Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!ukc!edcastle!ecwu61
From: ecwu61@castle.ed.ac.uk (R Renwick)
Newsgroups: comp.sys.acorn
Subject: Dividing
Message-ID: <8828@castle.ed.ac.uk>
Date: 2 Mar 91 19:47:07 GMT
Sender: ecwu61@castle.ed.ac.uk
Organization: Edinburgh University
Lines: 27


	Can anyone help me with this problem:
In my super-duper, wire-frame model animating 'C' program, I represent 
floating point numbers as 4-byte integers. This I do by shifting the number
being represented by 15 bits (multiply by 32768) in order that I can keep a
track of the fractional part. The reason for doing this is that I want
to make sure my program runs as fast as possible by not using floating
point arithmetic...

	What I want to do is do a divide a number by another number
and not lose alot of the precision.
	The way I do this at the moment is by
		 result=((a/b)<<15)
But if a and b are close together then (a/b) loses alot of precison
(since the result is an int).

Does anyone know how I could do an integer division without losing
precision. I cannot use result=(((a<<8)/b)<<7) or any other tricks
because a may be close to 32768<<15 already...
Perhaps if I can find the reciprocal of b in the above representation
and multiply it by a then I'll not lose the as much precision, but I
don't know how to do this (find the reciprocal) :-(

Anyway, if anyone knows how to solve this one, I'd be very grateful to
hear from them :-).

Rik