[lammps-users] Problem with multi-processor runs using long rang PPPM/tip4p

Dear all

I am running LAMMPs in parallel using the PPPM/tip4p with AIREBO and LJ/longrange/TIP4P potential. Despite specifying the mpirun -n8 setting which specifies 8 processors, the log file still indicates:

   orthogonal box = (-97.4982 -97.4982 -61.4878) to (97.4982 97.4982 62.9078)
   1 by 1 by 1 processor grid

I do not have this issue when I compute using other potentials.

Also, I have collegues running LAMMPS on other machines and they have had issues computing using the PPPM/tip4p solver in parallel. Sometimes it works, at other times it doesn;t depending on the no. of processors you select.

Anyone have the same issues?


Huck Beng

Re: 1 vs 8 procs. I think this would be an MPI issue on your
box with launching your job. I can't imagine how it would depend
on the potential you were using in LAMMPS. These lines

orthogonal box = (-97.4982 -97.4982 -61.4878) to (97.4982 97.4982 62.9078)
1 by 1 by 1 processor grid

are often printed before a potential is ever selected.

Re: other people's problem with PPPM/TIP4P. You'll have to
be more specific about "doesn't work". Does it give an error
message. Any PPPM variant can complain if you use too many
procs so that the grid stencil overlaps a neighboring proc's domain
too far.