这几天在两台机器上跑lammps，偶然发现分别跑了两个一样的算例，但结果却有不一样，后来查了下手册，发现12.1 Common problems中写了如下一段话：

If two LAMMPS runs do not produce the exact same answer on different machines or different numbers of processors, this is typically not a bug.  In theory you should get identical answers on any number of processors and on any machine.  In practice, numerical round-off can cause slight differences and eventual divergence of molecular dynamics phase space trajectories within a few 100s or few 1000s of timesteps. However, the statistical properties of the two runs (e.g. average energy or temperature) should still be the same.

个人理解，瞬时量没有太大意义，做某个时间段的平均更有统计意义。

下面摘自lammps mailing test关于这种问题一个答复：

“you are comparing apples with oranges, i.e. numbers that have been averaged over 99 steps each 5 steps apart, with numbers from a single step.​ why *should* they be the same?”