Version 1.8.0
-
Fixed a bug in the error estimation. This bug lead to errors being overestimated for time steps smaller than 1 and underestimated for time steps larger than 1. As a result, dynamics with time scales smaller than one were integrated overly accurately (and slowly), while dynamics with larger time scales were integrated less accurately than intended. It is unlikely, though not impossible, that this had relevant impacts on results where the step size was larger than 1. There is a reason why I did not notice this bug earlier.
-
step_on_discontinuities
now adapts time steps. -
generate_lambda
now allows to control simplification. -
Some improvements to documentation.