To:         J3 Members                                       J3/16-133
From:       Van Snyder
Subject:    Clause 13 questions without answers or edits
Reference:  16-007
Date: 2016 January 25

[349:27 18.8.37p7] says "BGE(0,-1) has the value false."  13.3.1p3 says
the interpretation of negative integers as sequences of bits is
processor dependent.  Is the assertion correct?

[350:1 18.8.38p7] says "BGT(0,-1) has the value false."  13.3.1p3 says
the interpretation of negative integers as sequences of bits is
processor dependent.  Is the assertion correct?

[350:21 18.8.40p7] says "BLE(0,-1) has the value false."  13.3.1p3 says
the interpretation of negative integers as sequences of bits is
processor dependent.  Is the assertion correct?

[350:33 18.8.41p7] says "BLT(0,-1) has the value false."  13.3.1p3 says
the interpretation of negative integers as sequences of bits is
processor dependent.  Is the assertion correct?

[375:19 13.8.96p3 SHIFT] The requirement that |SHIFT| <= SIZE is not
needed.  Can it be removed?

[376:1-14 13.8.89-99] IS_IOSTAT_END and IS_IOSTAT_EOR were added because
IBM might provide more than one value to signify these conditions.  But
13.9.2.14-15 are scalars.  Which way is it?

[13.8.72 13.8.119 13.8.125 13.8.139 13.8.145 13.8.153 13.8.176] Do we
need to take care with the statement that the result values depend upon
the values of ARRAY for which corresponding elements of MASK are true?
What if MASK is a scalar?  Some of these, e.g., PRODUCT, REDUCE, SUM,
could be described using PACK, as are, e.g., MAXVAL.

[401:2+1-4 NOTE 13.21] NOTE 13.21 leaves the impression that results
will be repeatable if ORDERED is present with the value true.  This is
not true unless we say so in 13.8.153p5 (because floating-point
arithmetic is not necessarily associative, and at least some modern
processors do dynamic reordering).  Requiring the results to be
repeatable if ORDERED is present might impose a significant burden
and/or performance penalty on some processors.

[417:4,7 13.9.2.2-3] Did we intend to restrict a processor to provide
only one kind of type integer and logical for which atomic operations
work?

[419:16-23 13.9.14-15] IS_IOSTAT_END and IS_IOSTAT_EOR were added because
IBM might provide more than one value to signify these conditions.  But
13.9.2.14-15 are scalars.  Which way is it?