[lammps-users] Extending Lammps

Dear Steve and users,

I am planning to port some of my subroutines (in Fortran) for computing the normal mode decomposition to Lammps and I would like your advise. These are really expensive computations that require the following:

  • Read large input data sets involving complex numbers at the beginning of the run
  • Initial atom positions should be stored
  • Perform calculations involving complex numbers
  • Parallel exchange of information on relatively small arrays
  • Print large data sets (> 1-4GBs)
  • Calculations should be performed every 8-16 time steps and only one call is required per step

I was thinking that both fix or input script commands could be suitable to do these calculations, any suggestion?

Regards, Javier

A fix with an end_of_step() method (to do the heavy computation, output)
is the LAMMPS way to do this. You shouldn't need any additional
input script commands if you can set all relevant arguments
as part of the fix command. I'd have to understand what you need
to communicate in parallel to know if anything else is required.