Path: chuka.playstation.co.uk!news
From: george_bain@playstation.sony.com (George Bain)
Newsgroups: scee.yaroze.programming.2d_graphics
Subject: Re: Collision detection with rotating sprites
Date: 4 Dec 1998 09:31:47 GMT
Organization: Sony Computer Entertainment Europe
Lines: 42
Message-ID: <748a63$c3e3@chuka.playstation.co.uk>
References: <01be1d86$03a1a160$0a3ddec2@guildfoc> <742u50$9lv3@chuka.playstation.co.uk> <01be1e90$e7e94160$0a3ddec2@guildfoc> <746eq9$c3e2@chuka.playstation.co.uk> <3666DDF3.EBEFDB1E@cthullu.freeserve.co.uk> <01be1ef5$5bfd4ba0$LocalHost@guildfoc>
NNTP-Posting-Host: mailgate.scee.sony.co.uk
Mime-Version: 1.0
Content-Type: Text/Plain; charset=US-ASCII
X-Newsreader: WinVN 0.99.8 (x86 32bit)


Hi Miles,

I think you might have to decide whether you want
your routine to be very accurate or very fast.  I 
tend to stay far clear from sqrt().

George



In article <01be1ef5$5bfd4ba0$LocalHost@guildfoc>, miles@guildfrd.demon.co.uk 
says...
>
>Yes, I've had a go with some broad tests. The method I use is less accurate
>that yours, but probably faster - check a bounding square based on the
>longest axis of the sprite first and then use a more precise bounding
>sphere test (using the anti-christ of fast code - Sqrt).
>
>I like bounding spheres but there is just no quick way to calculate square
>root as far as I can see. Even with a fixed point maths routine I still
>have to perform at least 12+ iterations to get an reasonable approximation
>of sqrt. Has anyone out there found a really quick and dirty way of
>calculating it?
>
>By the way, for anyone reading this who hasn't got into fixed point maths
>yet - do so, very quickly! For the last few years I've been used to
>programming PCs with powerful maths co-processors, I did some tests using
>floating point values on the playstation and couldn't believe how sloooooow
>it is - but use fixed point and it really motors!
>
>- Miles
>
>Jake Turner <jaket@cthullu.freeserve.co.uk> wrote:
>
>> How about doing simple rectangle bounding box detection first (simple
>> approximation) and then doing the proper tests.
>
>
>
>