P0019r3

• Align proposal with content of corresponding sections in N5131, 2016-07-15.
• Remove the one root wrapping constructor requirement from atomic_array_view.
• Other minor revisions responding to feedback from SG1 @ Oulu.

P0019r4

• wrapper constructor strengthen requires clause and omit throws clause
• Note types must be trivially copyable, as required for all atomics
• 2016-11-09 Issaquah SG1 decision: move to LEWG targeting Concurrency TS V2

P0019r5

• 2017-03-01 Kona LEWG review
  • Merge in P0440 Floating Point Atomic View because LEWG consensus to move P0020 Floating Point Atomic to C++20 IS
  • Rename from atomic_view and atomic_array_view; authors' selection atomic_ref<T> and atomic_ref<T[]>, other name suggested atomic_wrapper.
  • Remove constexpr qualification from default constructor because this qualification constrains implementations and does not add apparent value.
• Remove default constructor, copy constructor, and assignment operator for tighter alignment with atomic<T> and prevent empty references.
• Revise syntax to align with P0558r1, Resolving atomic<T> base class inconsistencies
• Recommend feature next macro

P0019r6 : for 2018-03-Jacksonville

• 2017-11-07 Albuquerque LEWG review
  • Settle on name atomic_ref
  • Split out atomic_ref<T[]> into a separate paper, apply editorial changes accordingly
  • Restore copy constructor; not assignment operator
  • add Throws: Nothing to constructor but do not add noexcept
  • Remove wrapping terminology
  • Address problem of CAS on atomic_ref<T> where T is a struct containing padding bits
  • With these revisions move to LWG

Motivation: Atomic Operations on a Single Non-atomic Object

An atomic reference is used to perform atomic operations on referenced non-atomic object. The intent is for atomic reference to provide the best-performing implementation of atomic operations for the non-atomic object type. All atomic operations performed through an atomic reference on a referenced non-atomic object are atomic with respect to any other atomic reference that references the same object, as defined by equality of pointers to that object. The intent is for atomic operations to directly update the referenced object. An atomic reference constructor may acquire a resource, such as a lock from a collection of address-sharded locks, to perform atomic operations. Such atomic reference objects are not lock-free and not address-free. When such a resource is necessary subsequent copy and move constructors and assignment operators may reduce overhead by copying or moving the previously acquired resource as opposed to re-acquiring that resource.

Introducing concurrency within legacy codes may require replacing operations on existing non-atomic objects with atomic operations such that the non-atomic object cannot be replaced with an atomic object.

An object may be heavily used non-atomically in well-defined phases of an application. Forcing such objects to be exclusively atomic would incur an unnecessary performance penalty.

Motivation: Atomic Operations on Members of a Very Large Array

High performance computing (HPC) applications use very large arrays. Computations with these arrays typically have distinct phases that allocate and initialize members of the array, update members of the array, and read members of the array. Parallel algorithms for initialization (e.g., zero fill) have non-conflicting access when assigning member values. Parallel algorithms for updates have conflicting access to members which must be guarded by atomic operations. Parallel algorithms with read-only access require best-performing streaming read access, random read access, vectorization, or other guaranteed non-conflicting HPC pattern.

An atomic array reference is used to perform atomic operations on the non-atomic members of the referenced array. The intent is for atomic array reference to provide the best-performing implementation of atomic operations for the members of the array.

Reference-ability Constraints

An object referenced by an atomic reference must satisfy, possibly architecuture specific, constraints. For example that the object is properly aligned in memory or does not reside in GPU register memory. We do not enumerate all potential constraints or specify behavior when these constraints are violated. It is quality-of-implementation to generate appropriate information when constraints are violated.

Concern with atomic<T> and padding bits in T

A concern has been discussed for atomic<T> where T is a class type that contains padding bits and how construction and compare_exchange operations are effected by the value of those padding bits.

This is less of a concern for atomic_ref<T> because the referenced object is not constructed and is unchanged by the construction of an atomic_ref<T> object. In regard to the compare_exchange operations, the atomic_ref<T> capability does not introduce additional effects to existing concerns with atomic<T>.

Recommended feature text macro

__cpp_lib_atomic_ref

add to Header <atomic> synopsis

namespace std {

namespace experimental {

template< class T > struct atomic_ref ;

template< class T > struct atomic_ref< T * >;

}}

add section to Class template atomic

template< class T > struct atomic_ref {

using value_type = T;

static constexpr size_t required_alignment = implementation-defined ;

static constexpr bool is_always_lock_free = implementation-defined ;

bool is_lock_free() const noexcept;

void store( T , memory_order = memory_order_seq_cst ) const noexcept;

T load( memory_order = memory_order_seq_cst ) const noexcept;

operator T() const noexcept ;

T exchange( T , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_weak( T& , T , memory_order , memory_order ) const noexcept;

bool compare_exchange_strong( T& , T , memory_order , memory_order ) const noexcept;

bool compare_exchange_weak( T& , T , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_strong( T&, T, memory_order = memory_order_seq_cst ) const noexcept;

atomic_ref() = delete ;

atomic_ref( const atomic_ref & );

atomic_ref & operator = ( const atomic_ref & ) = delete ;

explicit atomic_ref( T & obj );

T operator=(T) const noexcept ;

};

static constexpr size_t required_alignment = implementation-defined ;

The required alignment of an object to be referenced by an atomic reference, which is at least alignof(T). [Note: An architecture may support lock-free atomic operations on objects of type T only if those objects meet a required alignment. The intent is for atomic_ref to provide lock-free atomic operations whenever possible. For example, an architecture may be able to support lock-free operations on std::complex<double> only if aligned to 2*alignof(double) and not alignof(double) . - end note]

atomic_ref( T & obj );

Construct an atomic reference that references the non-atomic object. Atomic operations applied to object through a referencing atomic reference are atomic with respect to atomic operations applied through any other atomic reference that references that object.

Requires: The referenced non-atomic object shall be aligned to required_alignment. The lifetime (3.8) of *this shall not exceed the lifetime of the referenced non-atomic object. While any atomic_ref instance exists that references the object all accesses of that object shall exclusively occur through those atomic_ref instances. If the referenced object is of a class or aggregate type then members of that object shall not be concurrently referenced by an atomic_ref object.

Throws: Nothing

Effects: *this references the non-atomic object* [Note: The constructor may acquire a shared resource, such as a lock associated with the referenced object, to enable atomic operations applied to the referenced non-atomic object. - end note]

add to Specializations for integers

template<> struct atomic_ref< integral > {

using value_type = integral ;

using difference_type = value_type;

static constexpr size_t required_alignment = implementation-defined ;

static constexpr bool is_always_lock_free = implementation-defined ;

bool is_lock_free() const noexcept;

void store( integral , memory_order = memory_order_seq_cst ) const noexcept;

integral load( memory_order = memory_order_seq_cst ) const noexcept;

operator integral () const noexcept ;

integral exchange( integral , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_weak( integral & , integral , memory_order , memory_order ) const noexcept;

bool compare_exchange_strong( integral & , integral , memory_order , memory_order ) const noexcept;

bool compare_exchange_weak( integral & , integral , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_strong( integral &, integral , memory_order = memory_order_seq_cst ) const noexcept;

integral fetch_add( integral , memory_order = memory_order_seq_cst) const noexcept;

integral fetch_sub( integral , memory_order = memory_order_seq_cst) const noexcept;

integral fetch_and( integral , memory_order = memory_order_seq_cst) const noexcept;

integral fetch_or( integral , memory_order = memory_order_seq_cst) const noexcept;

integral fetch_xor( integral , memory_order = memory_order_seq_cst) const noexcept;

atomic_ref() = delete ;

atomic_ref( const atomic_ref & ) = delete ;

atomic_ref & operator = ( const atomic_ref & ) = delete ;

explicit atomic_ref( integral & obj );

integral operator=( integral ) const noexcept ;

integral operator++(int) const noexcept;

integral operator--(int) const noexcept;

integral operator++() const noexcept;

integral operator--() const noexcept;

integral operator+=( integral ) const noexcept;

integral operator-=( integral ) const noexcept;

integral operator&=( integral ) const noexcept;

integral operator|=( integral ) const noexcept;

integral operator^=( integral ) const noexcept;

};

add to Specializations for floating-point

template<> struct atomic_ref< floating-point > {

using value_type = floating-point ;

using difference_type = value_type;

static constexpr size_t required_alignment = implementation-defined ;

static constexpr bool is_always_lock_free = implementation-defined ;

bool is_lock_free() const noexcept;

void store( floating-point , memory_order = memory_order_seq_cst ) const noexcept;

floating-point load( memory_order = memory_order_seq_cst ) const noexcept;

operator floating-point () const noexcept ;

floating-point exchange( floating-point , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_weak( floating-point & , floating-point , memory_order , memory_order ) const noexcept;

bool compare_exchange_strong( floating-point & , floating-point , memory_order , memory_order ) const noexcept;

bool compare_exchange_weak( floating-point & , floating-point , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_strong( floating-point &, floating-point , memory_order = memory_order_seq_cst ) const noexcept;

floating-point fetch_add( floating-point , memory_order = memory_order_seq_cst) const noexcept;

floating-point fetch_sub( floating-point , memory_order = memory_order_seq_cst) const noexcept;

atomic_ref() = delete ;

atomic_ref( const atomic_ref & );

atomic_ref & operator = ( const atomic_ref & ) = delete ;

explicit atomic_ref( floating-point & obj ) noexcept ;

floating-point operator=( floating-point ) noexcept ;

floating-point operator+=( floating-point ) const noexcept ;

floating-point operator-=( floating-point ) const noexcept ;

};

add to Partial specializations for pointers

template<class T> struct atomic_ref< T * > {

using value_type = T * ;

using difference_type = ptrdiff_t;

static constexpr size_t required_alignment = implementation-defined ;

static constexpr bool is_always_lock_free = implementation-defined ;

bool is_lock_free() const noexcept;

void store( T * , memory_order = memory_order_seq_cst ) const noexcept;

T * load( memory_order = memory_order_seq_cst ) const noexcept;

operator T * () const noexcept ;

T * exchange( T * , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_weak( T * & , T * , memory_order , memory_order ) const noexcept;

bool compare_exchange_strong( T * & , T * , memory_order , memory_order ) const noexcept;

bool compare_exchange_weak( T * & , T * , memory_order = memory_order_seq_cst ) const noexcept;

bool compare_exchange_strong( T * &, T * , memory_order = memory_order_seq_cst ) const noexcept;

T * fetch_add( difference_type , memory_order = memory_order_seq_cst) const noexcept;

T * fetch_sub( difference_type , memory_order = memory_order_seq_cst) const noexcept;

~atomic_ref();

atomic_ref() = delete ;

atomic_ref( const atomic_ref & );

atomic_ref & operator = ( const atomic_ref & ) = delete ;

explicit atomic_ref( T * & obj );

T * operator=( T * ) const noexcept ;

T * operator++(int) const noexcept;

T * operator--(int) const noexcept;

T * operator++() const noexcept;

T * operator--() const noexcept;

T * operator+=( difference_type ) const noexcept;

T * operator-=( difference_type ) const noexcept;

};