Re: Motion P1788/M0009.01_ExactDotProduct YES

[Ulrich Kulisch <Ulrich.Kulisch@xxxxxxxxxxx>]

Nate Hayes schrieb:
> YES
>
> Nate Hayes
>
> P.S. I would also vote YES even if Section 3 was dropped, per Arnold's 
> comments.
Nate:

I responded on Anold's comments. You may have missed it. So I repeat 
what I wrote:

If an "exact" dot product is available, Section 3 of the proposal is a 
trivial consequence and it could be dropped. It was added in order to 
demonstrate that in a pure interval computation (which is free of 
exceptions) also computing the interval dot product is free of 
exceptions. All the exceptions that had to be dealt with in section 
"2.2.1 Convert" of the proposal do not occur in a pure interval 
computation. A pure interval computation only makes use of non 
exceptional flaoting-point numbers. In contrast to this an "exact" dot 
product, if it is based on IEEE 754, has to consider the IEEE 754 
exceptions also.

Best wishes
Ulrich