From: Mark Probert <probertm@...>
Date: 2002-12-28T12:59:57+09:00
Subject: Re: 1210 / 100 = 12? what?


Hi, Mark.

At 04:12 AM 12/28/2002 +0900, you wrote:

>... why one might want to use fixed point rather than floating point 
>arithmetic

You basically have it: speed and accuracy (for some types of calculations).

If you know the scaling factor that you are going to use, then the use
of fixed point can be a big speed factor, if you don't have floating
point acceleration.  If you do, then your mileage will vary, as will
your accuracy.

For a modern general purpose computer, that means that fixed point is
mostly useful for accuracy -- none of the lingering doubts about rounding
errors.  On an embedded micro-controller, where cycles and/or memory are
precious, then fixed point is the way to go.

Personally, if I can get away with fixed point, then I will use it.

The other part of your question, the part about remainders, deals with 
scaling.
If I do:

  irb: 1210 / 100    # 12
  irb: 1210 % 100    # 10   (scaled: 10 / 100)

So, if I was working in dollars and cents (scaled to cents), then the modulus
returns me the working number of cents.

Hope this helps a little.



-mark

-----
Mark Probert
GNPS - Optera Metro 3000
Tel (613) 768-1082 (ESN: 398-1082)
Mob (613) 715-2045
email: probertm@nortelnetworks.com