AMG2023 status

Components

1. Source code, Build system: CMake, Hypre uses Autotools. Debugging Problem 2 issue - Arjun and Rui Peng to debug mid-June.
2. Source code released
3. Spack package.py
4. Ramble application.py
5. Benchpark experiment.py
6. +cuda experiment.py: Link errors for cublas #1067
7. +rocm experiment.py: Tuolumne build error #1068 -- seems to be resolved?
8. Dependencies: Hypre, Umpire. Umpire currently not on #1069 https://github.com/LLNL/AMG2023/pull/14 #1075. OOM error https://github.com/LLNL/AMG2023/issues/15 https://github.com/hypre-space/hypre/pull/1421. Allocation outside of the memory pool in Setup https://github.com/LLNL/AMG2023/issues/16. @rfhaque to follow up with Rui Peng. https://github.com/hypre-space/hypre/pull/1421. https://github.com/llnl/AMG2023/pull/17
9. small test size problem for CI
10. Scaling (Weak, Strong, or both)
11. Large scale problem size defined. @rfhaque to follow up with Arjun.
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: will not pursue
16. +caliper experiment.py
17. CPU-only performance analysis: Dane FOM low for Problem 1 #1070
18. CUDA performance analysis
19. ROCm performance analysis
20. Documentation (build, run instructions, performance demo): https://lanl.github.io/benchmarks/02_amg/amg.html
21. https://github.com/llnl/benchpark/issues/875
22. Integer overflow #1252
23. MPI bug in Problem 2 https://github.com/llnl/AMG2023/pull/18, also hangs - use STAT to diagnose where.
24. Need to test AMG2023 as part of Hypre testing so it is not out of date
25. Core dump on 64 nodes, 256 ranks, 360^3 elements. Different problem sizes fails at different node count. Need to give Arjun the logs