[ODE] Re[2]: Opcode's LSS vs OBB

[Nguyen Binh]

Hi Pierre,

PT> Why don't you just try ?
    Just too lazy and I'd like to get some idea from you first. Maybe you
    have some good thing on your TODO list but too busy to update :).

>>    For the case of LSS-trimesh collider, is it better to use LSS or
>>    OBB for broad phase?

PT> Same as above. The faster, the better, but I don't remember which one wins.
PT> Because, frankly, they're both slower than.... the planes-collider,
PT> surprise. Create an OBB around your LSS, then use the PlanesCollider with
PT> the 6 planes around the OBB (there's a function doing this already). It
PT> should be faster.
    Yes, so it really better to ask you first. :). I'll play with your
    suggestion.

PT> (shrug)

PT> As I already said on this list, the fastest way to do it is to skip the
PT> collision query completely => temporal coherence + caches. So there's no
PT> point in "optimizing" it further.

    Hmmm..that's another point. I didn't see the use temporal
    coherence plus caching in ODE. I remembered Erwin had a codepath
    that use temporal coherence but he didn't suggest that way -> I
    will make some experiments to use it in ODE...
    
PT> That being said, I'm not sure what you mean exactly. It's "optimized" since
PT> it's already a O(log n) version instead of O(n). The implementation could be
PT> a bit faster maybe, for example using a SAT-based LSS-box routine instead of
PT> a distance query, but there's very little to win out there.

PT> Now of course, if you have a better way to drastically "optimize" it, well,
PT> feel free to share.
    :) . Nope!

-- 
Best regards,

---------------------------------------------------------------------
   Nguyen Binh
   Software Engineer
   Glass Egg Digital Media
   
   E.Town Building  
   7th Floor, 364 CongHoa Street
   Tan Binh District,
   HoChiMinh City,
   VietNam,

   Phone : +84 8 8109018
   Fax   : +84 8 8109013

     www.glassegg.com
---------------------------------------------------------------------