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

Re: resend [IEEE P1788] A second required: Motion P1788/M0002.01_ProcessStructure

Dear Nate and forum members,
I suggest that keeping the same general structure with the following addition 


A level 1' is added
This level should become the interface to conceptors of agorithms instead of level 1 since it describe the different views of level 1 
including

         *The Vienna proposal intervals
          *The Kaucher's interval and the modal intervals
           *The Kahan's intervals
            *May be The midpoint radius model
From this point of view the objets of level 1' are deduced from object of level 1 fy focusing the attention on some specific features Conversely objet of level 1 are deduced from the objets of level 1 by some integration of features
The extra level is not numbered 0 since objects of level 1 are are derived from the objects of the extra level.
On the other hand objects of levels 2 and following are derived from objects of level 1 and not from level 1' 

IMHO, it should not to an overloaded and the unsable standard.
In contrast it might make the thing more simple, due to a deeper understanding of interval arithmetic
Nate Hayes a écrit : 
Dear John and forum members:
"The levels framework, here proposed, does not give much help to discussing modal or Kaucher intervals, because these are more than “plain sets” (of numbers);"
It is sufficient. Kaucher intervals can be thought of as "plain sets" that simply have algebraic properties which are otherwise lacking in the traditional set of intervals [a,b] with a <= b. They are not too much unlike the Kahan intervals in this respect.
Modal intervals, with quantifiers "for all" and "there exists," coinicide entirely with the Kaucher intervals. This is also true for the "directed" intervals of Popova, Markov, et. al. So from a standards perspective, it is sufficient to simply consider the Kaucher intervals.
In my upcoming paper, I show there is a complete mapping between the Level 1 and Level 2 layers for the Kaucher intervals. This mapping has many desireable properties, e.g., it has correct overflow tracking, minimal deviation from IEEE 754, and it provides a standardized meaning for every interval comprised of any possible combination of two IEEE 754 values. It is compatible with several different methods of underflow tracking as well as classical set-theoretic interval algorithms, such as the extended interval Newton method. At Sunfish, we have already applied this mapping with success, for example, to prototypes of a deeply pipelined Kaucher interval processor. 

Nate Hayes
Sunfish Studio, LLC
----- Original Message ----- From: "Ralph Baker Kearfott" <rbk@xxxxxxxxxxxxx> 
To: "stds-1788" <stds-1788@xxxxxxxxxxxxxxxxx>
Sent: Sunday, February 01, 2009 3:41 PM
Subject: resend [IEEE P1788] A second required: Motion P1788/M0002.01_ProcessStructure 



P1788 members:

Evidently there was something wrong witth the attachment, the
first time I sent it.  Thus, I am re-sending it.

Baker

===========================================================
IEEE P1788 working group members:

John Pryce (one of our technical editors), has made the following
motion, with appended rationale and attached document.
For this motion to succeed, according to our Policies and
Procedures paragraphs 10.5 and 11.2.  Seconding and voting
on this motion may be public, that is, the second and votes
may be posted to this list.

To formally start the voting, a "second" needs to be posted.

Sincerely,

R. Baker Kearfott
(acting chair, P1788)

P.S. In general, I will not attach large position papers
      to email to the group, but will reference them
      by URL.  However, I have judged this paper, at roughly 60KB,
      not to be inordinately large.  Please inform me privately
      if sending papers that size seems to be a problem.

===Motion P1788/M0002.01_ProcessStructure===
Proposer: John Pryce
Seconder: required

===Motion text===
The P1788 Working Group adopts the principles set out in
sections 1 and 2 of Position Paper PP008, "A proposed structure
for the process of constructing the P1788 standard".

===Rationale===
For a rationale, please read Section 1 of the position paper.
The paper is available on the P1788 web site.

Note that this is a vote on principles, NOT on the detail in the
following section 3 and appendix. P1788 members may see various
defects of commission or omission in those, and are asked to
point them out, as a separate issue from this motion.
=====



--

---------------------------------------------------------------
R. Baker Kearfott,    rbk@xxxxxxxxxxxxx   (337) 482-5346 (fax)
(337) 482-5270 (work)                     (337) 993-1827 (home)
URL: http://interval.louisiana.edu/kearfott.html
Department of Mathematics, University of Louisiana at Lafayette
(Room 217 Maxim D. Doucet Hall, 1403 Johnston Street)
Box 4-1010, Lafayette, LA 70504-1010, USA
---------------------------------------------------------------









--
Dr Dominique LOHEZ
ISEN
41, Bd Vauban
F59046 LILLE
France

Phone : +33 (0)3 20 30 40 71
Email: Dominique.Lohez@xxxxxxx