extended LAMMPS commands support

[cz4rs]

Make it possible to define multiple regions (and the size of simulation box will be adjusted accordingly). This is currently rectricted to fcc lattice style only.

Notable changes:

• use create_atoms to fill the region with atoms of specific type
• use mass to assign mass to atom type
• use velocity to assign temperature to atom type

See in.lb deck and the screenshots below for practical demonstration.

You can use paraview --script utils/paraview_process_data.py to speed up analysis of produced data. It will try to load all dump*, domain_act* and domain_lb* data files from your current directory and do some basic ParaView visualization.

---

TODO (mostly optional):

• run the simulations for existing decks in CI to ensure no parsing or runtime errors
• deduce the padding width for filenames from number of steps instead of using hardcoded default
• improve compatibility with LAMMPS (to be able to run the decks with LAMMPS for comparison)
• improve group command support
• improve communication statistics

---

Multiple regions with different atom types. Note that when the regions overlap, random type will be assigned:

Multiple regions with different velocities:

Types after 100 steps:

Velocities after 100 steps:

Types after 1000 steps:

[cz4rs]

@streeve This is remaining part of our work on CabanaMD. Marking this as a draft, as there are some limitations (mostly being restricted to fcc lattice style only) and in progress bits.
Would you be interested in getting this merged as well?

[streeve]

This is remaining part of our work on CabanaMD. Marking this as a draft, as there are some limitations (mostly being restricted to fcc lattice style only) and in progress bits. Would you be interested in getting this merged as well?

Yes definitely, I will take a look

[streeve]

Same here - always better to have more output, but I doubt we always want to output this

[cz4rs]

Yes, definitely. I will do some testing and polishing to make sure that the information is actually useful.

[cz4rs]

Relevant cabanaMD.out fragment after integrating this with the other statistics:

#Timestep Temperature PotE ETot Time Atomsteps/s LBImbalance Exchanged
0   13.038466   -4.081539   15.472248   0.00    0.00e+00    -nan    -   
10  11.276282   -1.442460   15.468581   0.52    9.67e+04    -nan    -   
20  10.717567   -0.607034   15.466101   1.13    8.11e+04    7.55e-01    1057
30  10.846505   -0.798659   15.467845   1.87    6.77e+04    7.55e-01    -   
40  10.953565   -0.960041   15.467020   2.73    5.80e+04    7.14e-01    1011
50  11.091998   -1.167610   15.467059   3.15    1.19e+05    7.14e-01    -   
60  11.050713   -1.105665   15.467089   3.65    1.02e+05    6.50e-01    704 
70  11.114292   -1.200594   15.467509   4.00    1.41e+05    6.50e-01    -   
80  11.153192   -1.258347   15.468094   4.40    1.26e+05    5.65e-01    639 
90  11.169153   -1.281515   15.468864   4.66    1.89e+05    5.65e-01    -   
100 11.207249   -1.339408   15.468103   4.96    1.66e+05    4.70e-01    555 

Exchanged column will show up only when Load Balancing is enabled.

[cz4rs]

I went with total values since we don't seem to need a breakdown per rank.

[cz4rs]

@cwschilly This should help with modifying the load balancing test case.

@streeve Thanks for the review. Leaving this as a draft for now, I will fix the output and possibly some other minor issues next week 🤞

[streeve]

@cz4rs just a note that I updated the formatting to 14 here to match Cabana (#117) and in the near future we'll plan to bump that again to match Kokkos at 16

[cz4rs]

@cz4rs just a note that I updated the formatting to 14 here to match Cabana (#117) and in the near future we'll plan to bump that again to match Kokkos at 16

Alright, I'll rebase this as needed.

[cz4rs]

@streeve I have addressed remaining comments. This is also rebased and re-formatted.

[streeve]

This looks great - couple of wording suggestions, but I think it's ready to merge after that