On Decimal Floating-Point

On Decimal Floating-Point

Post by jsavar » Mon, 02 May 2005 17:26:29



If one could just leave it at that, they wouldn't have put an "inexact"
flag in IEEE 754!

algorithm:

Actually, the fast divide by 10 algorithm is discussed in the *latter*
location as well.

John Savard
 
 
 

On Decimal Floating-Point

Post by Dik T. Win » Tue, 03 May 2005 09:24:54

In article < XXXX@XXXXX.COM > XXXX@XXXXX.COM writes:

> >To me that's almost a bug, not a feature:
> >People should learn that fp isn't exact. :-)
>
> You have a good point. For many purposes, calculator-style
> floating-point is misleading.

What is calculator-style floating-point? Where (say) 15 digits are
calculated but only 12 displayed?
--
dik t. winter, cwi, kruislaan 413, 1098 sj amsterdam, nederland, +31205924131
home: bovenover 215, 1025 jn amsterdam, nederland; http://www.yqcomputer.com/ ~dik/

 
 
 

On Decimal Floating-Point

Post by jsavar » Tue, 03 May 2005 12:07:02


Although that certainly is a feature of floating-point arithmetic on
many calculators, and an entirely legitimate one, I was just referring
to decimal floating-point in general, as opposed to binary floating
point, which rather more often proves itself not to be doing what we
might expect.

John Savard