[lammps-users] limit on the number of atoms


I was wondering if there's a limit on the maximum number of atoms that lammps can simulate.

While my lammps executable (version 21Jul06) was running perfect for small systems (up to 15000 atoms), for large systems (50 000 atoms or more) it has started to give an error that looks like (on a Beowulf cluster):

p1_22628: p4_error: net_recv read: probable EOF on socket: 1
      p2_22629: p4_error: net_recv read: probable EOF on socket: 1
      bm_list_22627: (215.570312) net_send: could not write to fd=5, errno =
          p4_error: latest msd from perror: Bad file descriptor
bm_list_22627: p4_error: net_send write: -1

Any help is truly appreciated,
thank you,


The only limit on # of atoms is what will fit in memory. 50000
isn't that many so I doubt you are running out of memory.
LAMMPS prints the memory usage (per processor) when it
starts a run, so you can check.

The error looks more like a message-passing or file error.
So I would check that MPI is configured correctly for your Beowulf.