Re: Please listen to Ulrich here...
Baker,
Thanks for that clarification.
Everyone:
When P1788 started I was working on high accuracy arithmetic with people like Paul Zimmermann, and
Norbert Muller. I've since changed direction a bit (!!!), and have had less time to devote to P1788
than I would have liked, for which I apologize.
I think I have met most of you -- either at Arith, RNC or other workshops/conferences -- so I'd
prefer it if people didn't think of me as an outsider. Siegfried: we met at Dagstuhl.
Instead, the change of direction brings with it different perspectives, which I hope could be of
use to you all. I obviously don't speak for ARM, but their business model involves selling lots
of low power devices.
On the matter of HBP and the opportunities for interval arithemtic: it's not immediately clear
that we'll be using much by way of exact/interval arithmetic in our current work, but there are
certainly numeric instabilities inherent in almost all of the models in computational neuroscience.
I'm more than happy to discuss these issues further, but I suggest we do so privately, so as not
divert too much from our core agenda which is to nail down a precise and usable standard for
interval arithmetic.
Thanks,
David Lester
On 8 Aug 2013, at 14:26, Ralph Baker Kearfott wrote:
> David,
>
> On 08/08/2013 05:41 AM, David Lester wrote:
>
>> So my question is: is our proposed chip P1788 compliant? Provided that we do everything in software, I cannot see why it should not be. Am I missing something?
>>
>>
>
> We currently have no specification that requires anything be done in hardware. Indeed,
> nor does 754-2008 have any such mandate. Thus, your proposed chip would be neither
> compliant nor non-compliant. A software system built for it may be compliant or
> non-compliant, depending on the attributes of that system. (People, please correct
> me if I am wrong.)
>
> Baker
>