I am trying to run adsorption of acetone in UiO-66 using a hybrid GCMC/MD scheme using 3March20 LAMMPS software.
I am able to run it on a single smp core. But, if I try running it on 12 smp core. I get the following error:
ERROR: fix gcmc does currently not support full_energy option with molecules on more than 1 MPI process. (…/fix_gcmc.cpp:728)
Is there a way to make it run in parallel?
I want to attach my input file but I am unable.
No, there have been lengthy discussions about this topic in the past.
This is a principal problem of the algorithm and would lead to inconsistent results if ignored.
In general, monte carlo operations and parallelization with domain decomposition as used in LAMMPS rarely work together. You would have to exchange data between all subdomains at every MC attempt and that would make it prohibitively expensive. Already the “full_energy” option is very time consuming.
Thus the only parallelization available in this case is that through multi-threading with the KOKKOS (using OpenMP) or OPENMP packages (INTEL is not likely to work since it depends on cached data that would become inconsistent through MC steps).
I have added my input files here -
in.Rogge_AceHbond.lammps (7.8 KB)
log.lammps (9.3 KB)
slurm-6667610.out (2.4 KB)
Thank you so much for the quick response! I really appreciate it.