On Tue, 27 Apr 2004 15:57:53 -0300, Juan Vuletich 
<jmvsqueak at uolsinectis.com.ar> wrote:

>
> Yoshiki,
>
> Well, I did my homework (a bit too late). I googled "squeak-dev immediate
> oop" and "squeak-dev immediate float". Lots of interesting reading,
> extensive discussions on this topic. Apparently anything but 64bit IEEE
> Floats (as Squeak Floats) is almost useless.
>
> So, we probably won't have immediate Floats until we have 128 bits
> machines...
>

Wouldn't a 61 bit float with the 3 bits missing from the exponent be 
pretty useful?  I'd been lead to believe that the usual problem with 
floating point math was in rounding in the mantissa, but that the high 
bits of the exponent are very infrequently used.

-andrew
-- 
andrew_c_berg at yahoo.com