[lammps-users] Increasing NATTOTDEF, and NATDEF

LAMMPS LAMMPS Mailing List Mirror
1

Hi everyone,

I would like to increase NATTOTDEF, and NATDEF to use reax with
one hundred thousand atoms. What will I need? I am using intel fortran, and
c++ compiler. What flags are necessary to add to the compilation?
Is necessary to re-compile mpich, and if is there any flag that I have to add
to mpich compilation?

- My computer has 32Gb de RAM with 8+8 cores, I am using ubuntu 10.10 x86_64,
and the compilers are for x86_64 too.

Thanks,

2

Hi everyone,

I would like to increase NATTOTDEF, and NATDEF to use reax with
one hundred thousand atoms. What will I need? I am using intel fortran, and

those constants are defined in lib/reax/reax_defs.h

c++ compiler. What flags are necessary to add to the compilation?

you may need to increase other constants in the same file
to adjust for the increased system size.

Is necessary to re-compile mpich, and if is there any flag that I have to add

no. why should this be needed?
MPI knows nothing about application specific details.

to mpich compilation?

- My computer has 32Gb de RAM with 8+8 cores, I am using ubuntu 10.10 x86_64,
and the compilers are for x86_64 too.

have you made some tests/benchmarks as to how fast that machine will be
on a system that barely fits into the existing array dimensions?
reax is not exactly a "cheap" model and you want to model are quite
large system.
for that reason, you may want to try out the newer reax/c user package, that is
supposed to be faster.

axel.

3

for that reason, you may want to try out the newer reax/c user package, that is
supposed to be faster.

Moreover, you don't need to twiddle these settings for the C version.

Steve