runs fine with 15-6000 atoms on various platforms on CPUs.

Thanks!
Manish Agarwal
<zmanish@…24…>
Postdoctoral Research Scientist
Chemical Engineering
Columbia University

2

Energy decreases - and oscillates about a minimum (negative value) - as it
should - the simulation runs fine with 15-6000 atoms on various platforms on
CPUs.

well, then you'll have to post some complete example input
so that somebody can try and reproduce this behavior.
with the extremely limited amount of information that you are
providing, it is impossible to identify any source of error.

axel.

3

Yes, indeed!
I was really hoping it was a common oversight on my part - which could be resolved by experience of the community.

just a thought : is the “pair_modify mix arithmetic” a possible culprit?

The file config(initital configuration with atoms, angles, etc) and control(controlling script) files are attached.

Thanks and Best wishes!
Manish Agarwal
<zmanish@…24…>
Postdoctoral Research Scientist
Chemical Engineering
Columbia University

config (3.7 KB)

control (854 Bytes)

4

Yes, indeed!
I was really hoping it was a common oversight on my part - which could be
resolved by experience of the community.
just a thought : is the "pair_modify mix arithmetic" a possible culprit?

no. the problem is that you are using "run_style respa"
that is not supported with gpu acceleration. you currently
_have_ to use "run_style verlet" for and /gpu pair style.

axel.

5

Excellent ! so it was indeed an oversight - I did not read the manual completely!!

Thanks a ton, Axel, all

Best Wishes,
Manish Agarwal
<zmanish@…24…>
Postdoctoral Research Scientist
Chemical Engineering
Columbia University