I have no problem running that script on my box in serial
or parallel (with a Ni EAM potential in potentials) The size of the
dump:rbuf allocation is small
and correct. Are you using the current version of LAMMPS?
I tested the input script below with two most recent versions of lammps. It runs fine with May 19 2011 version with no error and dumping cfg files as expected. But June 03 2011 version (which is what I was running previously) is giving me the error with no dumps.
Please try the most current version of LAMMPS.
Since I don't get any issue I can't debug it.
If you still get the same error (huge memory),
I suggest you put a print statement in the code
jn dump_cfg.cpp where the error string is printed
(dump:rbuf), and see what sizes it thinks
the 2 dimensional array is. From there you
could backtrack to how that huge size got set.
If you can try the most current version (tarball that
you download today), that would be better.
Since I can't reproduce the problem, you will probably
need to instrument the code as I indicated and print
out some values to see what is going wrong.