(j3.2006) (SC22WG5.6008) Additional forthcoming unauthorised editorial/technical changes

Malcolm Cohen malcolm
Fri Dec 15 03:38:47 EST 2017


Hi folks,

 

While I was looking at the list of IEEE functions we require support for in
p1 of 17.9 (page 453) I noticed some more problems.

 

1.	There is no such function as "copysign".  It's spelled "copySign".
2.	There is no such function as "scalb".  There is a "scaleB", but
unlike IEEE_SCALB that is required to have the second argument of the same
type and type parameters as the result of "logB" (we chose REAL for the
result of IEEE_LOGB and INTEGER for the argument of IEEE_SCALB, so we fail
the consistency requirement).  Also unlike IEEE_SCALB it is rounded
according to the "applicable" rounding mode, so should not always produce
INF with overflow.  Perhaps we weaselled "scaleB" by saying it could be done
by a code sequence, but anyway we have nothing suitable to require, so we
have to delete it from the list.
3.	There is no such function as "logb".  It's spelled "logB".
4.	There is no such function as "nextafter" at all.  The 754 committee
did not include this function, and they have no function that matches
IEEE_NEXT_AFTER.  They did include nextUp and nextDown as improved
replacements for the functionality (and we have those).  So we need to
delete "nextafter" from the list, the only question is whether we want to
require nextUp and nextDown (viz IEEE_NEXT_UP and IEEE_NEXT_DOWN) instead.
5.	There is no such function as "unordered".  It's spelled
"compareQuietUnordered".

 

So that is three renamings and two deletions.  As the text is undeniably
wrong as is, these would seem to be uncontroversial.

 

And there is one open question: Should we require support for IEEE_NEXT_UP
and IEEE_NEXT_DOWN?  In the absence of direction I will assume the answer is
"No", i.e. I will not insert anything to require such support.

 

Cheers,

-- 

..............Malcolm Cohen, NAG Oxford/Tokyo.

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.j3-fortran.org/pipermail/j3/attachments/20171215/09afdf7c/attachment.html>



More information about the J3 mailing list