LON computation is currently quite slow, so we should think about ways to speed it up.
Parallelize independent runs
Sampling runs are independent, so this is low-hanging fruit. We can run them in parallel (e.g., via multiprocessing). That alone should reduce wall-clock time significantly.
Things to keep in mind:
- Make sure random seeds are handled properly for reproducibility.
- Avoid shared-state issues.
- Benchmark scaling (speedup vs number of workers).
This is mostly an engineering improvement and should be relatively straightforward.
LON computation is currently quite slow, so we should think about ways to speed it up.
Parallelize independent runs
Sampling runs are independent, so this is low-hanging fruit. We can run them in parallel (e.g., via multiprocessing). That alone should reduce wall-clock time significantly.
Things to keep in mind:
This is mostly an engineering improvement and should be relatively straightforward.