Lammps partition feature


Could someone simplify the explanation for the partition feature in lammps used along with NEB package?

Thank you

Why don’t you first explain what you have understood rather than forcing people to guess or waste time trying to explain things that are already clear to you or both?


I am beginner to lammps and haven’t understood how would partition feature help unlike using lammps under normal scenarios when requested for x number of CPU cores( say using 8 cpu cores out of 16 cpu cores) using SLURM scheduler

I went through its documentation and it said “This ( partition ) command invokes the specified command on a subset of the partitions of processors you have defined via the -partition command-line switch.”
What does it imply by “subset of partitions of processors”?

Thank You

When you do an NEB calculation (or other commands from the REPLICA) package, your are not doing one calculation, but multiple coordinated calculations at the same time. For NEB you have to do one calculation for each “point” of the band. Those are essentially the same calculation of the entire system but with just some small changes in the geometry.

Thus when you normally would do a parallel calculation with 8 processors and you want to do a NEB calculation with 6 points in the band, you need to request 48 processors and run 6 partitions of 8 processors, or request 24 processors and run 6 partitions of 4 processors or request 12 processors and run 6 partitions of 2 processors or request 6 processors and run 6 partitions of 1 processor.
So the number of “points” determines the number of partitions and then you can run one or more processors per partition.

Thank you Axel for explaining it to me.