Re: Discussion about accuracy modes
On 2010-10-14 10:00:29 +0100, N.M. Maclaren wrote:
> On Oct 14 2010, Vincent Lefevre wrote:
> >Note that in FP arithmetic, it is very easy to get huge values:
> >just consider the result of an overflow rounded to zero. But
> >perhaps things are different in interval arithmetic (however
> >intervals can also be built from FP numbers). This might not be
> >a problem in most applications in practice, but for standard
> >conformance and security (where users could provide ridiculous
> >data to an online application to yield a DoS), it will.
>
> Yes, but I regard that as poor software engineering, and there is a
> better solution. Such an application should impose an arbitrary
> upper bound, and raise an exception if that is exceeded every time
> it becomes relevant.
This would mean that it would no longer be conforming to the standard.
> For example, refusing to handle speeds faster than that of light is
> a safe rule in 99.99% of all known programs :-)
Not all applications are related to physics.
> Anyway, I am not disagreeing with your point - there are LOTS of
> functions with no known realistic algorithm for arbitrary precision.
Here the problem is *not* arbitrary precision, but a large exponent
range. So, the problem would exist if IA were implemented with the
DPE[*] format even though it has a 53-bit precision.
[*] https://gforge.inria.fr/projects/dpe
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / Arénaire project (LIP, ENS-Lyon)