Thanks for your reply
If there is no periodic boundary condition LAMMPS and me will give the same results. In my way, I treat a specific similation box as some real thing, which is unreasonable. In PBC, if average over all nonequivalent simulation box, my way will give the same value as LAMMPS does. The way LAMMPS uses is independent of a specific similation box, if there is only pair potential, e.g. my system.
However, in my view, for similation with PBC pressure is illed defined, especially for polymer melts. We should care about stess tensor other than pressure.
Thanks!