Empty base optimization

From cppreference.com
< cpplrm; | language

Allows the size of an empty base subobject to be zero.

Explanation

The size of any object or member subobject (unless [[no_unique_address]] -- see below) (since C++20) is required to be at least 1 even if the type is an empty class type (that is, a class or struct that has no non-static data members), in order to be able to guarantee that the addresses of distinct objects of the same type are always distinct.

However, base class subobjects are not so constrained, and can be completely optimized out from the object layout:

#include <cassert>

struct Base {}; // empty class

struct Derived1 : Base {
    int i;
};

int main()
{
    // the size of any object of empty class type is at least 1
    assert(sizeof(Base) >= 1);

    // empty base optimization applies
    assert(sizeof(Derived1) == sizeof(int));
}

Empty base optimization is prohibited if one of the empty base classes is also the type or the base of the type of the first non-static data member, since the two base subobjects of the same type are required to have different addresses within the object representation of the most derived type.

A typical example of such situation is the naive implementation of std::reverse_iterator (derived from the empty base std::iterator), which holds the underlying iterator (also derived from std::iterator) as its first non-static data member.

#include <cassert>

struct Base {}; // empty class

struct Derived1 : Base {
    int i;
};

struct Derived2 : Base {
    Base c; // Base, occupies 1 byte, followed by padding for i
    int i;
};

struct Derived3 : Base {
    Derived1 c; // derived from Base, occupies sizeof(int) bytes
    int i;
};

int main()
{
    // empty base optimization does not apply,
    // base occupies 1 byte, Base member occupies 1 byte
    // followed by 2 bytes of padding to satisfy int alignment requirements
    assert(sizeof(Derived2) == 2*sizeof(int));

    // empty base optimization does not apply,
    // base takes up at least 1 byte plus the padding
    // to satisfy alignment requirement of the first member (whose
    // alignment is the same as int)
    assert(sizeof(Derived3) == 3*sizeof(int));
}

Empty base optimization is required for StandardLayoutTypes in order to maintain the requirement that the pointer to a standard-layout object, converted using reinterpret_cast, points to its initial member, which is why one of the requirements for a standard layout type is "has no base classes with non-static data members and has no base classes of the same type as the first non-static data member".

(since C++11)

The empty member subobjects are permitted to be optimized out just like the empty bases if they use the attribute [[no_unique_address]]. Taking the address of such member results in an address that may equal the address of some other member of the same object.

#include <cassert>

struct Empty {}; // empty class

struct X {
    int i;
    [[no_unique_address]] Empty e;
};

int main()
{
    // the size of any object of empty class type is at least 1
    assert(sizeof(Empty) >= 1);

    // empty member optimized out
    assert(sizeof(X) == sizeof(int));
}
(since C++20)

Notes

Empty base optimization is commonly used by allocator-aware standard library classes (std::vector, std::function, std::shared_ptr, etc) to avoid occupying any additional storage for its allocator member if the allocator is stateless. This is achieved by storing one of the required data members (e.g., begin, end, or capacity pointer for the vector) in an equivalent of boost::compressed_pair with the allocator.


References

  • C++11 standard (ISO/IEC 14882:2011):
  • 5.10 Equality operators [expr.eq](p: 2)
  • 5.3.3 Sizeof [expr.sizeof](p: 2)
  • 9 Classes [class](p: 4,7)
  • 9.2 Class members [class.mem](p: 20)
  • C++98 standard (ISO/IEC 14882:1998):
  • 5.10 Equality operators [expr.eq](p: 2)
  • 5.3.3 Sizeof [expr.sizeof](p: 2)
  • 9 Classes [class](p: 3)