[vox-tech] ODE solvers in C

[Matt Holland]

Pete,

Thanks for the perspective.  Still, I'm cautiously optimistic.  This is 
really just a slight tweak on a pretty simple ecological model.  The 
system is nonlinear with coupled equations, but I doubt that it's going 
to be as troublesome as your system was.

> on C and fortran...
>
> before going down this route, be aware that there are some serious
> issues with integrating fortran and C object files.  serious enough that
> when i was in your position a few years ago, i opted to NOT do the
> fortran/C thing.  for a good discussion on the many pitfalls of
> integrating C and fortran, check out "computational physics" by landau
> and paez (wiley 1997) chapter 15.  you'll hate what you see.

Thanks for the warning.  I'll look into that.  Hell, maybe I'll just 
learn Fortran, I know it can't be that hard... just antiquated and 
weird, no?

> on using canned code...
>
> i didn't mention anything because i want to discourage using other
> peoples' code (especially when you have no idea what a "good" answer
> will look like ahead of time).

I tend to differ with you there.  If there's well-established code that 
solves a problem, and I can understand the interface to that code, I'm 
inclined to use it.  The trouble comes when the code is poorly 
documented (which, admittedly, may be the case more often than not) or 
hasn't been tested sufficiently.  Unfortunately, those may both be 
problems when it comes to C libraries for numerics, because it seems 
like most people continue to use Fortran for those sorts of things.  And 
the old tried and true code is probably much better than anything I 
could write with out changing my area of specialization.

<snippage>


> 7. don't get distracted by linux.   ;)

Actually, I'm much more likely to get distracted by OS X, since that's 
what I'm running on my new iceBook that I picked up a few weeks ago.  
Oops, did I say that on a LUGOD list?

> disclaimer...
>
> take what i say with a grain of salt.

Done :)

Matt