[lammps-users] How to run a simulation on different nodes (or on partitions of processors) from one input script?

Dear LAMMPS users and developers,
Recently I am working on a test case whose simplified version is the mail attachment (in.test). In this test case, I use a loop variable ‘a’ to output chunk properties to different files named ‘a.dat’. The output 10 data files are independent of each other.
In my actual simulation, because the simulation steps are so large, so I want to run the simulation on different nodes. For the test case, If I have 2 nodes, each node has 24 cores, I hope the 2 nodes can work together, and each node can complete the calculation of 5 data files. The simulation results of 2-node-simulation will be the same as 1-node-simulation, but the simulation time will be reduced by nearly half.
Also, the problem can be solved by running simulation on partitions of processors. For the test case, if I have 48 cores, I hope the cores can be partitioned into 2 groups, each has 24 cores. The 2 groups will work together, each completes the calculation of half of the data files.
I was wondering if I could do this by using only one input script. I would appreciate it very much if anyone could give me some help about this problem.
Best regards !
Qihan Ma

in.test (1.24 KB)

N2.dat (220 Bytes)

The manual explains how to run LAMMPS with multiple partitions… and you can use a uloop variable instead of a loop variable.
But in this case it would be just as well to just start two calculations. If there would be and issue with on run somehow, you will have only part of the jobs interrupted.
Axel