[ODE] Re: Cross compiling ODE for Pocket PC

[GMANE]

Nate,

Thank you for the info.

I've compiled the core ODE library for Pocket PC now. It compiled pretty
much out of the box after I added a custom makefile and removed the
configurator stuff. Initial tests on my ARM based PDA indicate that it will
be fast enough for my purposes (1500 worldsteps per second on two bodies and
one joint with emulated fp).

I'll create diffs and submit them at some point, in case anyone else wants
to run ODE on Pocket PC :) I'm not planing on trying to compile any of the
ODE test stuff though. That would be much harder. I guess for optimization
it might also be a good idea at some point to look at setting up the fbuild
stuff (I'm just using the default files now).

Roger

"Nate W" <coding@natew.com> wrote in message
news:Pine.LNX.4.21.0307022211540.4430-100000@spiral...
> On Wed, 2 Jul 2003, GMANE wrote:
>
> > [...] I would have to compile ODE with the free embedded Visual Studio
> > cross compiler. Taking a quick look at the compilation process for
> > ODE, it looks like it is assumed that ODE will be compiled on the same
> > platform that it will run on, as an executable is first created and
> > then the executable itself controls compilation of the rest of the
> > library. Anyone know why it is done this way and how much work it
> > would be to change it?
>
> The first exe you speak of just generates config.h, but you can also
> create that header file my hand if you wish.  Try making one the usual
> way, and look for #defines and const declarations and such that might need
> to be changed for the PocketPC platform.  That will probably be a
> trial-and-error process, but if you make it work, it would be very cool to
> update configurator.c with your findings.
>
> > The other hurdle would be that floating point is emulated on most (if
not
> > all?) Pocket PC PDAs. The physics that I plan on doing are so simple
that
> > that might be ok. If it's not, I would have to create support for
integer
> > math in ODE. How much work would that be?
>
> I would expect floating point emulation to be available, perhaps as a
> linker option, but I'm just guessing.
>
> -- 
>
> Nate Waddoups
> Redmond WA USA
> http://www.natew.com