Skip to content

Running OWL

Jump to bottom
Ying Wai Li edited this page Jul 27, 2020 · 6 revisions

Running a new simulation

Serial run

  1. You need an input file to run OWL. Please refer to the examples in the examples/ directory to set one up.

  2. To start a serial simulation, execute

mpirun -np 1 ./owl owl.input

This launches one MPI rank to execute OWL.

Parallel run

To run a parallel simulation using 5 MPI ranks, for example, execute

mpirun -np 5 ./owl owl.input

The mpirun command might be different depending on the machine you are running on. Equivalents could be, e.g., mpiexec, srun for Slurm workload manager, or aprun on Cray machines.

Restarting a simulation

  1. Change the restartFlag field in the input file owl.input to 1.

  2. Execute OWL using the same command as above:

mpirun -np 1 ./owl owl.input
  1. OWL will look for the following checkpoint files to pick up from the previous run:

  • [algorithm]_checkpoint.dat: This file contains information about where OWL stops in the previous run. File name is different for different algorithms, e.g.: metropolis_checkpoint.dat for Metropolis sampling; hist_dos_checkpoint.dat for Wang-Landau sampling. If this file is missing, OWL will start a new simulation. Do NOT change the name of this checkpoint file.

  • configurations/config_checkpoint.dat: The config_checkpoint.dat file inside the configurations directory contains the configuration of the last MC step in the previous run. If this file and the algorithm checkpoint file are both present, If this file is missing, OWL will start from a random initial configuration, and the thermalization procedure will be repeated before entering the sampling stage (even the previous simulation was already in the sampling stage).

Clone this wiki locally