Nicholaswogan/numbalsoda

Potential interpolant issue

Patol75 opened this issue · 2 comments

Patol75 commented

Hi @Nicholaswogan,

Great work on this repository, it looks like you managed a consistent and quite high speed-up compared to SciPy. I have been working with LSODA recently using solve_ivp, and I have encountered an issue relative to the interpolant (PR at SciPy). I am wondering if this implementation of LSODA also has a similar issue, as it looks to me that the root of the problem lies within the FORTRAN implementation. In particular, LSODA::intdy seems very similar to me to DINTDY in ODEPACK, which could well mean that the issue is present. Let me know if you can have a look.

Nicholaswogan commented

Thanks! This is interesting and seem hard to pin down and understand. I'll spend a bit of time on it. It might be worth opening an issue at LSODA.jl : https://github.com/rveltz/LSODA.jl , if the issue persists there. They might shed some light on the problem that would help this repository.

Patol75 commented

I have had a look at LSODA.jl, and I can confirm there also looks to be an interpolant problem there - I will open an issue.

  • 👍1