On Saturday 12 February 2005 19:56, Daniel Phillips wrote:
> Hi Lourens,
>
> On Saturday 12 February 2005 08:09, Lourens Veen wrote:
> > On Saturday 12 February 2005 04:56, Daniel Phillips wrote:
> > > Excuse me, I attached attached yesterday's code to today's email.
> > > Here's the updated code.
> >
> > Took me a while to realise, but I've found the problem. You get 18
> > bits relative to the [1, 2) range, I get 17 bits relative to the
> > (0.5, 1] range. These are equivalent, because the exponent changes.
>
> This stands to reason because your interpolation expression is the same
> as mine.  I presume you've arranged for your multiply to be unsigned.
> There might be differences in the way the table samples are calculated,
> I didn't check that.

It doesn't really matter that much anyway. As long as it works, it's fine with 
me :-).

> The interpolation should be done in the [1, 2) range to match the range
> of the mantissa.

Actually, I'm inputting the range [0, 65536), and outputting the range [0, 
65536). That's what we need in hardware. The exponent is just a subtraction, 
and the sign remains the same.

Lourens
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to