Intel had an ode solver library (dodesol) around 2008, which was incredibly great for stiff problems. Where lsoda or cvode would fail, dodesol would power through.
I believe internal development was discontinued and the code was provided at some point. I have tried searching for it but to no avail. I only came across similar sentiments and experience from others as I had.
Does anyone here have the source code?
Best I could find was the download page with the author’s names and a filename. I didn’t have much luck with search engines. Maybe email the authors? Let me know if you can find it.
Thanks. I will email the authors.
For stiff ODE solving, perhaps also have a look at the options FATODE provides. I have used their Rodas3 method in cases where dvode would fail; FATODE site
Apparently you can download the library from here. However it only contains dll files which I have not been able to use so far.
something worth noting is that an easy way for an ode solver to appear really robust is to just give up on controlling the error to the extent desired by the user. unless you had work precision diagrams that showed better performance I would be very wary.
Found this on an intel forum. It contains lib files that seem to work.
I can now confirm that the implicit solver in the library works at points where LSODA, DVODE, RADAU5, RODAS, SEULEX all fail. Also seems to be slightly more accurate at the same tolerance (compared to DVODE). The mechanism to select an explicit/implicit solver automatically doesnt seem to work well though.