Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: As simple as it is now, I am still against motion 24.03...



> John Pryce wrote:
> On 10 Jun 2011, at 16:15, Joel C. Salomon wrote:
> > Seems to me the list of functions provided in interval form needs to
> > be either exhaustive or extendible.
> >
> > Consider any of the functions this group has rejected from being
> > required for inclusion, or any we haven't thought of.  If I need
> > them, am I stuck implementing them with interval arithmetic on the
> > sub-expressions, giving me a wider-than-necessary result, or does the
> > platform provide a mechanism through which an analyst can extend it?
>
> Does 4.8.4 "User-supplied functions", in my just-circulated draft
> 20110608DecorationSystem.pdf, address your concerns at all?  Or
> are you asking a different question?

I can't speak for Joel, but I thought Joel raised the same issue I did:
Some standardized primitives not strictly needed for end-user IA would
be useful to implementers of IA libraries.  Section 4.8.4 mentions the
use of monotonicity arguments to compute tighter enclosures, but not
how they would be used, thus perhaps implying that only IA primitives
would be used.  Now, if an implementation uses explicit directed rounding
on point arguments, it will of course bear the responsibility of also
dealing with decorations, and in general preserving the properties needed
for proper enclosure: it won't happen automatically.  So this should only
be done by experts.  Experts like to have sharp tools at their disposal,
and having a standard set of such tools would permit fairly generic IA
libraries to be written, allowing the expert programmer to concentrate on
the mathematics without having to cater to numerous idiosyncratic details
on different platforms.

Michel.
---Sent: 2011-06-10 19:07:39 UTC