Re: IEEEP1788
On Tue, 9 Jun 2015 14:23:58 +0200 Vincent Lefèvre wrote:
> BTW, I would rather see *simpler* building blocks for *more common*
> algorithms, starting with the error on a sum of two floating-point
> numbers (which otherwise has to be computed with 5 or 6 operations
> in the general case). But in any case, this is more something for
> IEEE 754 than for IEEE 1788.
If the 754 WG is reactivated (see recent post by Baker), re-instating
the "distill" operations (which were dropped in 2006 or thereabouts)
would be the first thing I would push for.
> Well, correctly-rounded dot product should have been included in
> IEEE 754 seven years ago.
That too.
That would also be the place where complete arithmetic formats and
operations could be standardized as one of the flavours of arithmetic
(besides BFP and DFP, binary and decimal floating-point formats and
arithmetic).
Michel.
---Sent: 2015-06-09 17:07:46 UTC