Re: double money

From: CBFalconer (cbfalconer_at_yahoo.com)
Date: 04/15/04 

Date: Thu, 15 Apr 2004 03:58:45 GMT

Keith Thompson wrote:
>> Dan Pop wrote:
>>
>>> Floating point types can use *any* integer base greater or equal
>>> to two.
>>
... snip ...
>
> But there's a good chance that a new form of decimal floating-point
> may catch on in the not too distant future. It uses 10 bits for
> each 3 decimal digits, so the mantissa representation is nearly
> (1000/1024) as efficient as pure binary.

That is not the point. Normalizing decimal values requires
multiply/divide by 10, so the eventual range, at best, is from 1.0
to 10.0- (9.999...999) for some value of the exponent. For base 2
the corresponding range is from 1.0 to 2.0-. Besides the loss of
a bit position the maximum possible percentage deviation is higher
by 5 for the decimal organization. The extra bit makes that a
factor of 10.

If the normalization is by factors of 1000, as suggested by the 3
dec. digits above, the situation is worse.

So you need a good reason to use any such decimal organization. 

-- 
Churchill and Bush can both be considered wartime leaders, just
as Secretariat and Mr Ed were both horses.  - James Rhodes.