Hi all,
I want to simulate the explosive behaviour of some solid materials, so I need to leave much space around the system or the two end of the system in one direction. Thus here is the question: Can LAMMPS do adaptive or dynamic spatial domain decomposition to run the CPUs more efficiently? Because in normal spatial decomposition only very little CPUs will be actually used( no atom in the space which the CPU calculates.)
Thanks for any answer or suggestion!
Xiaopeng
If you have free surfaces then using boundary "s" for
shrink wrap will adjust the box size as the simulation
progresses. Once LAMMPS decides (at the beginning)
to partition the box onto Px by Py by Pz procs, that
will not change. If your box shape changes so much
that that is a bad decomposition, then you should write
a restart file, restart with it, and continue. In that
case LAMMPS will compute Px by Py by Pz again.
Steve
2009/5/12 Xiaopeng Huang <[email protected]...>: