Path: chuka.playstation.co.uk!news
From: Toby Hutton <thutton@cybec.com.au>
Newsgroups: scee.yaroze.beginners
Subject: Re: Clever bit manipulation routines?
Date: 27 Jul 1998 13:14:14 +1000
Organization: PlayStation Net Yaroze (SCEE)
Lines: 22
Sender: thutton@TECH10
Message-ID: <uyatg7zp5.fsf@cybec.com.au>
References: <35B79502.BC7489D6@nospam.easynet.co.uk>
	<35B7BB71.D869219@compuserve.com> <ug1fsatq2.fsf@cybec.com.au>
	<slrn6rg196.6j6.tjs@indy19.cs.monash.edu.au>
NNTP-Posting-Host: 203.103.154.235
X-Newsreader: Gnus v5.3/Emacs 19.34

tjs@cs.monash.edu.au (Toby Sargeant) writes:

> 
> On 24 Jul 1998 12:05:57 +1000, Toby Hutton <thutton@cybec.com.au> wrote:
> [snip]
> >The libraries often use 12 bit fixed point math.  This means that 4096
> >is 1, 2048 is 1/2 and 512 will be an eighth.  512 in hex is 0x0200 and
> >511 is 0x01ff.  To limit a value between 0 and +511 you simply:
> >
> >val &= 0x1ff;
> 
> sadly, that won't do what was wanted. That will wrap all values into a range,
> whereas the intention was to compute f(x), where f(x)=-45 if x<-45, f(x)=x
> if -45<=x<=45, f(x)=45 if 45<x.

f(x) that you just described is a range, no?  Don't let x get less
than -45 or greater than 45.  What?  Hey?

If you used +/-31 instead of +/-45, why wouldn't 
x = ((x + 0x20) & 0x3f) - 0x20; work?
-- 
Toby.