[ODE] ODE .NET Bindings Performance Issues, Unsafe code, etc...

[Jason Perkins]

On 10/30/06, Gonçalo Lopes <goncaloclopes at gmail.com> wrote:
> Since my concerns considering efficiency and usability of .NET
> wrappers to ODE held even after strong argumentation in this list, I
> decided to pursue the implementation of a complete and fully
> documented lower- (P/Invoke bindings) and high-level (Class hierarchy)
> managed API to this amazing library.

First of all, I appreciate the work you did on this binding. Wrapping
the entirety of ODE is a big undertaking, and you have clearly put a
lot of work into it. However...

(sorry)

The decisions you have made in your binding, particularly in simple
data objects like dMass and dMatrix, not only make the unnecessarily
complex, it also makes it much slower. You are doing numerous
unnecessary allocations, and completely removing any possibility of
JIT optimizations. I think you have might have gotten some bad
information on the performance bottlenecks of .NET.

Obviously there is interest in getting an "official" .NET binding
done, and now that I am finally crawling out from the depths here I
will make it a priority. I had hoped to test everything before I added
it; I'll skip that step to get it to you more quickly.

I am going to continue along the path of my original binding,
currently in SVN. I am trying to make it programmer-friendy, and also
as fast as possible. I'm open to suggestions for improvements on
either front.

Jason