[lammps-users] Mismatch between results

Dear all,

I ran granular particles’ simulation using LAMMPS with lmp_serial and lmp_linux executables. I used
two processors for lmp_linux. The simulation parameters and initial conditions are same for both runs.
I am looking for average kinetic energy. But the value calculated by both runs (serial and parallel) is different.

KE value by lmp_serial = 81 LJ units
KE value by lmp_linux = 86 LJ units

How is it possible for KE value to be different if my initial conditions and parameters are same?

Please elaborate on this problem.

Thanks.

With best wishes,

Ashish

Dear all,

[...]

KE value by lmp_serial = 81 LJ units
KE value by lmp_linux = 86 LJ units

How is it possible for KE value to be different if my initial
conditions and parameters are same?

perhaps your conditions and parameters are not well chosen?
perhaps you have miscompiled executables?

there are a lot of things that can go wrong when running
MD simulations and problems/bugs first need to be confirmed.

unfortunately, you provide *no* information that can be
used to quickly attempt to reproduce your observations.
thus until you provide evidence to the contrary, one has
to assume that it is your fault. :wink:

cheers,
   axel.

Dear Axel,

Thanks for your reply.

I saw the difference in the KE values in the Pour example given in the examples folder (came with LAMMPS package).
There is also a difference between KE when same input script file is used with one processor and four processor simulation run.
Please have a look at it.

best wishes,
Ashish

If you run on different numbers of processors you can get slightly
different answers
with diverge over time for long runs (lots of timesteps). See
doc/Section_errors.html,
sub-section 9.1.

Steve