ISO/IEC JTC1 SC22 WG21
P0840R2
Richard Smith
richard@metafoo.co.uk
2018-03-12

Language support for empty objects

For motivation, see P0840R0.

Proposal

We propose the addition of an attribute, [[no_unique_address]] to indicates that a unique address is not required for an non-static data member of a class. A non-static data member with this attribute may share its address (or its tail padding) with another object, if it could when used as a base class.

FAQ

Q: Does this allow multiple subobjects of the same type to have the same address? No. Just like with base classes, if two members are of the same type or have subobjects of the same type, the common subobjects will be given distinct addresses.
Q: Can a standard library switch from EBO to this attribute without an ABI break? The intent is that an ABI can specify the same layout rule for a member with the attribute as it does for a base class. In an ABI that makes that choice, yes.
Q: Does this allow reuse of tail padding? (Eg, three bytes at the end of struct A { int n; char c; };) The general rule is that the layout is just like for a base class. Tail padding reuse is permitted for base classes, so it's also permitted for members with the attribute.
Q: When is a type using this attribute considered "empty"? (As visible via std::is_empty) This is up to the ABI, so it will be implementation-defined whether a type is considered empty if it has no data members (transitively) other than those marked by the attribute.
Q: Does the attribute affect whether a type is standard-layout? No.
Q: Does the attribute affect the "common initial sequence" rule? Yes. For two structs to be considered to have a common initial sequence, their initial sequences of common members must make consistent use of the attribute.
Q: Suppose I have members a, b, c (in that order, with the same access). Today we guarantee that &a < &b < &c. What happens if b has the attribute? Two cases:
  1. If the type of b is empty, then there is no guarantee about the address of b (other than that it is somewhere within the containing object).
  2. If the type of b is nonempty, then we still guarantee that &a < &b < &c.
Q: Applying a [[no_unique_address]] attribute to a function, constant variable, etc, could be used to permit ICF ("Identical Code Folding" / "Identical Constant Folding" / "Identical COMDAT Folding"). Is this proposed? Not as part of this proposal.

Wording

Add a new paragraph before [intro.object] (6.6.2) paragraph 7:

A potentially-overlapping subobject is either:

Change in [intro.object] (6.6.2) paragraph 7:

An object has nonzero size if it Otherwise, if the object is a base class subobject of a standard-layout class type with no non-static data members, it has zero size. Otherwise, the circumstances under which the object has zero size are implementation-defined. Unless it is a bit-field (12.2.4), a most derived an object shall have a with nonzero size and shall occupy one or more bytes of storage, including every byte that is occupied in full or in part by any of its subobjects. Base class subobjects may have zero size. An object of trivially copyable or standard-layout type (6.9) shall occupy contiguous bytes of storage.

Change in [intro.object] (6.6.2) paragraph 8:

Unless an object is a bit-field or a base class subobject of zero size, the address of that object is the address of the first byte it occupies. Two objects a and b with overlapping lifetimes that are not bit-fields may have the same address if one is nested within the other, or if at least one is a base class subobject of zero size and they are of different types; otherwise, they have distinct addresses and shall occupy disjoint bytes of storage. [Footnote] [Example] The address of a non-bit-field subobject of zero size is that of a byte of storage occupied by the complete object of the subobject.

Change in [basic.life] (6.6.3) bullet 8.4:

neither the original object nor the new object is a potentially-overlapping subobject (6.6.2 [intro.object]) was a most derived object (6.6.2) of type T and the new object is a most derived object of type T (that is, they are not base class subobjects).

Change in [basic.types] (6.7) paragraph 2:

For any object (other than a base-class potentially-overlapping subobject) of trivially copyable type […]

Change in [basic.types] (6.7) paragraph 3:

For any trivially copyable type T, if two pointers to T point to distinct T objects obj1 and obj2, where neither obj1 nor obj2 is a base-class potentially-overlapping subobject, if the underlying bytes (6.6.1) making up obj1 are copied into obj2, obj2 shall subsequently hold the same value as obj1. […]

Change in [expr.sizeof] (8.5.2.3) paragraph 1:

The sizeof operator yields the number of bytes occupied by a non-potentially-overlapping object of the type in the object representation of its operand. […]

Change in [expr.sizeof] (8.5.2.3) paragraph 2:

When applied to a reference or a reference type, the result is the size of the referenced type. When applied to a class, the result is the number of bytes in an object of that class including any padding required for placing objects of that type in an array. The size of a most derived class shall be greater than zero (6.6.2). The result of applying sizeof to a base class potentially-overlapping subobject is the size of the base class type, not the size of the subobject. [ Footnote: The actual size of a base class potentially-overlapping subobject may be less than the result of applying sizeof to the subobject, due to virtual base classes and less strict padding requirements on base class potentially-overlapping subobjects. ] When applied to an array, the result is the total number of bytes in the array. This implies that the size of an array of n elements is n times the size of an element.

Change in [expr.rel] (8.5.9) paragraph 3:

Comparing unequal pointers to objects is defined as follows:

Add a new subclause [dcl.attr.nouniqueaddr] after [dcl.attr.noreturn]:

10.6.9 No unique address attribute [dcl.attr.nouniqueaddr]

The attribute-token no_unique_address specifies that a non-static data member is a potentially-overlapping subobject (6.6.2 [intro.object]). It shall appear at most once in each attribute-list and no attribute-argument-clause shall be present. The attribute may appertain to a non-static data member other than a bit-field.

[ Note: The non-static data member can share the address of another non-static data member or that of a base class, and any padding that would normally be inserted at the end of the object can be reused as storage for other members. — end note ] [ Example:

template<typename Key, typename Value,
         typename Hash, typename Pred, typename Allocator>
class hash_map {
  [[no_unique_address]] Hash hasher;
  [[no_unique_address]] Pred pred;
  [[no_unique_address]] Allocator alloc;
  Bucket *buckets;
  // ...
public:
  // ...
};
Here, hasher, pred, and alloc could have the same address as buckets if their respective types are all empty. — end example ]

Change in [class] (12) paragraph 4 and split into two paragraphs:

[ Note: Complete objects and member subobjects of class type shall have nonzero size. [ Footnote: Base class subobjects and members declared with the no_unique_address attribute ([dcl.attr.nouniqueaddr]) are not so constrained. ]

[ Note: ... ]

Change in [class] (12) paragraph 7:

[…] […] [Note: M(X) is the set of the types of all non-base-class subobjects that are guaranteed may be at a zero offset in a standard-layout class to be at a zero offset in X. —end note] […]

Change in [class.mem] (12.2) paragraph 21:

The common initial sequence of two standard-layout struct (Clause 12) types is the longest sequence of non-static data members and bit-fields in declaration order, starting with the first such entity in each of the structs, such that corresponding entities have layout-compatible types, either both entities are declared with the no_unique_address attribute ([dcl.attr.nouniqueaddr]) or neither is, and either neither entity is a bit-field or both entities are bit-fields with the same width or neither is a bit-field. [ Example ]

Change in [meta.unary.prop] (23.15.4.3) Table 42 ("Type property predicates"):

TemplateConditionPredicate
template <class T> struct is_empty; T is a class type, but not a union type, with no non-static data members other than bit-fields of length 0 subobjects of zero size, no virtual member functions, no virtual base classes, and no base class B for which is_empty_v<B> is false. If T is a non-union class type, T shall be a complete type.