🚀 go-pugleaf

In article <phoyer-0701961703010001@192.0.2.1>, phoyer@snakemail.hut.fi (Patrik Hoyer) writes:
> Hi,
> 
> I'm planning on using fixed point math for my 3d engine. (For those of you
> ready to recommend floating point on the pentium of powerpc... I'm
> targeting low end computers. No texture mapping, just flat-shaded
> polygons.) Now, I'd like to find some info on problems and solutions
> relating to using fixed instead of floating point. (The problems of
> overflow, and that of numbers close to zero losing all precision comes to
> mind). Source code would also help, I guess. I'd prefer online sources of
> information (cheap) but if you know of any good books covering the subject
> feel free to recommend them too!
> 

You might look at the Motorola Fast Floating Point (FFP) package.
It runs on plain 68000s and uses 24bit mantissa, 7 bit exponent, 1 bit sign.
It comes standard with every Amiga and should also be available from Motorola.