It's ~25% faster than before, especially when the learn_span is long.
The time complexity of simulate() is O(kn^2), where k is learn_limit and n is learn_span. It costs ~30s to find the optimal retention with learn_span=3650 and learn_limit=10. In previous implementation, it costs ~40s.
cargo test --release -- optimal_retention::tests::optimal_retention
It's not recommended to set a learn_span greater than 3650.
It costs ~250s to find the optimal retention with learn_span=10000.
I hope it is the last improvement on FSRS-rs before the release of Anki 24.04. @dae
It's ~25% faster than before, especially when the
learn_spanis long.The time complexity of
simulate()is O(kn^2), where k islearn_limitand n islearn_span. It costs ~30s to find the optimal retention withlearn_span=3650andlearn_limit=10. In previous implementation, it costs ~40s.It's not recommended to set a
learn_spangreater than 3650.It costs ~250s to find the optimal retention with
learn_span=10000.I hope it is the last improvement on FSRS-rs before the release of Anki 24.04. @dae