To: J3 07-333r1
From: Dan Nagle and John Reid
Subject: UTI 136 (p437): Re IEEE_SUPPORT_DENORMAL
Date: 2007 November 12
References: J3/07-007r3, ANSI/IEEE Std 754-1985
Discussion
The editor remarks that there can be no operations defined by
intrinsic functions. While this is true for Fortran operations,
it is not so for the operations of the IEEE standard. Section 5
of the IEEE standard states
"All conforming implementations of this standard shall provide
operations to add, subtract, multiply, divide, extract the square
root, find the remainder, round to integer in floating-point format,
convert between different floating-point formats, convert between
floating-point and integer formats, convert binary <-> decimal, and
compare."
Edits are provided to make this clear.
The editor also says that it would help to know exactly which
specifications are involved. We agree and propose the addition of a
note.
Edits to 07-007r3
[437:4-6,8-10,12-14] In 14.9 IEEE arithmetic, delete the second
sentence in each of paragraphs 4-6.
[437:14+] Insert a new paragraph:
"For each kind of real for which IEEE NaNs, IEEE infinities, or IEEE
denormals are supported, their behavior for the unary and binary
operations specified in Section 5 of the IEEE International Standard,
including those provided by intrinsic functions and by functions in
intrinsic modules, shall conform to the specifications in the IEEE
International Standard".
[437:14+] In 14.9 IEEE arithmetic, after paragraph 6, add
"NOTE 14.7a
The operations specified in Section 5 of the IEEE International
Standard are add, subtract, multiply, divide, extract the square root,
find the remainder, round to integer in floating-point format, convert
between different floating-point formats, convert between
floating-point and integer formats, convert binary <-> decimal, and
compare."
Note to editor: The 1985 IEEE standard uses "<->", not "between... and..."
or a double-arrow symbol.