Marco Caliari schrieb: >>> other solvers): at line 336, it should be >>> >>> vdelta = abs(y5 - y4); > >> So number of evaluations and steps increased dramatically. I can see >> the same for ode45..ode78. Can you send an example where I can see >> results becoming better if I use your modification? > > vdelta is an error estimate (in fact, it is compared few lines below > with the tolerance) > > if (all (vdelta <= vtau)) > > What happens if it is negative? Or, even worse, complex? Of course, > without the absolute value you have less time steps, only because the > comparison with the tolerance is a non-sense. > > Moreover, if you try to solve y'=sqrt(-1)*y you get the complex > conjugate solution. In fact, the conjugate-transpose symbol ' is > erroneously used instead of only-transpose .'
Then yes, you're absolutely right. I've never investigated complex problems before, so I expect there must be more changes that need to be done in the m-files to handle these kind of problems. > I think I will have some time to further investigate the codes (at > least, the m-files). Great, thanks, then I'll wait with that single modification to the m-file solvers until I either get a patch or modified m-files or whatever from you for all the other modifications. Thomas ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Octave-dev mailing list Octave-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/octave-dev
