Re: Motion 13
Ulrich Kulisch> Would'nt [SNaN, SNaN] be an appropriate encoding for the empty set?
I think [QNaN, QNaN] would be a better encoding for the empty set than [SNaN, SNaN].
A Signaling NaN normally terminates the program at the point it's used (unless you've written a trap handler for it), while a Quiet NaN does not. I wouldn't expect doing unions and intersections on intervals that happened to be empty to terminate my program. I would like using an uninitialized variable to, unless I explicitly check for it being uninitialized or recover from the exception, because it means the program is wrong and everything it produces from that variable is garbage.
- Ian McIntosh IBM Canada Lab Compiler Back End Support and Development
Ulrich Kulisch ---05/07/2010 11:33:30 AM---Ian McIntosh schrieb: >
From: |
Ulrich Kulisch <Ulrich.Kulisch@xxxxxxxxxxx> |
To: |
Ian McIntosh/Toronto/IBM@IBMCA |
Cc: |
stds-1788@xxxxxxxxxxxxxxxxx |
Date: |
05/07/2010 11:33 AM |
Subject: |
Re: Motion 13 |
Ian McIntosh schrieb:
>
> Detecting uninitialized variables is one of the primary uses of 754's
> signaling NaNs. In an IA implementation based on 754, the interval
> [SNaN, SNaN] would be the equivalent. I suppose the corresponding name
> would be SNaI. Uninitialized variables are one of programming's
> greatest evils, and SNaI would be better than silently using empty or
> entire or [0, 0].
>
Would'nt [SNaN, SNaN] be an appropriate encoding for the empty set?
>
>
> - Ian McIntosh IBM Canada Lab Compiler Back End Support and Development
>
>
> Inactive hide details for Vincent Lefevre ---05/07/2010 04:55:57
> AM---On 2010-05-05 18:40:08 +0200, Ulrich Kulisch wrote: > ThiVincent
> Lefevre ---05/07/2010 04:55:57 AM---On 2010-05-05 18:40:08 +0200,
> Ulrich Kulisch wrote: > This could be a possible use for the empty set
>
>
> From:
> Vincent Lefevre <vincent@xxxxxxxxxx>
>
> To:
> Ian McIntosh/Toronto/IBM@IBMCA
>
> Date:
> 05/07/2010 04:55 AM
>
> Subject:
> Re: Motion 13
>
> ------------------------------------------------------------------------
>
>
>
> On 2010-05-05 18:40:08 +0200, Ulrich Kulisch wrote:
> > This could be a possible use for the empty set:
> > Occasionally it happens in programming that a variable is used
> > before a definite value has been assigned to it. This is a terrible
> > error in interval arithmetic and should be avoided by checking for
> > it. A solution could be to set every variable at definition to be
> > the empty set.
>
> It could be NaI if defined. Or it could be the entire set: this would
> allow to correctly handle variables that should have been initialized
> to some value (or interval) but this was forgotten, but programming
> errors (such as typos in variable names) may not be detected.
>
> --
> 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)
>
>
--
Karlsruher Institut für Technologie (KIT)
Institut für Angewandte und Numerische Mathematik (IANM2)
D-76128 Karlsruhe, Germany
Prof. Ulrich Kulisch
Telefon: +49 721 608-2680
Fax: +49 721 608-6679
E-Mail: ulrich.kulisch@xxxxxxx
www.kit.edu
www.math.kit.edu/ianm2/~kulisch/
KIT - Universität des Landes Baden-Württemberg und nationales Großforschungszentrum in der Helmholtz-Gemeinschaft