Hello everyone,
I was wondering what is the way lammps assigns atoms to the processors in a parallel run. For instance, say that an fcc box is defined with dimensions -6 6 -6 6 -6 6 with lattice spacing 4 and that 16 processors are available.
Thanks
Jaime Sanchez
LAMMPS does something simple. It takes the box
size and chops it into a 3d grid of procs, in such
a way that Px * Py * Pz = P, and such that
the surface area of each sub-box is minimum. I.e.
the sub-boxes are as cubic as possible. Note that
if you run on 17 procs, LAMMPS doesn't have a lot
of choices.
Steve