The code from ODE.jl is clearly based on the Hairer and Wanner monograph, but
it is not clear at times how it has been changed. Things that need to be
investigated:
Some of the code has some magical constants like timeout_const = 5 which
feel unmotivated in the current form.
It is not clear that the stiffness check has been ported over.
Features that are missing that are known include:
[] the Hermite interpolation is of the incorrect order
[X] Dopri5 should have 4th order
[] Dop853 should have 7th order
[X] Really there are special forms derived by Shampine for the forms of these
Hermite coefficients that take advantage of the ode formulation. These
should simply be ported over from the original codes.
The code from
ODE.jl
is clearly based on the Hairer and Wanner monograph, but it is not clear at times how it has been changed. Things that need to be investigated:timeout_const = 5
which feel unmotivated in the current form.Features that are missing that are known include: