[Date Prev][Date Next][Thread Prev][Thread Next]
[Date Index]
[Thread Index]
- Subject: Re: not all numbers are treated the same way
- From: David Kastrup <dak@...>
- Date: Thu, 15 Jul 2010 08:36:58 +0200
Pantelis Kavallieratos <kavallieratos@gmail.com> writes:
> one would expect that b=1-(1-a) leads to b=a
Huh? Since when does the associative law hold with floating point
arithmetic?
With _very_ well picked rounding and arithmetic, you can get to the
point where (evaluated from left to right)
a+b-b = a+b-b+b-b
(namely you don't lose _additional_ precision when trying the same
"neutral" operation a second time, having rounded to a more stable
starting point the first time round) but it is not possible to arrive at
a+b-b = a
since the whole point of floating point arithmetic is the acceptance of
small losses of precision during operations.
--
David Kastrup