To: J3                                                     J3/18-190
From: Malcolm Cohen
Subject: Country comment GB-051
Date: 2018-June-12

1. Introduction

The fourth paragraph of subclause 17.1 "Overview of IEEE arithmetic
support" contains unnecessary verbiage and a mistake.
These should be fixed/deleted.

2. Response to GB-051

Accepted with modification.

3. Discussion of the comment

Unnecessary verbiage is as follows:
(a) specifying "and return true" from various functions

Identified error is as follows:
(b) rounding poorly specified.

4. Additional changes

(A)
Incorrect use of "may" (should be "might" or "can").

(B)
"has access to X" is clumsier than "X is accessible".

(C)
"of IEEE_FEATURES" is not needed when it immediately follows
the discussion about things being influenced from IEEE_FEATURES.


5. Edits to N2146

Page 443, 17.1 "Overview of IEEE arithmetic support", paragraph 4,

Change "may be influenced" to "can be influenced".

Change "If a scoping unit has access to IEEE_DATATYPE of IEEE_FEATURES,"
to "If IEEE_DATATYPE is accessible,".

Delete "and return true from IEEE_SUPPORT_DATATYPE(X) (17.11.48)".
Delete "and return true from the corresponding inquiry function".

After "In the case of IEEE_ROUNDING,"
change "return true for" to "support".

This will make the whole paragraph read:
  "Additional ISO/IEC/IEEE 60559:2011 facilities are available from the
   module IEEE_ARITHMETIC.  The extent of support can be influenced by the
   accessibility of the named constants of the module IEEE_FEATURES.  If
   IEEE_DATATYPE is accessible, within the scoping unit the processor
   shall support IEEE arithmetic for at least one kind of real. Similarly,
   if IEEE_DENORMAL, IEEE_DIVIDE, IEEE_INF, IEEE_NAN, IEEE_ROUNDING,
   IEEE_SQRT, or IEEE_SUBNORMAL is accessible, within the scoping unit the
   processor shall support the feature for at least one kind of real. In
   the case of IEEE_ROUNDING, it shall support the rounding modes
   IEEE_NEAREST, IEEE_TO_ZERO, IEEE_UP, and IEEE_DOWN; support for
   IEEE_AWAY is also required if there is at least one kind of real X for
   which IEEE_SUPPORT_DATATYPE (X) is true and RADIX (X) is equal to ten.
   Note that the effect of IEEE_DENORMAL is the same as that of
   IEEE_SUBNORMAL."

===END===