Integration Method?

0 views
Skip to first unread message

Matt Hayes

unread,
Oct 13, 2008, 6:14:48 PM10/13/08
to glaze Engine
So, I've been doing as much research as I can on Glaze/Chipmunk and
Box2DFlash/Box2D, and it's kind of starting to make sense a little. I
had a thought though... have done some research into physics for game
prior to AS3, I remember reading a lot about the Verlet integration
method, and how fast/stable it was over time, and how easy to
implement... I get the impression that the main time-saver in these
engines is the contact caching... so first question I guess, is that
true?

I understand also, that Glaze/Chipmunk use a hash table for broad-
phase collision detection by default, I assume this is because it's
the fastest overall, was a lot of testing done, or is it a programming
preference that works? Does Box2D just use the "Brute Force" method?

Would it be possible to vary the integration method... or to create an
engine that used Verlet integration, contact caching, and a hash table
(or quad-tree, or something else even?) for broad phase collisions?
Would this engine then be super fast? Could it also easily deal with
particles?

Maybe Verlet _just_ is for particles, and it's no good here because
the whole thing with contact caching means dealing with impulses, and
therefore velocity _has_ to be a part of the equation... I'll admit
I'm out of my depth, but am I intuiting it correctly?

Thanks in advance,
//Matt Hayes

Matt Hayes

unread,
Oct 14, 2008, 12:09:21 AM10/14/08
to glaze Engine
Hi,

I already thought of a follow up question, this one may be for Richard
Jewson himself, maybe for any contributor... I guess I don't know.

Anyway, I was wondering about what went into optimizing the code for
the AVM 2? I read about changing arrays to linked lists, but is that
all? What else?

Thanks again,
//Matt Hayes
Reply all
Reply to author
Forward
0 new messages