LAMMPS PACE status

[pearce8]

We will keep the existing LJ potential workload, and add a new workload to use the ACE model.

Components

1. Source code, Build system: CMake
2. Source code released. The source code is in the latest LAMMPS release.
3. Spack package.py: Compile LAMMPS with the PACE package installed -D PKG_ML-PACE=yes Add lammps pace workload #1106
4. Ramble application.py
5. Benchpark experiment.py
6. +cuda experiment.py
7. +rocm experiment.py
8. Benchmark dependencies: Kokkos, Pace
9. small test size problem for CI: Test problem for the PACE and the hns-reaxff workloads runs on dane, matrix and tuolumne
10. Scaling (Weak, Strong, or both)
11. Large scale problem size defined
12. CI running on CPUs/CDash
13. CI running on NVIDIA GPUs/CDash
14. CI running on AMD GPUs/CDash
15. +single_memory experiment.py
16. +caliper experiment.py
17. CPU-only performance analysis
18. CUDA performance analysis
19. ROCm performance analysis
20. Documentation (build, run instructions, performance demo): https://docs.lammps.org/Build_extras.html#ml-pace. https://github.com/lammps/lammps/blob/develop/examples/PACKAGES/pace/in.pace.product
21. FOM computation script