J3/01-130r1

Date:    12 Apr. 2001
To:      J3
From:    JOR/Stan Whitlock
Subject: Issue 212 - the content and placement of note 15.6

In its text issue 212 of J3/01-007 refers to note 15.15, but its
placement and the rest of its text makes it clear that it actually
refers to note 15.6, and I will use that reference in subsequent
discussion.

Issue 212 notes that neither the content nor the placement of note
15.6 makes the subject of the note clear, that section 15.8.2 may
not be the appropriate place for this note, and the note's content
is not well justified.

An examination of section 15.8.2 suggests that the content of note
15.6 can apply to all the elemental functions of 15.8.2, but the
interpretation of IEEE_LOG, IEEE_NEXT_AFTER, IEEE_REM, IEEE_RINT,
IEEE_SCALB, and IEEE_UNORDERED benefit the most from its comments.

There are only two sections where this note might be justifiably
be placed, section 15.8.2, which lists only the functions that
are the subject of note 15.6, or section 15.9, which discusses
the semantics of all the procedures.  As 15.8 is otherwise a
simple tabulation of procedures, and this note refers to
semantics, my preference is to place the contents of this note
in section 15.9.

The justification of this note appears to be similar to that of
15.7 and would appear to benefit from similar wording. The
committee appears to desire implementations with the semantics
of IEEE arithmetic, but allow the efficient implementation on
processors for a system which does not have complete hardware
support for IEEE arithmetic.

Edits:

Delete Note 15.6 near the end of 15.8.2, just before Issue 212,
on page 369.  Renumber subsequent notes.

Delete Issue 212 at the end of 15.8.2 on page 369.

Add the following text between what is now Note 15.7 and the
start of section 15.9.1 on page 370 [370:16+].

For the elemental functions of IEEE_ARITHMETIC, as tabulated in
section 15.8.2, if X or Y has a value that
is an infinity or a NaN, the result shall be consistent with
the general rules in sections 6.1 and 6.2 of the IEEE standard.
For example, the result for an infinity shall be constructed
as the limiting case of the result with a value of arbitrarily
large magnitude, when such a limit exists.