Dear lammps-users,
I am using Lammps version (15 may 2015). I added a new class by modifying fix/qeq/reax. It works properly when I run my simulation with a single core, but in parallel mode it does not. I did not change any parallel function in the source file. I would be thankful if you could tell me where the problem comes from.
Here is my input script:
Dear lammps-users,
I am using Lammps version (15 may 2015). I added a new class by modifying
fix/qeq/reax. It works properly when I run my simulation with a single core,
but in parallel mode it does not. I did not change any parallel function in
the source file. I would be thankful if you could tell me where the problem
comes from.
most likely your modifications. nobody here will debug your code for
you, and it is particularly difficult to make any statement without
any kind of description on what your modifications are and how they
are not "proper" in parallel. on top of that, your input looks flawed,
too, since you seem to be double counting interactions.
i would suggest to test your modifications on a (much) simpler input.
axel.
If you have really not touched any code for parallel operations, then
the discrepancy probably comes from insufficient testing on your part.
If you run more than one simulation, or pay closer attention, I expect
you will see problems in both serial and parallel runs. LAMMPS will
generate identical forces on any number of processors, modulo very
small differences due to non-associative arithmetic and larger
differences due to random number sequences that depend on the
processor count. The latter effect can be eliminated by adding
optional keywords to your velocity command (see doc page). The former
effect only matters after you have run for many timesteps, so you
should observe effective processor independence over short runs.