[lammps-users] LAMMPS : heat flux with mutiproc

I'm not sure what you're asking. I presume LAMMPS itself
is doing the right thing on 1 or 4 or 8 procs, i.e. the output
thermodynamics is correct. Note that 2 runs can
be different due to round-off or different inititialization in
parallel, yet still both be statistically correct.

Then you are measuring heat flux by some means (you don't say
how), but I presume it is just a diagnostic, and getting different
answers. If it is your method for doing this, then you will need
to debug your 2 answers and figure out what is making them


Hi Yann,
Did you try regular way to calculate heat flux, compute heat_flux, for
something simle
like Ar with LJ potential. I did it for 1, 2 and 8 processors with
two methods,
step 0.002 ps. The results was pretty close. It can help if it's
possible to understand
if the difference is the same for two types of calculations.