This is cool. Questions: - I think the original dassl was limited to banded Jacobians but in this implementation any (sparse)-Jacobian works, right? - Are there any of the good old licencing issues here? DASSL comes without one.
On Tuesday, May 6, 2014 9:21:39 AM UTC+1, Ivar Nesje wrote: > > I think a DAE solver would be a great fit for ODE.jl. Currently I think it > will be best to develop it outside, until a few important cleanups is > performed on the ODE solvers. The state of ODE.jl is between a breaking API > discussion and implementation of that described API. It would be very nice > if you could read some of the > discussion<https://github.com/JuliaLang/ODE.jl/search?q=api&ref=cmdform&type=Issues> > and > documentation <https://github.com/JuliaLang/ODE.jl/blob/master/doc/api.md> > and > comment from your point of view. Some consensus has been reached, but there > are still possible to argue for different solutions. > > Ivar > > kl. 10:02:40 UTC+2 tirsdag 6. mai 2014 skrev Alex følgende: >> >> Or submit a PR for including it in ODE.jl :-) >> >> Best, >> >> Alex. >> >> >> On Tuesday, 6 May 2014 09:57:34 UTC+2, Viral Shah wrote: >>> >>> It's great to see a DASSL implementation in Julia! Do register it with >>> METADATA. >>> >>> -viral >>> >>> On Tuesday, May 6, 2014 12:41:20 PM UTC+5:30, Paweł Biernat wrote: >>>> >>>> For the last several weeks I have been implementing [1] a DASSL >>>> algorithm in Julia. The DASSL algorithm was developed by Linda >>>> Petzold in late 80s to solve Differential Algebraic Equations (DAE) >>>> and its Fortran 77 implementation is one of the most popular DAE >>>> solvers. My implementation in Julia bases mainly on a book [3] and >>>> partly on the source of the original implementation [2] (mostly in >>>> places where the book was ambiguous). I tried to mimic the behavior >>>> of the original implementation as much as I could but my code is not >>>> in a one to one correspondence with the original (I aimed at >>>> readability of the algorithm), so don't be surprised if you get >>>> slightly different results from both versions. >>>> >>>> This is my first larger project in Julia so don't expect high end >>>> quality or performance, but I tested my implementation for convergence >>>> on various simple index-0 and index-1 problems and it works just >>>> fine. That said, I am open to any critique and suggestions, and any >>>> help with improving the code. Ultimately, I would like to see this >>>> code merged into the ODE.jl library, if the maintainers of the latter >>>> allow it. >>>> >>>> [1] https://github.com/pwl/dassl.jl >>>> [2] http://www.netlib.org/ode/ddassl.f >>>> [3] "Numerical Solution of Initial'Value Problems in >>>> Differential-Algebraic Equations" K. E. Brenan, S. L. Campbell, L. R. >>>> Petzold >>>> >>>>
