ODE的线程模型

dxWorld线程模型

dxWorld继承dxThreadingBase,dxIThreadingDefaultImplProvider用来实现多线程调度。
dxIThreadingDefaultImplProvider仅用来将dxWorld中的实现传送给dxThreadingBase。
dxThreadingBase封装了一个多线程调度接口，但是内部它将具体的事务委托给具体的实现。
dxThreadingBase通过接口dxThreadingFunctionsInfo调用具体的多线程调度实现。

dxThreadingFunctionsInfo

dxThreadingFunctionsInfo其实就是一个函数表：

static const dxThreadingFunctionsInfo g_builtin_threading_functions = {    sizeof(dxThreadingFunctionsInfo), // unsigned struct_size;    &AllocMutexGroup, // dMutexGroupAllocFunction *alloc_mutex_group;    &FreeMutexGroup, // dMutexGroupFreeFunction *free_mutex_group;    &LockMutexGroupMutex, // dMutexGroupMutexLockFunction *lock_group_mutex;    &UnlockMutexGroupMutex, // dMutexGroupMutexUnlockFunction *unlock_group_mutex;    &AllocThreadedCallWait, // dThreadedCallWaitAllocFunction *alloc_call_wait;    &ResetThreadedCallWait, // dThreadedCallWaitResetFunction *reset_call_wait;    &FreeThreadedCallWait, // dThreadedCallWaitFreeFunction *free_call_wait;    &PostThreadedCall, // dThreadedCallPostFunction *post_call;    &AlterThreadedCallDependenciesCount, // dThreadedCallDependenciesCountAlterFunction *alter_call_dependencies_count;    &WaitThreadedCall, // dThreadedCallWaitFunction *wait_call;    &RetrieveThreadingThreadCount, // dThreadingImplThreadCountRetrieveFunction *retrieve_thread_count;    &PreallocateResourcesForThreadedCalls, // dThreadingImplResourcesForCallsPreallocateFunction *preallocate_resources_for_calls;    // &TryLockMutexGroupMutex, // dMutexGroupMutexTryLockFunction *trylock_group_mutex;};

dxThreadingFunctionsInfo将具体的调用委托给具体的多线程实现，它自己仅仅起到一个接口的作用。
如它的一个函数AllocMutexGroup，可以看到它仅仅是委托给具体的实现去做。

static dMutexGroupID AllocMutexGroup(dThreadingImplementationID impl, dmutexindex_t Mutex_count, const char *const *Mutex_names_ptr/*=NULL*/){    dIMutexGroup *mutex_group = ((dxIThreadingImplementation *)impl)->AllocMutexGroup(Mutex_count);    return (dMutexGroupID)mutex_group;}

实现层

具体的多线程调度实现有一组模板类来完成，它们是dxtemplateJobListContainer,dxtemplateJobListSelfHandler,dxtemplateThreadingImplementation
通过模板类的参数来完成类型组装。dxSelfThreadedThreading就是一个具体的实现。它其实是一个伪装的多线程实现。
真正的多线程实现是dxMultiThreadedThreading。下面是代码片段，被剪接到以前便于查看。

typedef dxtemplateJobListContainer<dxFakeLull, dxFakeMutex, dxFakeAtomicsProvider> dxSelfThreadedJobListContainer;typedef dxtemplateJobListSelfHandler<dxSelfWakeup, dxSelfThreadedJobListContainer> dxSelfThreadedJobListHandler;typedef dxtemplateThreadingImplementation<dxSelfThreadedJobListContainer, dxSelfThreadedJobListHandler> dxSelfThreadedThreading;/*extern */dThreadingImplementationID dThreadingAllocateSelfThreadedImplementation(){    dxSelfThreadedThreading *threading = new dxSelfThreadedThreading();    if (threading != NULL && !threading->InitializeObject())    {        delete threading;        threading = NULL;    }    dxIThreadingImplementation *impl = threading;    return (dThreadingImplementationID)impl;}typedef dxtemplateJobListContainer<dxtemplateThreadedLull<dxCondvarWakeup, dxOUAtomicsProvider, false>, dxMutexMutex, dxOUAtomicsProvider> dxMultiThreadedJobListContainer;typedef dxtemplateJobListThreadedHandler<dxCondvarWakeup, dxMultiThreadedJobListContainer> dxMultiThreadedJobListHandler;typedef dxtemplateThreadingImplementation<dxMultiThreadedJobListContainer, dxMultiThreadedJobListHandler> dxMultiThreadedThreading;/*extern */dThreadingImplementationID dThreadingAllocateMultiThreadedImplementation(){#if dBUILTIN_THREADING_IMPL_ENABLED    dxMultiThreadedThreading *threading = new dxMultiThreadedThreading();    if (threading != NULL && !threading->InitializeObject())    {        delete threading;        threading = NULL;    }#else    dxIThreadingImplementation *threading = NULL;#endif // #if dBUILTIN_THREADING_IMPL_ENABLED    dxIThreadingImplementation *impl = threading;    return (dThreadingImplementationID)impl;}

在分析这些模板类前，先看看它的模板参数类是干什么的。

dxEventWakeup,dxCondvarWakeup

dxEventWakeup(win32)，dxCondvarWakeup(posix)
dxEventWakeup创建一个自动Event对象(win32)，当几个线程调用WaitWakeup等在那里时。
其他线程可以通过调用WakeupAThread来唤醒一个线程，或者通过WakeupAllThreads
来唤醒全部等待的线程。ResetWakeup重新将Event设置成无信号状态。
dxCondvarWakeup实现类似机制。

dxtemplateThreadedLull

有4个函数
RegisterToLull增加计数
UnregisterFromLull减少计数
WaitForLullAlarm等待醒来
SignalLullAlarmIfAnyRegistrants让一个等待的线程醒来

dxCriticalSectionMutex

LockMutex 进入临界区对象
TryLockMutex 如果临界区对象没有其他线程拥有，则进入临界区。
否则立刻返回bool。
UnlockMutex 释放临界区对象

dxtemplateJobListContainer

有一个成员dxThreadedJobInfo *m_job_list。dxThreadedJobInfo是一个单向链表。
dxThreadedJobInfo仅仅是一个数据结构，并没有封装算法。