Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
|
|
|
|
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
|
|
|
|
/* This Source Code Form is subject to the terms of the Mozilla Public
|
|
|
|
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
|
|
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
|
|
|
|
|
|
|
|
#include "mozilla/ThreadEventQueue.h"
|
|
|
|
#include "mozilla/EventQueue.h"
|
2017-09-25 12:59:47 +03:00
|
|
|
#include "LabeledEventQueue.h"
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
|
|
|
|
#include "LeakRefPtr.h"
|
|
|
|
#include "nsComponentManagerUtils.h"
|
|
|
|
#include "nsIThreadInternal.h"
|
|
|
|
#include "nsThreadUtils.h"
|
|
|
|
#include "PrioritizedEventQueue.h"
|
|
|
|
#include "ThreadEventTarget.h"
|
|
|
|
|
|
|
|
using namespace mozilla;
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
class ThreadEventQueue<InnerQueueT>::NestedSink : public ThreadTargetSink
|
|
|
|
{
|
|
|
|
public:
|
|
|
|
NestedSink(EventQueue* aQueue, ThreadEventQueue* aOwner)
|
|
|
|
: mQueue(aQueue)
|
|
|
|
, mOwner(aOwner)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
bool PutEvent(already_AddRefed<nsIRunnable>&& aEvent,
|
|
|
|
EventPriority aPriority) final
|
|
|
|
{
|
|
|
|
return mOwner->PutEventInternal(Move(aEvent), aPriority, this);
|
|
|
|
}
|
|
|
|
|
|
|
|
void Disconnect(const MutexAutoLock& aProofOfLock) final
|
|
|
|
{
|
|
|
|
mQueue = nullptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
friend class ThreadEventQueue;
|
|
|
|
|
|
|
|
// This is a non-owning reference. It must live at least until Disconnect is
|
|
|
|
// called to clear it out.
|
|
|
|
EventQueue* mQueue;
|
|
|
|
RefPtr<ThreadEventQueue> mOwner;
|
|
|
|
};
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
ThreadEventQueue<InnerQueueT>::ThreadEventQueue(UniquePtr<InnerQueueT> aQueue)
|
|
|
|
: mBaseQueue(Move(aQueue))
|
|
|
|
, mLock("ThreadEventQueue")
|
|
|
|
, mEventsAvailable(mLock, "EventsAvail")
|
|
|
|
{
|
|
|
|
static_assert(IsBaseOf<AbstractEventQueue, InnerQueueT>::value,
|
|
|
|
"InnerQueueT must be an AbstractEventQueue subclass");
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
ThreadEventQueue<InnerQueueT>::~ThreadEventQueue()
|
|
|
|
{
|
|
|
|
MOZ_ASSERT(mNestedQueues.IsEmpty());
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
bool
|
|
|
|
ThreadEventQueue<InnerQueueT>::PutEvent(already_AddRefed<nsIRunnable>&& aEvent,
|
|
|
|
EventPriority aPriority)
|
|
|
|
{
|
|
|
|
return PutEventInternal(Move(aEvent), aPriority, nullptr);
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
bool
|
|
|
|
ThreadEventQueue<InnerQueueT>::PutEventInternal(already_AddRefed<nsIRunnable>&& aEvent,
|
|
|
|
EventPriority aPriority,
|
|
|
|
NestedSink* aSink)
|
|
|
|
{
|
|
|
|
// We want to leak the reference when we fail to dispatch it, so that
|
|
|
|
// we won't release the event in a wrong thread.
|
|
|
|
LeakRefPtr<nsIRunnable> event(Move(aEvent));
|
|
|
|
nsCOMPtr<nsIThreadObserver> obs;
|
|
|
|
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
|
|
|
|
if (mEventsAreDoomed) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (aSink) {
|
|
|
|
if (!aSink->mQueue) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
aSink->mQueue->PutEvent(event.take(), aPriority, lock);
|
|
|
|
} else {
|
|
|
|
mBaseQueue->PutEvent(event.take(), aPriority, lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
mEventsAvailable.Notify();
|
|
|
|
|
|
|
|
// Make sure to grab the observer before dropping the lock, otherwise the
|
|
|
|
// event that we just placed into the queue could run and eventually delete
|
|
|
|
// this nsThread before the calling thread is scheduled again. We would then
|
|
|
|
// crash while trying to access a dead nsThread.
|
|
|
|
obs = mObserver;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (obs) {
|
|
|
|
obs->OnDispatchedEvent();
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
already_AddRefed<nsIRunnable>
|
|
|
|
ThreadEventQueue<InnerQueueT>::GetEvent(bool aMayWait,
|
|
|
|
EventPriority* aPriority)
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
|
|
|
|
nsCOMPtr<nsIRunnable> event;
|
|
|
|
for (;;) {
|
|
|
|
if (mNestedQueues.IsEmpty()) {
|
|
|
|
event = mBaseQueue->GetEvent(aPriority, lock);
|
|
|
|
} else {
|
|
|
|
// We always get events from the topmost queue when there are nested
|
|
|
|
// queues.
|
|
|
|
event = mNestedQueues.LastElement().mQueue->GetEvent(aPriority, lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (event || !aMayWait) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
mEventsAvailable.Wait();
|
|
|
|
}
|
|
|
|
|
|
|
|
return event.forget();
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
bool
|
|
|
|
ThreadEventQueue<InnerQueueT>::HasPendingEvent()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
|
|
|
|
// We always get events from the topmost queue when there are nested queues.
|
|
|
|
if (mNestedQueues.IsEmpty()) {
|
2017-07-29 00:56:49 +03:00
|
|
|
return mBaseQueue->HasReadyEvent(lock);
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
} else {
|
2017-07-29 00:56:49 +03:00
|
|
|
return mNestedQueues.LastElement().mQueue->HasReadyEvent(lock);
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
bool
|
|
|
|
ThreadEventQueue<InnerQueueT>::ShutdownIfNoPendingEvents()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
2017-07-29 00:56:49 +03:00
|
|
|
if (mNestedQueues.IsEmpty() && mBaseQueue->IsEmpty(lock)) {
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
mEventsAreDoomed = true;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
void
|
|
|
|
ThreadEventQueue<InnerQueueT>::EnableInputEventPrioritization()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
mBaseQueue->EnableInputEventPrioritization(lock);
|
|
|
|
}
|
|
|
|
|
2017-07-28 10:14:54 +03:00
|
|
|
template<class InnerQueueT>
|
|
|
|
void
|
|
|
|
ThreadEventQueue<InnerQueueT>::FlushInputEventPrioritization()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
mBaseQueue->FlushInputEventPrioritization(lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
void
|
|
|
|
ThreadEventQueue<InnerQueueT>::SuspendInputEventPrioritization()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
mBaseQueue->SuspendInputEventPrioritization(lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
void
|
|
|
|
ThreadEventQueue<InnerQueueT>::ResumeInputEventPrioritization()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
mBaseQueue->ResumeInputEventPrioritization(lock);
|
|
|
|
}
|
|
|
|
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
template<class InnerQueueT>
|
|
|
|
already_AddRefed<nsISerialEventTarget>
|
|
|
|
ThreadEventQueue<InnerQueueT>::PushEventQueue()
|
|
|
|
{
|
|
|
|
auto queue = MakeUnique<EventQueue>();
|
|
|
|
RefPtr<NestedSink> sink = new NestedSink(queue.get(), this);
|
|
|
|
RefPtr<ThreadEventTarget> eventTarget = new ThreadEventTarget(sink, NS_IsMainThread());
|
|
|
|
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
|
|
|
|
mNestedQueues.AppendElement(NestedQueueItem(Move(queue), eventTarget));
|
|
|
|
return eventTarget.forget();
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
void
|
|
|
|
ThreadEventQueue<InnerQueueT>::PopEventQueue(nsIEventTarget* aTarget)
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
|
|
|
|
MOZ_ASSERT(!mNestedQueues.IsEmpty());
|
|
|
|
|
|
|
|
NestedQueueItem& item = mNestedQueues.LastElement();
|
|
|
|
|
|
|
|
MOZ_ASSERT(aTarget == item.mEventTarget);
|
|
|
|
|
|
|
|
// Disconnect the event target that will be popped.
|
|
|
|
item.mEventTarget->Disconnect(lock);
|
|
|
|
|
|
|
|
AbstractEventQueue* prevQueue =
|
|
|
|
mNestedQueues.Length() == 1
|
|
|
|
? static_cast<AbstractEventQueue*>(mBaseQueue.get())
|
|
|
|
: static_cast<AbstractEventQueue*>(mNestedQueues[mNestedQueues.Length() - 2].mQueue.get());
|
|
|
|
|
|
|
|
// Move events from the old queue to the new one.
|
|
|
|
nsCOMPtr<nsIRunnable> event;
|
|
|
|
EventPriority prio;
|
|
|
|
while ((event = item.mQueue->GetEvent(&prio, lock))) {
|
|
|
|
prevQueue->PutEvent(event.forget(), prio, lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
mNestedQueues.RemoveElementAt(mNestedQueues.Length() - 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
already_AddRefed<nsIThreadObserver>
|
|
|
|
ThreadEventQueue<InnerQueueT>::GetObserver()
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
2017-10-26 05:46:50 +03:00
|
|
|
return do_AddRef(mObserver);
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
already_AddRefed<nsIThreadObserver>
|
|
|
|
ThreadEventQueue<InnerQueueT>::GetObserverOnThread()
|
|
|
|
{
|
2017-10-26 05:46:50 +03:00
|
|
|
return do_AddRef(mObserver);
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
template<class InnerQueueT>
|
|
|
|
void
|
|
|
|
ThreadEventQueue<InnerQueueT>::SetObserver(nsIThreadObserver* aObserver)
|
|
|
|
{
|
|
|
|
MutexAutoLock lock(mLock);
|
|
|
|
mObserver = aObserver;
|
|
|
|
}
|
|
|
|
|
|
|
|
namespace mozilla {
|
|
|
|
template class ThreadEventQueue<EventQueue>;
|
|
|
|
template class ThreadEventQueue<PrioritizedEventQueue<EventQueue>>;
|
2017-09-02 02:39:13 +03:00
|
|
|
template class ThreadEventQueue<PrioritizedEventQueue<LabeledEventQueue>>;
|
Bug 1382922 - Refactor event queue to allow multiple implementations (r=erahm)
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
2017-06-21 05:42:13 +03:00
|
|
|
}
|