Re: Comments on the IEEE-SA draft P1788/D9.5
On Fri, 09 Jan 2015 10:23:01 -0600, Baker posted Vincent's comments.
I agree that Definition 3.2.55 of "roundoff unit" should simply be dropped,
because (a) as defined, it is absolute and only applies to the value 1.0
(which is in fact what is used in the two examples), and (b) I think the
meaning is clear in the context of those examples.
(In actual use, "machine epsilon" is multiplied by the weight of the unit
bit, whether hidden or not.)
Michel.
P.S. Just one day left to post those official Sponsor Ballot comments!
---Sent: 2015-01-09 16:46:07 UTC