getting per-atom pe by final integrate

FYI, I've got this resolved. It was a combination of resetting the compute's invoked flag and getting our push update system to play nicely with Lammps' update system.
Jeremy