Re: Relation between P1788 and P1788.1

[Ned Nedialkov <nedialk@xxxxxxxxxxx>]

Dima, 

On Dec 2, 2014, at 8:28 AM, Dmitry Nadezhin <dmitry.nadezhin@xxxxxxxxxx> wrote:

John,

It cannot be a flavor, because it doesn't have the full complement of operations required in §9 of the full standard.

This is my suggestion to make it a flavor.

The list of absent operations is not large (5 operations):
RoundTiesToEven(x)
RoundTiesAway(x)
cancelPlus(x,y)
cancelMinus(x,y)
textToInterval(s) that can parse the full syntax, though with relaxed accuracy on some literals

It is more than just the 5 operations above; see also the attached. 

Is it a trouble to include them in 1788.1 ?

Their implementation requires some tricks, but we can suggest
implementation templates under permissive license.

One of the goals is to keep the basic standard as simple as possible.  

This is from the PAR I submitted to IEEE

5.2 Scope: This standard specifies interval arithmetic (IA) operations based on intervals whose endpoints are IEEE binary64 floating-point numbers and a decoration system for exception-free computations and propagation of properties of the computed results.

This standard is a subset of the (full) IEEE P1788 Standard for Interval Arithmetic and includes those operations that in the the editors' view are most commonly used. A program built on top of an implementation of the basic standard should compile and run using an implementation of the full standard, or any superset of the former.

5.4 Purpose: This standards aims to be minimalistic, compared to the full standard, and to cover most of the functionality needed for interval computations. As such, the basic standard is more accessible, will be much easier to implement, and will speed up production of implementations.

Ned

Attachment: basicops.pdf
Description: Adobe PDF document

 -Dima

----- Original Message -----
From: j.d.pryce@xxxxxxxxxx
To: rbk5287@xxxxxxxxxxxxx, dmitry.nadezhin@xxxxxxxxxx
Cc: STDS-1788@xxxxxxxxxxxxxxxxx
Sent: Tuesday, December 2, 2014 9:43:58 AM GMT +04:00 Abu Dhabi / Muscat
Subject: Re: Relation between P1788 and P1788.1

Dima, Baker

On 2 Dec 2014, at 03:59, Ralph Baker Kearfott <rbk5287@xxxxxxxxxxxxx> wrote:

It is meant by its champion (Ned Nedialkov) to be C):
1788.1 is meant to be a proper subset of the 1788 set-based flavor.
A P1788.1-conforming program is meant to be 1788-conforming, but not
visa versa.  This will become clear when Ned posts the document, or from
the Project Authorization Request Ned has submitted.

1788 is meant to be a normative reference of 1788.1, but not
visa versa.

Which is easier to check: that it be a subset or that it be a flavor?

It cannot be a flavor, because it doesn't have the full complement of operations required in §9 of the full standard.

I would call it a subset, but one must understand what that means. In my view the *only* required (mathematical) relation between the standards is:
(*)     A 1788.1-conforming program shall be 1788-conforming.

Maybe judge 1788.1 on how easy it is to understand, implement and teach: these are crucial, but subjective. But if (*), which is objective, is false then there is an error somewhere.

Of course there is also an administrative relation in that IEEE regards them as related.

John P