Dear Lammps list,
I tried to run the examples provided to compute the thermal conductivity KAPPA with the Lammps distribution of 28th june 2014.
With the input script “in.heatflux” on 8 cores, I obtained a result identical to the reference provided with LAMMPS. However, if I run this input script on different on number of processors, I obtained rather different results :
1 core => Kappa= 2.85
6 cores => Kappa = 3.97
8 cores => Kappa = 3.78 (output identical to the reference provided with the Lammps distribution).
16 cores => Kappa = 4.16
I don’t understand why Kappa differ when I increased the number of processors/cores. Is there any bug or misunderstanding on my side ?
Thanks for your help,
Tom
Dear Lammps list,
I tried to run the examples provided to compute the thermal conductivity
KAPPA with the Lammps distribution of 28th june 2014.
With the input script "in.heatflux" on 8 cores, I obtained a result
identical to the reference provided with LAMMPS. However, if I run this
input script on different on number of processors, I obtained rather
different results :
1 core => Kappa= 2.85
6 cores => Kappa = 3.97
8 cores => Kappa = 3.78 (output identical to the reference provided with
the Lammps distribution).
16 cores => Kappa = 4.16
I don't understand why Kappa differ when I increased the number of
processors/cores. Is there any bug or misunderstanding on my side ?
please note the README file where it says:
These scripts are provided for illustration purposes. No guarantee is
made that the systems are fully equilibrated or that the runs are long
enough to generate good statistics and highly accurate results.
axel.
And note that any 2 MD trajectories will diverge
over time due to round-off numerical issues
which include effects due to running on different
#s of procs. The doc/Section_errors.html doc
page explains this.
Steve