Re: Motion P1788/M0004.01_Keep_to_754 NO
It also limits consideration and discussion to:
"an environment that conforms to 754-2008"
I don't believe its good idea to limit discussion this way. It prevents any
consideration of non-standard extentions or simplifications of 754 to be
used in the interval arithmetic, such as 0*X=0 for any interval X (even when
X is unbounded).
In my view, the selling point of intervals is not 754-compliance, but
reliable computing. So I think this motion puts the emphasis in the wrong
place and may lead to missed opportunities.
Nate
----- Original Message -----
From: "Joel C. Salomon" <joelcsalomon@xxxxxxxxx>
To: <STDS-1788@xxxxxxxxxxxxxxxxx>
Sent: Friday, May 08, 2009 10:02 AM
Subject: Re: Motion P1788/M0004.01_Keep_to_754 NO
Nate Hayes wrote:
> I would vote "YES" for a motion to adopt the electrical engineering terms
> and concepts of IEEE 754 in our discussions. This includes denormalized
> numbers, NaNs, binary fraction, exponent, rounding modes, etc. However,
> any
> motion to encourage "compliance" or "independence" from IEEE 754 is, in my
> view, putting the cart before the horse. So I don't see the current motion
> (or Arnold's alternative) is necessary or useful
Rereading the motion, it seems that’s exactly what’s under
consideration: using the terms & concepts from 754-2008 to discuss the
semantics of 1788.
—Joel Salomon
>> ===Motion P1788/M0004.01_Keep_to_754===
>> Proposer: John Pryce
>> Seconder: Dan Zuras
>>
>> ===Motion text===
>>
>> In our work on 1788 we will limit the scope of our consideration
>> to those systems which present an environment that conforms to
>> 754-2008. This limitation is done without prejudice and for the
>> purpose of K.I.S.S. (Keeping It Simple Stupid) to simplify our
>> discussions and specifications.
>>
>> ===Rationale===
>>
>> This motion is offered in the spirit of simplifying our task. We
>> are all familiar with the characteristics of 754 environments.
>> Confining our thoughts and discussions to such systems will give
>> us common ground for understanding. And 1788 specifications
>> couched in terms of that understanding can be made simpler than
>> without it. Issues of things like the role and use of NaNs and
>> infinity, signed zeros, transcendental functions, support of
>> basic formats, and the like are all details that are not part of
>> this motion. But all of those details are made easier to discuss
>> in the future if this motion passes.
>>
>> The issue of whether or not a non-754-conforming system may yet
>> conform to 1788 is also not part of this motion. That will
>> depend on all the other normative statements we make in the
>> future. We will almost certainly not use all the specifications
>> of 754 to specify 1788. But it is expected that most such
>> systems will not conform. The Cell chip may be an exception.
>>
>> Passing this motion saves us time and misunderstanding in the
>> future. That's what this motion is about.
>> ssing this motion saves us time and misunderstanding in the future.
>> That's what this motion is about.