239 lines
6.5 KiB
C
239 lines
6.5 KiB
C
|
/***
|
||
|
* ==++==
|
||
|
*
|
||
|
* Copyright (c) Microsoft Corporation. All rights reserved.
|
||
|
*
|
||
|
* ==--==
|
||
|
* =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
|
||
|
*
|
||
|
* pplinterface.h
|
||
|
*
|
||
|
* Parallel Patterns Library - PPL interface definitions
|
||
|
*
|
||
|
* =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
|
||
|
****/
|
||
|
|
||
|
#pragma once
|
||
|
|
||
|
#ifndef _PPLINTERFACE_H
|
||
|
#define _PPLINTERFACE_H
|
||
|
|
||
|
#if defined(_CRTBLD)
|
||
|
#elif defined(_MS_WINDOWS)
|
||
|
#if (_MSC_VER >= 1700)
|
||
|
#define _USE_REAL_ATOMICS
|
||
|
#endif
|
||
|
#else // GCC compiler
|
||
|
#define _USE_REAL_ATOMICS
|
||
|
#endif
|
||
|
|
||
|
#include <memory>
|
||
|
#ifdef _USE_REAL_ATOMICS
|
||
|
#include <atomic>
|
||
|
#endif
|
||
|
|
||
|
namespace Concurrency
|
||
|
{
|
||
|
|
||
|
/// <summary>
|
||
|
/// An elementary abstraction for a task, defined as <c>void (__cdecl * TaskProc_t)(void *)</c>. A <c>TaskProc</c> is called to
|
||
|
/// invoke the body of a task.
|
||
|
/// </summary>
|
||
|
/**/
|
||
|
typedef void (__cdecl * TaskProc_t)(void *);
|
||
|
|
||
|
/// <summary>
|
||
|
/// Scheduler Interface
|
||
|
/// </summary>
|
||
|
struct __declspec(novtable) scheduler_interface
|
||
|
{
|
||
|
virtual void schedule( TaskProc_t, void* ) = 0;
|
||
|
};
|
||
|
|
||
|
/// <summary>
|
||
|
/// Represents a pointer to a scheduler. This class exists to allow the
|
||
|
/// the specification of a shared lifetime by using shared_ptr or just
|
||
|
/// a plain reference by using raw pointer.
|
||
|
/// </summary>
|
||
|
struct scheduler_ptr
|
||
|
{
|
||
|
/// <summary>
|
||
|
/// Creates a scheduler pointer from shared_ptr to scheduler
|
||
|
/// </summary>
|
||
|
explicit scheduler_ptr(std::shared_ptr<scheduler_interface> scheduler) : m_sharedScheduler(std::move(scheduler))
|
||
|
{
|
||
|
m_scheduler = m_sharedScheduler.get();
|
||
|
}
|
||
|
|
||
|
/// <summary>
|
||
|
/// Creates a scheduler pointer from raw pointer to scheduler
|
||
|
/// </summary>
|
||
|
explicit scheduler_ptr(_In_opt_ scheduler_interface * pScheduler) : m_scheduler(pScheduler)
|
||
|
{
|
||
|
}
|
||
|
|
||
|
/// <summary>
|
||
|
/// Behave like a pointer
|
||
|
/// </summary>
|
||
|
scheduler_interface *operator->() const
|
||
|
{
|
||
|
return get();
|
||
|
}
|
||
|
|
||
|
/// <summary>
|
||
|
/// Returns the raw pointer to the scheduler
|
||
|
/// </summary>
|
||
|
scheduler_interface * get() const
|
||
|
{
|
||
|
return m_scheduler;
|
||
|
}
|
||
|
|
||
|
/// <summary>
|
||
|
/// Test whether the scheduler pointer is non-null
|
||
|
/// </summary>
|
||
|
operator bool() const { return get() != nullptr; }
|
||
|
|
||
|
private:
|
||
|
|
||
|
std::shared_ptr<scheduler_interface> m_sharedScheduler;
|
||
|
scheduler_interface * m_scheduler;
|
||
|
};
|
||
|
|
||
|
|
||
|
/// <summary>
|
||
|
/// Describes the execution status of a <c>task_group</c> or <c>structured_task_group</c> object. A value of this type is returned
|
||
|
/// by numerous methods that wait on tasks scheduled to a task group to complete.
|
||
|
/// </summary>
|
||
|
/// <seealso cref="task_group Class"/>
|
||
|
/// <seealso cref="task_group::wait Method"/>
|
||
|
/// <seealso cref="task_group::run_and_wait Method"/>
|
||
|
/// <seealso cref="structured_task_group Class"/>
|
||
|
/// <seealso cref="structured_task_group::wait Method"/>
|
||
|
/// <seealso cref="structured_task_group::run_and_wait Method"/>
|
||
|
/**/
|
||
|
enum task_group_status
|
||
|
{
|
||
|
/// <summary>
|
||
|
/// The tasks queued to the <c>task_group</c> object have not completed. Note that this value is not presently returned by
|
||
|
/// the Concurrency Runtime.
|
||
|
/// </summary>
|
||
|
/**/
|
||
|
not_complete,
|
||
|
|
||
|
/// <summary>
|
||
|
/// The tasks queued to the <c>task_group</c> or <c>structured_task_group</c> object completed successfully.
|
||
|
/// </summary>
|
||
|
/**/
|
||
|
completed,
|
||
|
|
||
|
/// <summary>
|
||
|
/// The <c>task_group</c> or <c>structured_task_group</c> object was canceled. One or more tasks may not have executed.
|
||
|
/// </summary>
|
||
|
/**/
|
||
|
canceled
|
||
|
};
|
||
|
|
||
|
namespace details
|
||
|
{
|
||
|
/// <summary>
|
||
|
/// Atomics
|
||
|
/// </summary>
|
||
|
#ifdef _USE_REAL_ATOMICS
|
||
|
typedef std::atomic<long> atomic_long;
|
||
|
typedef std::atomic<size_t> atomic_size_t;
|
||
|
|
||
|
template<typename _T>
|
||
|
_T atomic_compare_exchange(std::atomic<_T>& _Target, _T _Exchange, _T _Comparand)
|
||
|
{
|
||
|
_T _Result = _Comparand;
|
||
|
_Target.compare_exchange_strong(_Result, _Exchange);
|
||
|
return _Result;
|
||
|
}
|
||
|
|
||
|
template<typename _T>
|
||
|
_T atomic_exchange(std::atomic<_T>& _Target, _T _Value)
|
||
|
{
|
||
|
return _Target.exchange(_Value);
|
||
|
}
|
||
|
|
||
|
template<typename _T>
|
||
|
_T atomic_increment(std::atomic<_T>& _Target)
|
||
|
{
|
||
|
return _Target.fetch_add(1) + 1;
|
||
|
}
|
||
|
|
||
|
template<typename _T>
|
||
|
_T atomic_decrement(std::atomic<_T>& _Target)
|
||
|
{
|
||
|
return _Target.fetch_sub(1) - 1;
|
||
|
}
|
||
|
|
||
|
template<typename _T>
|
||
|
_T atomic_add(std::atomic<_T>& _Target, _T value)
|
||
|
{
|
||
|
return _Target.fetch_add(value) + value;
|
||
|
}
|
||
|
|
||
|
#else // not _USE_REAL_ATOMICS
|
||
|
|
||
|
typedef long volatile atomic_long;
|
||
|
typedef size_t volatile atomic_size_t;
|
||
|
|
||
|
template<class T>
|
||
|
inline T atomic_exchange(T volatile& _Target, T _Value)
|
||
|
{
|
||
|
return _InterlockedExchange(&_Target, _Value);
|
||
|
}
|
||
|
|
||
|
inline long atomic_increment(long volatile & _Target)
|
||
|
{
|
||
|
return _InterlockedIncrement(&_Target);
|
||
|
}
|
||
|
|
||
|
inline long atomic_add(long volatile & _Target, long value)
|
||
|
{
|
||
|
return _InterlockedExchangeAdd(&_Target, value) + value;
|
||
|
}
|
||
|
|
||
|
inline size_t atomic_increment(size_t volatile & _Target)
|
||
|
{
|
||
|
#if (defined(_M_IX86) || defined(_M_ARM))
|
||
|
return static_cast<size_t>(_InterlockedIncrement(reinterpret_cast<long volatile *>(&_Target)));
|
||
|
#else
|
||
|
return static_cast<size_t>(_InterlockedIncrement64(reinterpret_cast<__int64 volatile *>(&_Target)));
|
||
|
#endif
|
||
|
}
|
||
|
|
||
|
inline long atomic_decrement(long volatile & _Target)
|
||
|
{
|
||
|
return _InterlockedDecrement(&_Target);
|
||
|
}
|
||
|
|
||
|
inline size_t atomic_decrement(size_t volatile & _Target)
|
||
|
{
|
||
|
#if (defined(_M_IX86) || defined(_M_ARM))
|
||
|
return static_cast<size_t>(_InterlockedDecrement(reinterpret_cast<long volatile *>(&_Target)));
|
||
|
#else
|
||
|
return static_cast<size_t>(_InterlockedDecrement64(reinterpret_cast<__int64 volatile *>(&_Target)));
|
||
|
#endif
|
||
|
}
|
||
|
|
||
|
inline long atomic_compare_exchange(long volatile & _Target, long _Exchange, long _Comparand)
|
||
|
{
|
||
|
return _InterlockedCompareExchange(&_Target, _Exchange, _Comparand);
|
||
|
}
|
||
|
|
||
|
inline size_t atomic_compare_exchange(size_t volatile & _Target, size_t _Exchange, size_t _Comparand)
|
||
|
{
|
||
|
#if (defined(_M_IX86) || defined(_M_ARM))
|
||
|
return static_cast<size_t>(_InterlockedCompareExchange(reinterpret_cast<long volatile *>(_Target), static_cast<long>(_Exchange), static_cast<long>(_Comparand)));
|
||
|
#else
|
||
|
return static_cast<size_t>(_InterlockedCompareExchange64(reinterpret_cast<__int64 volatile *>(_Target), static_cast<__int64>(_Exchange), static_cast<__int64>(_Comparand)));
|
||
|
#endif
|
||
|
}
|
||
|
#endif // _USE_REAL_ATOMICS
|
||
|
|
||
|
}} // namespace Concurrency
|
||
|
|
||
|
#endif // _PPLINTERFACE_H
|