Benchmarking

[PMeira]

Random thoughts for future reproducible benchmarks. EDIT: work has started and results should be published in the coming weeks.

• [x] Select sample networks for small, medium and large networks, ideally among test cases from the official OpenDSS repository. Whole substation, single low voltage networks, etc.
• [ ] Consider different kinds of machines: mobile i5/i7, desktop i5/i7, low cost Celeron, large Xeons
• API level benchmarks:
  • [x] COM
  • [ ] DDLL
  • [x] DSS_CAPI-CFFI
  • [x] DSS_CAPI-Cython
  • [x] Retest with PyPy without using NumPy arrays (slow on PyPy) or wait for cpyext to be finished.
• FPC vs Delphi comparisons
  • [ ] runtime
  • [ ] memory
• [ ] FPC Windows vs FPC Linux (same machine, probably a laptop)
• [ ] Provide recommendations according to the results.

[PMeira]

Initial memory observations: from running the benchmark, COM uses 2x the memory of DSS Python.

[PMeira]

Moving this to a private issue tracker.