pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T > Class Template Reference

#include <Concurrent.h>

Inherited by pongasoft::Utils::Concurrent::LockFree::AtomicValue< T >, and pongasoft::Utils::Concurrent::LockFree::SingleElementQueue< T >.

Classes
struct	Element

Public Member Functions
bool	__isLockFree () const
bool	isEmpty () const
	SingleElementStorage (std::unique_ptr< T > iElement, bool iIsEmpty) noexcept
	~SingleElementStorage ()

Protected Member Functions
std::unique_ptr< Element >	__newElement () const
std::unique_ptr< T >	__newT () const
std::unique_ptr< Element >	load (std::unique_ptr< Element > iElement)
std::unique_ptr< Element >	store (std::unique_ptr< Element > iElement)

Private Attributes
std::atomic< Element * >	fSingleElement

Detailed Description

template<typename T>
class pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >

This (internal) class stores a single element. This implementation is NOT fully thread safe. It is only thread safe as long as load is called by a single thread and store is called by another single thread as well (if load and store are called by the same thread it is obviously thread safe since it would be mono thread...)

Todo:

the implementation somehow assumes that T is a real type, not a primitive.. maybe there is a way to write a primitive version (if that becomes a necessity)

Constructor & Destructor Documentation

SingleElementStorage()

template<typename T>

pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::SingleElementStorage ( std::unique_ptr< T >  iElement, bool  iIsEmpty  )

inlinenoexcept

~SingleElementStorage()

template<typename T>

pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::~SingleElementStorage ( )

inline

Member Function Documentation

__isLockFree()

template<typename T>

bool pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::__isLockFree ( ) const

inline

Used (from test) to make sure that it is a lock free implementation. Relies on an atomic pointer and in general the implementation should be lock free but we make sure here.

__newElement()

template<typename T>

std::unique_ptr<Element> pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::__newElement ( ) const

inlineprotected

__newT()

template<typename T>

std::unique_ptr<T> pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::__newT ( ) const

inlineprotected

isEmpty()

template<typename T>

bool pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::isEmpty ( ) const

inline

load()

template<typename T>

std::unique_ptr<Element> pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::load ( std::unique_ptr< Element >  iElement )

inlineprotected

Loads an element from storage. In order to avoid copy and memory allocation, the api actually takes an element to replace the one that is returned. The correct usage pattern should be:

if(!isEmpty()) myPtr = std::move(load(std::move(myPtr)));

That way you are guaranteed that what load returns will have a fNew flag set to true...

Returns

element with flag fNew set to true if there was a new element, false otherwise

store()

template<typename T>

std::unique_ptr<Element> pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::store ( std::unique_ptr< Element >  iElement )

inlineprotected

Stores an element in the storage. Replaces the current one if there is one. In order to avoid copy and memory allocation, the storage assumes ownership of the element provided and as result returns the one it replaces.

Member Data Documentation

fSingleElement

template<typename T>

std::atomic<Element *> pongasoft::Utils::Concurrent::LockFree::SingleElementStorage< T >::fSingleElement

private

---

The documentation for this class was generated from the following file:

• pongasoft/Utils/Concurrent/Concurrent.h