Re: fixed point math?

[Ranjit Mathew <rmathew at hotmail dot com>]

> Does anybody know if there's a quick, easy way to compile a program
> that uses lots of float's so that the floats are represented as
> fixed-point values?  An added bonus would be continuing to generate
> floating point ops for double's, but that's not a requirement.
> 
> The only idea I've come up with is rewriting libfloat and using
> -msoft-float.  Is there an easier or cleaner way?

Some dumb questions related to this:

Does -msoft-float imply using the SoftFloat library:

    http://www.jhauser.us/arithmetic/SoftFloat.html

If not, you might want to take a dekko at this library.

However, I'm a bit confused by your requirements - you
obviously are not targetting an embedded platform without
an FPU, as you actually would like the doubles to be
using floating-point operations, so why do you need
this?

Is it for speed? AFAIK, x86 processors since the Pentium
have had very good and fast floating-point support - this
is what made John Carmack switch from using fixed-point
for DOOM to using floating-point for Quake and subsequent
games from iD Software.

I guess you realise that floating-point and fixed-point
numbers have very different ranges and representability
(for large numbers), so you should be prepared for
these in your arithmetic.

Sorry if I seem to have completely missed the point.

Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/