std::atomic
<atomic>
struct atomic;
struct atomic<U*>;
<memory>
struct atomic<std::shared_ptr <U>>;
struct atomic<std::weak_ptr <U>>;
<stdatomic.h>
Each instantiation and full specialization of the std::atomic
template defines an atomic type. If one thread writes to an atomic object while another thread reads from it, the behavior is well-defined (see memory model for details on data races).
In addition, accesses to atomic objects may establish inter-thread synchronization and order non-atomic memory accesses as specified by std::memory_order .
std::atomic
is neither copyable nor movable.
The compatibility macro _Atomic
is provided in <stdatomic.h> such that _Atomic(T)
is identical to std::atomic<T>
while both are well-formed.
It is unspecified whether any declaration in namespace std
is available when <stdatomic.h> is included.
Contents
[edit] Specializations
[edit] Primary template
The primary std::atomic
template may be instantiated with any TriviallyCopyable type T
satisfying both CopyConstructible and CopyAssignable. The program is ill-formed if any of following values is false:
- std::is_trivially_copyable <T>::value
- std::is_copy_constructible <T>::value
- std::is_move_constructible <T>::value
- std::is_copy_assignable <T>::value
- std::is_move_assignable <T>::value
- std::is_same <T, typename std::remove_cv <T>::type>::value
struct Counters { int a; int b; }; // user-defined trivially-copyable type std::atomic<Counters> cnt; // specialization for the user-defined type
std::atomic<bool> uses the primary template. It is guaranteed to be a standard layout struct and has a trivial destructor.
[edit] Partial specializations
The standard library provides partial specializations of the std::atomic
template for the following types with additional properties that the primary template does not have:
std::atomic<U*>
for all pointer types. These specializations have standard layout, trivial default constructors,(until C++20) and trivial destructors. Besides the operations provided for all atomic types, these specializations additionally support atomic arithmetic operations appropriate to pointer types, such as fetch_add
, fetch_sub
.See std::atomic<std::shared_ptr> and std::atomic<std::weak_ptr> for details.
(since C++20)[edit] Specializations for integral types
When instantiated with one of the following integral types, std::atomic
provides additional atomic operations appropriate to integral types such as fetch_add
, fetch_sub
, fetch_and
, fetch_or
, fetch_xor
:
- The character types char, char8_t(since C++20), char16_t, char32_t, and wchar_t;
- The standard signed integer types: signed char, short, int, long, and long long;
- The standard unsigned integer types: unsigned char, unsigned short, unsigned int, unsigned long, and unsigned long long;
- Any additional integral types needed by the typedefs in the header <cstdint>.
Additionally, the resulting std::atomic<Integral>
specialization has standard layout, a trivial default constructor,(until C++20) and a trivial destructor. Signed integer arithmetic is defined to use two's complement; there are no undefined results.
Specializations for floating-point types
When instantiated with one of the cv-unqualified floating-point types (float, double, long double and cv-unqualified extended floating-point types (since C++23)), std::atomic
provides additional atomic operations appropriate to floating-point types such as fetch_add
and fetch_sub
.
Additionally, the resulting std::atomic<Floating>
specialization has standard layout and a trivial destructor.
No operations result in undefined behavior even if the result is not representable in the floating-point type. The floating-point environment in effect may be different from the calling thread's floating-point environment.
(since C++20)[edit] Member types
value_type
T
(regardless of whether specialized or not)
difference_type
[1]
value_type
(only for atomic<Integral>
and atomic<Floating>
(since C++20) specializations)
std::ptrdiff_t (only for std::atomic<U*>
specializations)
- ↑
difference_type
is not defined in the primarystd::atomic
template or in the partial specializations for std::shared_ptr and std::weak_ptr .
[edit] Member functions
(public member function) [edit]
(public member function) [edit]
(public member function) [edit]
(public member function) [edit]
Constants
[edit] Specialized member functions
Specialized for integral, floating-point(since C++20) and pointer types
(public member function) [edit]
(public member function) [edit]
Specialized for integral and pointer types only
(public member function) [edit]
(public member function) [edit]
Specialized for integral types only
(public member function) [edit]
(public member function) [edit]
(public member function) [edit]
[edit] Type aliases
Type aliases are provided for bool and all integral types listed above, as follows:
Aliases for all std::atomic<Integral>
Aliases for special-purpose types
(typedef) [edit]
(typedef) [edit]
std::atomic_intN_t
, std::atomic_uintN_t
, std::atomic_intptr_t
, and std::atomic_uintptr_t
are defined if and only if std::intN_t
, std::uintN_t
, std::intptr_t , and std::uintptr_t are defined, respectively.
std::atomic_signed_lock_free
and std::atomic_unsigned_lock_free
are optional in freestanding implementations.
[edit] Notes
There are non-member function template equivalents for all member functions of std::atomic
. Those non-member functions may be additionally overloaded for types that are not specializations of std::atomic
, but are able to guarantee atomicity. The only such type in the standard library is std::shared_ptr <U>.
_Atomic
is a keyword and used to provide atomic types in C.
Implementations are recommended to ensure that the representation of _Atomic(T)
in C is same as that of std::atomic<T>
in C++ for every possible type T
. The mechanisms used to ensure atomicity and memory ordering should be compatible.
On GCC and Clang, some of the functionality described here requires linking against -latomic
.
Feature-test macro | Value | Std | Feature |
---|---|---|---|
__cpp_lib_atomic_ref |
201806L |
(C++20) | std::atomic_ref
|
__cpp_lib_constexpr_atomic |
202411L |
(C++26) | constexpr std::atomic and std::atomic_ref
|
[edit] Example
#include <atomic> #include <iostream> #include <thread> #include <vector> std::atomic_int acnt; int cnt; void f() { for (auto n{10000}; n; --n) { ++acnt; ++cnt; // Note: for this example, relaxed memory order is sufficient, // e.g. acnt.fetch_add(1, std::memory_order_relaxed); } } int main() { { std::vector <std::jthread > pool; for (int n = 0; n < 10; ++n) pool.emplace_back(f); } std::cout << "The atomic counter is " << acnt << '\n' << "The non-atomic counter is " << cnt << '\n'; }
Possible output:
The atomic counter is 100000 The non-atomic counter is 69696
[edit] Defect reports
The following behavior-changing defect reports were applied retroactively to previously published C++ standards.
DR | Applied to | Behavior as published | Correct behavior |
---|---|---|---|
LWG 2441 | C++11 | typedefs for atomic versions of optional fixed width integer types were missing |
added |
LWG 3012 | C++11 | std::atomic<T> was permitted for any T that is trivially copyable but not copyable |
such specializations are forbidden |
LWG 3949 | C++17 | the wording requiring std::atomic<bool> to have a trivial destructor was accidently dropped in C++17 |
added back |
LWG 4069 (P3323R1) |
C++11 | support for cv-qualified T was questionable
|
disallow T being cv-qualified
|
P0558R1 | C++11 | template argument deduction for some functions for atomic types might accidently fail; invalid pointer operations were provided |
specification was substantially rewritten: member typedefs value_type and difference_type are added
|