P0623r0
Final C++17 Parallel Algorithms Fixes

1. Overview

A few fixes for the C++17 parallel algorithms which should have been included in other papers slipped through the cracks.

This paper addresses NB comments US 161, US 162 and US 184.

2. Proposed Changes

The following changes are relative to the latest working paper as of 03-02-2017 ([N4640]) with the changes from [P0467r2] and [P0574r1] applied.

The � character is used to denote a placeholder section number.

2.1. Add Type Requirements From [P0574r1] To [P0452r1]'s New transform_reduce Signature

Modify the requirements for the new signatures of transform_reduce added by [P0452r1] to 26.8.4 [transform.reduce]:

� Requires:

• T shall be MoveConstructible (Table 23).
• All of binary_op1(init, binary_op2(*first1, *first2)), binary_op1(binary_op2(*first1, *first2), init), binary_op1(init, init), and binary_op1(binary_op2(*first1, *first2), binary_op2(*first1, *first2)) shall be convertible to T.

• Neither binary_op1 nor binary_op2 shall invalidate subranges, or modify elements in the ranges [first1, last1)] and [first2, first2 + (last1 - first1))].

Direction to the Editor: Add a footnote to the use of fully-closed ranges in the change above that says The use of fully closed ranges is intentional .

2.2. Forbid The result == first Case For Parallel adjacent_difference

Modify the remark in 26.8.11 paragraph 3 regarding whether the input and output ranges given to adjacent_difference can overlap:

3 Remarks: For the overloads with no ExecutionPolicy, result may be equal to first. For the overloads with an ExecutionPolicy, the ranges [first, last) and [result, result + (last - first)) shall not overlap.