Question

It seems that unlike the member functions, you don’t get to specify which memory_order to use, so there’s probably some ‘default’ that ends up being used.

Answer 1

According to the C++ Standard (§29.7/35), the effect of calling operator++ of an atomic type A with non-atomic base type C, i.e. one of these:

C A::operator++() volatile noexcept;
C A::operator++() noexcept;

is the same as that of calling the member function fetch_add(1), and according to §29.5 (initial declarations), the latter is declared with a default argument:

C fetch_add(C, memory_order = memory_order_seq_cst) volatile noexcept;
C fetch_add(C, memory_order = memory_order_seq_cst) noexcept;

for integral types C, and:

C* fetch_add(ptrdiff_t, memory_order = memory_order_seq_cst) volatile noexcept;
C* fetch_add(ptrdiff_t, memory_order = memory_order_seq_cst) noexcept;

for address types. In other words, the default memory order used by operator++ is memory_order_seq_cst.

The Standard does not make any such statement about operator+=, although it seems natural to assume that what is true of operator++ is also true of operator+= in this context.

Also note there is the general rule for functions of atomic types defined by the Standard:

(§29.6.5/2) […] the free functions not ending in _explicit have the semantics of their corresponding _explicit with memory_order arguments of memory_order_seq_cst.