[lammps-users] long range interaction:ewald or PPPM

Hi there,

I would like to know if anyone have succeeded in using the latest version of LAMMPS to run simulations with long range coul interaction via either ewald or pppm method. In my case, I cannot even run the example of peptide, a few steps later the coul energy would be nan! I cannot even run the bench of chute: it complains that the atom type is wrong.
I complied LAMMPS on Fedora with mpiCC, fftw 2.1.14, openmpi.

Thanks.

Konzern

Hi there,

hi!

I would like to know if anyone have succeeded in using the latest
version of LAMMPS to run simulations with long range coul

yes.

interaction via either ewald or pppm method. In my case, I cannot
even run the example of peptide, a few steps later the coul energy
would be nan! I cannot even run the bench of chute: it complains
that the atom type is wrong.

I complied LAMMPS on Fedora with mpiCC, fftw 2.1.14, openmpi.

i guess you are running a recent fedora release with
a gcc v4.x compiler.

current GCC compilers assume that code adheres to strict
ANSI-aliasing rules. some of files in LAMMPS do not do
this for a number of reasons. you can turn this off by
adding the -fno-strict-aliasing flag, which will make
LAMMPS run a little bit slower where it adheres to ANSI
aliasing rules or use a custom makefile with overrides
for only the files which are not compatible like the
attached one.

cheers,
   axel.

Makefile.openmpi (1.75 KB)

The issue with tabling (on by default) in the coul long pair routines is
discussed in the LAMMPS manual. See item (3) under Additional
Build Tips in section 2.2 of the manual. This leads to something
like's Axel's makefile.

The chute issue is that the granular package is not turned on by
default in LAMMPS. Try

mka yes-granular; make machine

Steve