Re: Please listen to Ulrich here...
On 2013-08-07 02:24:28 -0700, Dan Zuras Intervals wrote:
> When Ulrich talks about problems with exact dot product he has some
> experience in the matter. More than the rest of us put together.
> If we are about to have a standard that admits the possibility of
> a member that CANNOT do an exact EDP, all our work will be wasted.
> Please consider rewording this document in Ulrich's favor this time.
> He is not just blowing smoke. It is hard but it is necessary. Or,
> at least if we make it so.
I've already said that in the list in the past, but what Ulrich didn't
consider is multiple precision. He didn't show that multiple precision
was not sufficient, and there's no reason why it wouldn't be.
Then I agree with Paul. Anyone may have his own wishes for his own
applications. For instance, why not some some well-known floating-point
primitives, such as a TwoSum operation (you may remember that some
variant of this was considered in the revision of IEEE 754 and it was
removed just because of under-specification of exceptions[*], something
that never occurs in codes seen in practice). And FYI, evaluation of
math function is in every system library (and some of these functions
are even required by P1788), so that such primitives may be useful to
many users. Much more than CA/EDP.
[*] http://grouper.ieee.org/groups/754/email/msg02209.html
Now, I wouldn't like P1788 to deviate from interval arithmetic; let's
recall that the PAR title is "Standard for Interval Arithmetic" and
its scope is also exclusively on interval arithmetic.
If Ulrich wants CA to be implemented (in hardware), he can talk to a
processor vendor and try to convince him. He doesn't need a standard
for that. And if it is so useful, then other vendors will implement
it as well...
Also note that CA takes resources. Some processors, like the Itanium,
do not even have a division instruction.
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)