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 "PrioritizedEventQueue.h"
|
|
|
|
|
#include "mozilla/EventQueue.h"
|
2017-09-25 12:50:04 +03:00
|
|
|
#include "mozilla/ScopeExit.h"
|
2019-07-26 04:10:23 +03:00
|
|
|
#include "mozilla/StaticPrefs_threads.h"
|
2019-09-24 17:33:57 +03:00
|
|
|
#include "mozilla/ipc/IdleSchedulerChild.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 "nsThreadManager.h"
|
|
|
|
|
#include "nsXPCOMPrivate.h" // for gXPCOMThreadsShutDown
|
2017-09-25 12:50:04 +03:00
|
|
|
#include "InputEventStatistics.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
|
|
|
|
|
|
|
|
using namespace mozilla;
|
|
|
|
|
|
2019-09-24 17:33:57 +03:00
|
|
|
PrioritizedEventQueue::PrioritizedEventQueue(
|
2019-10-20 18:08:44 +03:00
|
|
|
already_AddRefed<nsIIdlePeriod>&& aIdlePeriod)
|
2019-09-24 17:33:57 +03:00
|
|
|
: mHighQueue(MakeUnique<EventQueue>(EventQueuePriority::High)),
|
2019-11-21 06:47:19 +03:00
|
|
|
mInputQueue(MakeUnique<EventQueueSized<32>>(EventQueuePriority::Input)),
|
2019-09-24 17:33:57 +03:00
|
|
|
mMediumHighQueue(MakeUnique<EventQueue>(EventQueuePriority::MediumHigh)),
|
2019-11-21 06:47:19 +03:00
|
|
|
mNormalQueue(MakeUnique<EventQueueSized<64>>(EventQueuePriority::Normal)),
|
2019-09-24 17:33:57 +03:00
|
|
|
mDeferredTimersQueue(
|
|
|
|
|
MakeUnique<EventQueue>(EventQueuePriority::DeferredTimers)),
|
|
|
|
|
mIdleQueue(MakeUnique<EventQueue>(EventQueuePriority::Idle)),
|
2019-10-20 18:08:44 +03:00
|
|
|
mIdlePeriodState(std::move(aIdlePeriod)) {}
|
2019-09-24 17:33:57 +03:00
|
|
|
|
2019-10-20 18:08:44 +03:00
|
|
|
PrioritizedEventQueue::~PrioritizedEventQueue() = default;
|
2019-09-24 17:33:57 +03:00
|
|
|
|
2019-05-25 20:46:15 +03:00
|
|
|
void PrioritizedEventQueue::PutEvent(already_AddRefed<nsIRunnable>&& aEvent,
|
|
|
|
|
EventQueuePriority aPriority,
|
2019-11-09 00:07:36 +03:00
|
|
|
const MutexAutoLock& aProofOfLock,
|
|
|
|
|
mozilla::TimeDuration* aDelay) {
|
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
|
|
|
// Double check the priority with a QI.
|
|
|
|
|
RefPtr<nsIRunnable> event(aEvent);
|
2019-01-26 20:18:05 +03:00
|
|
|
EventQueuePriority priority = aPriority;
|
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
|
|
|
|
2019-01-26 20:18:05 +03:00
|
|
|
if (priority == EventQueuePriority::Input &&
|
|
|
|
|
mInputQueueState == STATE_DISABLED) {
|
|
|
|
|
priority = EventQueuePriority::Normal;
|
2019-02-28 23:38:53 +03:00
|
|
|
} else if (priority == EventQueuePriority::MediumHigh &&
|
2019-05-26 18:00:09 +03:00
|
|
|
!StaticPrefs::threads_medium_high_event_queue_enabled()) {
|
2019-02-28 23:38:53 +03:00
|
|
|
priority = EventQueuePriority::Normal;
|
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
|
|
|
}
|
|
|
|
|
|
|
|
|
|
switch (priority) {
|
2019-01-26 20:18:05 +03:00
|
|
|
case EventQueuePriority::High:
|
2019-11-09 00:07:36 +03:00
|
|
|
mHighQueue->PutEvent(event.forget(), priority, aProofOfLock, aDelay);
|
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
|
|
|
break;
|
2019-01-26 20:18:05 +03:00
|
|
|
case EventQueuePriority::Input:
|
2019-11-09 00:07:36 +03:00
|
|
|
mInputQueue->PutEvent(event.forget(), priority, aProofOfLock, aDelay);
|
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
|
|
|
break;
|
2019-02-28 23:38:53 +03:00
|
|
|
case EventQueuePriority::MediumHigh:
|
2019-11-09 00:07:36 +03:00
|
|
|
mMediumHighQueue->PutEvent(event.forget(), priority, aProofOfLock,
|
|
|
|
|
aDelay);
|
2019-02-28 23:38:53 +03:00
|
|
|
break;
|
2019-01-26 20:18:05 +03:00
|
|
|
case EventQueuePriority::Normal:
|
2019-11-09 00:07:36 +03:00
|
|
|
mNormalQueue->PutEvent(event.forget(), priority, aProofOfLock, aDelay);
|
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
|
|
|
break;
|
2020-03-03 04:47:24 +03:00
|
|
|
case EventQueuePriority::DeferredTimers: {
|
|
|
|
|
if (NS_IsMainThread()) {
|
|
|
|
|
mDeferredTimersQueue->PutEvent(event.forget(), priority, aProofOfLock,
|
|
|
|
|
aDelay);
|
|
|
|
|
} else {
|
|
|
|
|
// We don't want to touch our idle queues from off the main
|
|
|
|
|
// thread. Queue it indirectly.
|
|
|
|
|
IndirectlyQueueRunnable(event.forget(), priority, aProofOfLock, aDelay);
|
|
|
|
|
}
|
2019-01-26 20:18:05 +03:00
|
|
|
break;
|
2020-03-03 04:47:24 +03:00
|
|
|
}
|
|
|
|
|
case EventQueuePriority::Idle: {
|
|
|
|
|
if (NS_IsMainThread()) {
|
|
|
|
|
mIdleQueue->PutEvent(event.forget(), priority, aProofOfLock, aDelay);
|
|
|
|
|
} else {
|
|
|
|
|
// We don't want to touch our idle queues from off the main
|
|
|
|
|
// thread. Queue it indirectly.
|
|
|
|
|
IndirectlyQueueRunnable(event.forget(), priority, aProofOfLock, aDelay);
|
|
|
|
|
}
|
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
|
|
|
break;
|
2020-03-03 04:47:24 +03:00
|
|
|
}
|
2019-01-26 20:18:05 +03:00
|
|
|
case EventQueuePriority::Count:
|
|
|
|
|
MOZ_CRASH("EventQueuePriority::Count isn't a valid priority");
|
2017-11-17 19:30:08 +03:00
|
|
|
break;
|
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
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2019-05-21 21:20:34 +03:00
|
|
|
EventQueuePriority PrioritizedEventQueue::SelectQueue(
|
2017-07-29 00:56:49 +03:00
|
|
|
bool aUpdateState, const MutexAutoLock& aProofOfLock) {
|
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
|
|
|
size_t inputCount = mInputQueue->Count(aProofOfLock);
|
|
|
|
|
|
2017-09-20 22:18:51 +03:00
|
|
|
if (aUpdateState && mInputQueueState == STATE_ENABLED &&
|
|
|
|
|
mInputHandlingStartTime.IsNull() && inputCount > 0) {
|
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
|
|
|
mInputHandlingStartTime =
|
|
|
|
|
InputEventStatistics::Get().GetInputHandlingStartTime(inputCount);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// We check the different queues in the following order. The conditions we use
|
|
|
|
|
// are meant to avoid starvation and to ensure that we don't process an event
|
|
|
|
|
// at the wrong time.
|
|
|
|
|
//
|
2020-02-18 19:03:37 +03:00
|
|
|
// HIGH:
|
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
|
|
|
// INPUT: if inputCount > 0 && TimeStamp::Now() > mInputHandlingStartTime
|
2019-02-28 23:38:53 +03:00
|
|
|
// MEDIUMHIGH: if medium high pending
|
|
|
|
|
// NORMAL: if normal pending
|
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
|
|
|
//
|
|
|
|
|
// If we still don't have an event, then we take events from the queues
|
|
|
|
|
// in the following order:
|
|
|
|
|
// INPUT
|
2019-09-24 17:33:57 +03:00
|
|
|
// DEFERREDTIMERS: if GetLocalIdleDeadline()
|
|
|
|
|
// IDLE: if GetLocalIdleDeadline()
|
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
|
|
|
//
|
|
|
|
|
// If we don't get an event in this pass, then we return null since no events
|
|
|
|
|
// are ready.
|
|
|
|
|
|
|
|
|
|
// This variable determines which queue we will take an event from.
|
2019-01-26 20:18:05 +03:00
|
|
|
EventQueuePriority queue;
|
2020-02-18 19:03:37 +03:00
|
|
|
if (!mHighQueue->IsEmpty(aProofOfLock)) {
|
2019-01-26 20:18:05 +03:00
|
|
|
queue = EventQueuePriority::High;
|
2017-07-28 10:14:54 +03:00
|
|
|
} else if (inputCount > 0 && (mInputQueueState == STATE_FLUSHING ||
|
|
|
|
|
(mInputQueueState == STATE_ENABLED &&
|
2017-09-20 22:18:51 +03:00
|
|
|
!mInputHandlingStartTime.IsNull() &&
|
2017-07-28 10:14:54 +03:00
|
|
|
TimeStamp::Now() > mInputHandlingStartTime))) {
|
2019-01-26 20:18:05 +03:00
|
|
|
queue = EventQueuePriority::Input;
|
2019-02-28 23:38:53 +03:00
|
|
|
} else if (!mMediumHighQueue->IsEmpty(aProofOfLock)) {
|
|
|
|
|
MOZ_ASSERT(
|
|
|
|
|
mInputQueueState != STATE_FLUSHING,
|
|
|
|
|
"Shouldn't consume medium high event when flushing input events");
|
|
|
|
|
queue = EventQueuePriority::MediumHigh;
|
2019-01-26 20:18:05 +03:00
|
|
|
} else if (!mNormalQueue->IsEmpty(aProofOfLock)) {
|
2017-07-28 10:14:54 +03:00
|
|
|
MOZ_ASSERT(mInputQueueState != STATE_FLUSHING,
|
2019-01-26 20:18:05 +03:00
|
|
|
"Shouldn't consume normal event when flushing input events");
|
2019-01-26 20:18:05 +03:00
|
|
|
queue = EventQueuePriority::Normal;
|
2017-07-28 10:14:54 +03:00
|
|
|
} else if (inputCount > 0 && mInputQueueState != STATE_SUSPEND) {
|
|
|
|
|
MOZ_ASSERT(
|
|
|
|
|
mInputQueueState != STATE_DISABLED,
|
|
|
|
|
"Shouldn't consume input events when the input queue is disabled");
|
2019-01-26 20:18:05 +03:00
|
|
|
queue = EventQueuePriority::Input;
|
2019-01-26 20:18:05 +03:00
|
|
|
} else if (!mDeferredTimersQueue->IsEmpty(aProofOfLock)) {
|
|
|
|
|
// We may not actually return an idle event in this case.
|
|
|
|
|
queue = EventQueuePriority::DeferredTimers;
|
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 {
|
|
|
|
|
// We may not actually return an idle event in this case.
|
2019-01-26 20:18:05 +03:00
|
|
|
queue = EventQueuePriority::Idle;
|
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
|
|
|
}
|
|
|
|
|
|
2017-07-28 10:14:54 +03:00
|
|
|
MOZ_ASSERT_IF(
|
2019-01-26 20:18:05 +03:00
|
|
|
queue == EventQueuePriority::Input,
|
2017-07-28 10:14:54 +03:00
|
|
|
mInputQueueState != STATE_DISABLED && mInputQueueState != STATE_SUSPEND);
|
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
|
|
|
|
2017-07-29 00:56:49 +03:00
|
|
|
return queue;
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-03 04:47:24 +03:00
|
|
|
void PrioritizedEventQueue::IndirectlyQueueRunnable(
|
|
|
|
|
already_AddRefed<nsIRunnable>&& aEvent, EventQueuePriority aPriority,
|
|
|
|
|
const MutexAutoLock& aProofOfLock, mozilla::TimeDuration* aDelay) {
|
|
|
|
|
nsCOMPtr<nsIRunnable> event(aEvent);
|
|
|
|
|
|
|
|
|
|
nsCOMPtr<nsIRunnable> mainThreadRunnable = NS_NewRunnableFunction(
|
|
|
|
|
"idle runnable shim",
|
|
|
|
|
[event = std::move(event), priority = aPriority]() mutable {
|
|
|
|
|
NS_DispatchToCurrentThreadQueue(event.forget(), priority);
|
|
|
|
|
});
|
|
|
|
|
mNormalQueue->PutEvent(mainThreadRunnable.forget(),
|
|
|
|
|
EventQueuePriority::Normal, aProofOfLock, aDelay);
|
|
|
|
|
}
|
|
|
|
|
|
2019-11-09 00:07:45 +03:00
|
|
|
// The delay returned is the queuing delay a hypothetical Input event would
|
|
|
|
|
// see due to the current running event if it had arrived while the current
|
|
|
|
|
// event was queued. This means that any event running at priority below
|
|
|
|
|
// Input doesn't cause queuing delay for Input events, and we return
|
|
|
|
|
// TimeDuration() for those cases.
|
2019-05-21 21:20:34 +03:00
|
|
|
already_AddRefed<nsIRunnable> PrioritizedEventQueue::GetEvent(
|
2019-11-09 00:07:36 +03:00
|
|
|
EventQueuePriority* aPriority, const MutexAutoLock& aProofOfLock,
|
2019-11-22 17:03:23 +03:00
|
|
|
TimeDuration* aHypotheticalInputEventDelay) {
|
|
|
|
|
MOZ_ASSERT_UNREACHABLE("Who is managing to call this?");
|
|
|
|
|
bool ignored;
|
|
|
|
|
return GetEvent(aPriority, aProofOfLock, aHypotheticalInputEventDelay,
|
|
|
|
|
&ignored);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
already_AddRefed<nsIRunnable> PrioritizedEventQueue::GetEvent(
|
|
|
|
|
EventQueuePriority* aPriority, const MutexAutoLock& aProofOfLock,
|
|
|
|
|
TimeDuration* aHypotheticalInputEventDelay, bool* aIsIdleEvent) {
|
2019-01-26 20:18:05 +03:00
|
|
|
EventQueuePriority queue = SelectQueue(true, aProofOfLock);
|
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
|
|
|
|
|
|
|
|
if (aPriority) {
|
|
|
|
|
*aPriority = queue;
|
|
|
|
|
}
|
2019-11-22 17:03:23 +03:00
|
|
|
*aIsIdleEvent = false;
|
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
|
|
|
|
2019-11-09 00:07:36 +03:00
|
|
|
// Since Input events will only be delayed behind Input or High events,
|
|
|
|
|
// the amount of time a lower-priority event spent in the queue is
|
|
|
|
|
// irrelevant in knowing how long an input event would be delayed.
|
|
|
|
|
// Alternatively, we could export the delay and let the higher-level code
|
|
|
|
|
// key off the returned priority level (though then it'd need to know
|
|
|
|
|
// if the thread's queue was a PrioritizedEventQueue or normal/other
|
|
|
|
|
// EventQueue).
|
|
|
|
|
nsCOMPtr<nsIRunnable> event;
|
|
|
|
|
switch (queue) {
|
|
|
|
|
default:
|
|
|
|
|
MOZ_CRASH();
|
|
|
|
|
break;
|
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
|
|
|
|
2019-11-09 00:07:36 +03:00
|
|
|
case EventQueuePriority::High:
|
2020-04-14 01:22:57 +03:00
|
|
|
event = mHighQueue->GetEvent(nullptr, aProofOfLock,
|
2019-11-09 00:07:45 +03:00
|
|
|
aHypotheticalInputEventDelay);
|
2019-11-09 00:07:36 +03:00
|
|
|
MOZ_ASSERT(event);
|
|
|
|
|
mInputHandlingStartTime = TimeStamp();
|
|
|
|
|
break;
|
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
|
|
|
|
2019-11-09 00:07:36 +03:00
|
|
|
case EventQueuePriority::Input:
|
2020-04-14 01:22:57 +03:00
|
|
|
event = mInputQueue->GetEvent(nullptr, aProofOfLock,
|
2019-11-09 00:07:45 +03:00
|
|
|
aHypotheticalInputEventDelay);
|
2019-11-09 00:07:36 +03:00
|
|
|
MOZ_ASSERT(event);
|
|
|
|
|
break;
|
2019-11-07 15:53:28 +03:00
|
|
|
|
2019-11-09 00:07:36 +03:00
|
|
|
// All queue priorities below Input don't add their queuing time to the
|
|
|
|
|
// time an input event will be delayed, so report 0 for time-in-queue
|
|
|
|
|
// if we're below Input; input events will only be delayed by the time
|
|
|
|
|
// an event actually runs (if the event is below Input event's priority)
|
|
|
|
|
case EventQueuePriority::MediumHigh:
|
2020-04-14 01:22:57 +03:00
|
|
|
event = mMediumHighQueue->GetEvent(nullptr, aProofOfLock);
|
2019-11-09 00:07:45 +03:00
|
|
|
*aHypotheticalInputEventDelay = TimeDuration();
|
2019-11-09 00:07:36 +03:00
|
|
|
break;
|
2019-11-07 15:53:28 +03:00
|
|
|
|
2019-11-09 00:07:36 +03:00
|
|
|
case EventQueuePriority::Normal:
|
2020-04-14 01:22:57 +03:00
|
|
|
event = mNormalQueue->GetEvent(nullptr, aProofOfLock);
|
2019-11-09 00:07:45 +03:00
|
|
|
*aHypotheticalInputEventDelay = TimeDuration();
|
2019-11-09 00:07:36 +03:00
|
|
|
break;
|
2019-11-07 15:53:28 +03:00
|
|
|
|
2019-11-09 00:07:36 +03:00
|
|
|
case EventQueuePriority::Idle:
|
|
|
|
|
case EventQueuePriority::DeferredTimers:
|
2019-11-09 00:07:45 +03:00
|
|
|
*aHypotheticalInputEventDelay = TimeDuration();
|
2019-11-09 00:07:36 +03:00
|
|
|
// If we get here, then all queues except deferredtimers and idle are
|
|
|
|
|
// empty.
|
|
|
|
|
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
if (!HasIdleRunnables(aProofOfLock)) {
|
2019-11-09 00:07:36 +03:00
|
|
|
return nullptr;
|
|
|
|
|
}
|
|
|
|
|
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
TimeStamp idleDeadline = mIdlePeriodState.GetCachedIdleDeadline();
|
2019-11-09 00:07:36 +03:00
|
|
|
if (!idleDeadline) {
|
|
|
|
|
return nullptr;
|
|
|
|
|
}
|
|
|
|
|
|
2020-04-14 01:22:57 +03:00
|
|
|
event = mDeferredTimersQueue->GetEvent(nullptr, aProofOfLock);
|
2019-11-09 00:07:36 +03:00
|
|
|
if (!event) {
|
2020-04-14 01:22:57 +03:00
|
|
|
event = mIdleQueue->GetEvent(nullptr, aProofOfLock);
|
2019-11-09 00:07:36 +03:00
|
|
|
}
|
|
|
|
|
if (event) {
|
2019-11-22 17:03:23 +03:00
|
|
|
*aIsIdleEvent = true;
|
2019-11-09 00:07:36 +03:00
|
|
|
nsCOMPtr<nsIIdleRunnable> idleEvent = do_QueryInterface(event);
|
|
|
|
|
if (idleEvent) {
|
|
|
|
|
idleEvent->SetDeadline(idleDeadline);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
break;
|
2019-11-09 00:07:45 +03:00
|
|
|
} // switch (queue)
|
|
|
|
|
|
|
|
|
|
if (!event) {
|
|
|
|
|
*aHypotheticalInputEventDelay = TimeDuration();
|
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
|
|
|
}
|
|
|
|
|
|
2020-04-14 01:22:57 +03:00
|
|
|
MOZ_ASSERT_IF(aPriority, *aPriority == queue);
|
|
|
|
|
|
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
|
|
|
return event.forget();
|
|
|
|
|
}
|
|
|
|
|
|
2019-09-24 17:33:57 +03:00
|
|
|
void PrioritizedEventQueue::DidRunEvent(const MutexAutoLock& aProofOfLock) {
|
|
|
|
|
if (IsEmpty(aProofOfLock)) {
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
// Certainly no more idle tasks. Unlocking here is OK, because
|
|
|
|
|
// our caller does nothing after this except return, which unlocks
|
|
|
|
|
// its lock anyway.
|
|
|
|
|
MutexAutoUnlock unlock(*mMutex);
|
|
|
|
|
mIdlePeriodState.RanOutOfTasks(unlock);
|
2019-09-24 17:33:57 +03:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2019-05-25 20:46:15 +03:00
|
|
|
bool PrioritizedEventQueue::IsEmpty(const MutexAutoLock& aProofOfLock) {
|
2017-07-29 00:56:49 +03:00
|
|
|
// Just check IsEmpty() on the sub-queues. Don't bother checking the idle
|
|
|
|
|
// deadline since that only determines whether an idle event is ready or not.
|
|
|
|
|
return mHighQueue->IsEmpty(aProofOfLock) &&
|
|
|
|
|
mInputQueue->IsEmpty(aProofOfLock) &&
|
2019-02-28 23:38:53 +03:00
|
|
|
mMediumHighQueue->IsEmpty(aProofOfLock) &&
|
2017-07-29 00:56:49 +03:00
|
|
|
mNormalQueue->IsEmpty(aProofOfLock) &&
|
2019-01-26 20:18:05 +03:00
|
|
|
mDeferredTimersQueue->IsEmpty(aProofOfLock) &&
|
2017-07-29 00:56:49 +03:00
|
|
|
mIdleQueue->IsEmpty(aProofOfLock);
|
|
|
|
|
}
|
|
|
|
|
|
2019-05-25 20:46:15 +03:00
|
|
|
bool PrioritizedEventQueue::HasReadyEvent(const MutexAutoLock& aProofOfLock) {
|
2019-10-20 18:08:44 +03:00
|
|
|
mIdlePeriodState.ForgetPendingTaskGuarantee();
|
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
|
|
|
|
2019-01-26 20:18:05 +03:00
|
|
|
EventQueuePriority queue = SelectQueue(false, aProofOfLock);
|
2017-07-29 00:56:49 +03:00
|
|
|
|
2019-01-26 20:18:05 +03:00
|
|
|
if (queue == EventQueuePriority::High) {
|
2017-07-29 00:56:49 +03:00
|
|
|
return mHighQueue->HasReadyEvent(aProofOfLock);
|
2019-01-26 20:18:05 +03:00
|
|
|
} else if (queue == EventQueuePriority::Input) {
|
2017-09-08 23:57:29 +03:00
|
|
|
return mInputQueue->HasReadyEvent(aProofOfLock);
|
2019-02-28 23:38:53 +03:00
|
|
|
} else if (queue == EventQueuePriority::MediumHigh) {
|
|
|
|
|
return mMediumHighQueue->HasReadyEvent(aProofOfLock);
|
2019-01-26 20:18:05 +03:00
|
|
|
} else if (queue == EventQueuePriority::Normal) {
|
2017-07-29 00:56:49 +03:00
|
|
|
return mNormalQueue->HasReadyEvent(aProofOfLock);
|
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
|
|
|
}
|
|
|
|
|
|
2019-01-26 20:18:05 +03:00
|
|
|
MOZ_ASSERT(queue == EventQueuePriority::Idle ||
|
|
|
|
|
queue == EventQueuePriority::DeferredTimers);
|
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
|
|
|
|
2017-07-29 00:56:49 +03:00
|
|
|
// If we get here, then both the high and normal queues are empty.
|
|
|
|
|
|
2019-01-26 20:18:05 +03:00
|
|
|
if (mDeferredTimersQueue->IsEmpty(aProofOfLock) &&
|
|
|
|
|
mIdleQueue->IsEmpty(aProofOfLock)) {
|
2017-07-29 00:56:49 +03:00
|
|
|
return false;
|
|
|
|
|
}
|
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
|
|
|
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
// Temporarily unlock so we can peek our idle deadline.
|
|
|
|
|
{
|
|
|
|
|
MutexAutoUnlock unlock(*mMutex);
|
2019-12-31 18:15:17 +03:00
|
|
|
mIdlePeriodState.CachePeekedIdleDeadline(unlock);
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
}
|
2019-12-31 18:15:17 +03:00
|
|
|
TimeStamp idleDeadline = mIdlePeriodState.GetCachedIdleDeadline();
|
|
|
|
|
mIdlePeriodState.ClearCachedIdleDeadline();
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
|
|
|
|
|
// Re-check the emptiness of the queues, since we had the lock released for a
|
|
|
|
|
// bit.
|
2019-10-20 18:08:44 +03:00
|
|
|
if (idleDeadline && (mDeferredTimersQueue->HasReadyEvent(aProofOfLock) ||
|
|
|
|
|
mIdleQueue->HasReadyEvent(aProofOfLock))) {
|
|
|
|
|
mIdlePeriodState.EnforcePendingTaskGuarantee();
|
|
|
|
|
return true;
|
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
|
|
|
}
|
|
|
|
|
|
2017-07-29 00:56:49 +03:00
|
|
|
return false;
|
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
|
|
|
}
|
|
|
|
|
|
2019-05-21 21:20:34 +03:00
|
|
|
bool PrioritizedEventQueue::HasPendingHighPriorityEvents(
|
2019-01-10 20:07:34 +03:00
|
|
|
const MutexAutoLock& aProofOfLock) {
|
|
|
|
|
return !mHighQueue->IsEmpty(aProofOfLock);
|
|
|
|
|
}
|
|
|
|
|
|
2019-05-25 20:46:15 +03:00
|
|
|
size_t PrioritizedEventQueue::Count(const MutexAutoLock& aProofOfLock) const {
|
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
|
|
|
MOZ_CRASH("unimplemented");
|
|
|
|
|
}
|
|
|
|
|
|
2019-05-21 21:20:34 +03:00
|
|
|
void PrioritizedEventQueue::EnableInputEventPrioritization(
|
2017-07-28 10:14:54 +03:00
|
|
|
const MutexAutoLock& aProofOfLock) {
|
|
|
|
|
MOZ_ASSERT(mInputQueueState == STATE_DISABLED);
|
|
|
|
|
mInputQueueState = STATE_ENABLED;
|
|
|
|
|
mInputHandlingStartTime = TimeStamp();
|
|
|
|
|
}
|
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
|
|
|
|
2019-05-21 21:20:34 +03:00
|
|
|
void PrioritizedEventQueue::FlushInputEventPrioritization(
|
2017-07-28 10:14:54 +03:00
|
|
|
const MutexAutoLock& aProofOfLock) {
|
|
|
|
|
MOZ_ASSERT(mInputQueueState == STATE_ENABLED ||
|
|
|
|
|
mInputQueueState == STATE_SUSPEND);
|
|
|
|
|
mInputQueueState =
|
|
|
|
|
mInputQueueState == STATE_ENABLED ? STATE_FLUSHING : STATE_SUSPEND;
|
|
|
|
|
}
|
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
|
|
|
|
2019-05-21 21:20:34 +03:00
|
|
|
void PrioritizedEventQueue::SuspendInputEventPrioritization(
|
2017-07-28 10:14:54 +03:00
|
|
|
const MutexAutoLock& aProofOfLock) {
|
|
|
|
|
MOZ_ASSERT(mInputQueueState == STATE_ENABLED ||
|
|
|
|
|
mInputQueueState == STATE_FLUSHING);
|
|
|
|
|
mInputQueueState = STATE_SUSPEND;
|
|
|
|
|
}
|
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
|
|
|
|
2019-05-21 21:20:34 +03:00
|
|
|
void PrioritizedEventQueue::ResumeInputEventPrioritization(
|
2017-07-28 10:14:54 +03:00
|
|
|
const MutexAutoLock& aProofOfLock) {
|
|
|
|
|
MOZ_ASSERT(mInputQueueState == STATE_SUSPEND);
|
|
|
|
|
mInputQueueState = STATE_ENABLED;
|
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
|
|
|
}
|
2019-11-22 17:06:29 +03:00
|
|
|
|
|
|
|
|
bool PrioritizedEventQueue::HasIdleRunnables(
|
Bug 1597158 part 4. Stop unlocking the provided mutex in IdlePeriodState methods. r=smaug
The basic idea, suggested by Olli, is that we can try to get a runnable in
ThreadEventQueue::GetEvent, and if that does not produce anything unlock our
mutex, do whatever idle state updates we need to do, re-lock our mutex. Then
always we need to try getting a runnable again, because a non-idle runnable
might have gotten queued while we had the lock unlocked. So we can't sleep on
our mutex, in the mayWait case, unless we try to get a runnable again first.
My notes on the current (pre this patch) unlocking setup follow.
------------------------------------------------------------
There are four places where we currently unlock:
1) IdlePeriodState::GetIdleDeadlineInternal. Needed only when !aIsPeek, to
RequestIdleToken, which can do IPC. The only caller, via
GetDeadlineForIdleTask, is PrioritizedEventQueue::GetEvent and only when we
selected the idle or deferred queue. We need this to set the proper deadline
on the idle event. In the cases when this unlock happens we currently _never_
return an idle event, because if we got here that means that we do not have an
idle token.
2) IdlePeriodState::GetLocalIdleDeadline. Needs to unlock to get the idle
period hint. The can get called from GetIdleDeadlineInternal in _both_ cases:
peek and get. The callstack for the get case is covered above. The peek case
is called from PrioritizedEventQueue::HasReadyEvent which is called from
ThreadEventQueue::HasPendingEvent.
3) IdlePeriodState::SetPaused, because it sends an IPC message. This is only
called from EnsureIsPaused, which is called from:
- IdlePeriodState::GetIdleDeadlineInternal. Only in the !aIsPeek case.
- IdlePeriodState::RanOutOfTasks called from:
- PrioritizedEventQueue::GetEvent if we fell into the idle case and our
queues are empty.
- PrioritizedEventQueue::DidRunEvent if we are empty.
4) IdlePeriodState::ClearIdleToken because it sends an IPC message. This is
called from:
- IdlePeriodState::RanOutOfTasks; see SetPaused.
- IdlePeriodState::GetIdleDeadlineInternal like EnsureIsPaused.
- IdlePeriodState::GetIdleToken if token is in the past. This is only
called from GetIdleDeadlineInternal, both cases.
- IdlePeriodState::FlagNotIdle called from PrioritizedEventQueue::GetEvent
if we find an event in a non-idle queue.
Or rewriting in terms of API entrypoints on IdlePeriodState that might need to
unlock:
* Anything to do with getting deadlines, whether we are peeking or getting.
Basically, if we need an updated deadline we need to unlock.
* When we have detected we are completely out of tasks (idle or not) to run.
Right now we do that when either we're asked for an event and don't have one
or if we run an event and are empty after that (before unlocking!). But the
unlocking or not happens in nsThreadEventQueue::DidRunEvent, so separately
from the getting of the event. In particular, we are unlocked before we
enter DidRunEvent, and unlock again before we return from it, so we can do
whatever updates we want there.
* When we have detected that we have a non-idle event to run; this calls
FlagNotIdle.
Differential Revision: https://phabricator.services.mozilla.com/D53631
--HG--
extra : moz-landing-system : lando
2019-11-22 17:06:17 +03:00
|
|
|
const MutexAutoLock& aProofOfLock) const {
|
2019-11-22 17:06:29 +03:00
|
|
|
return !mIdleQueue->IsEmpty(aProofOfLock) ||
|
|
|
|
|
!mDeferredTimersQueue->IsEmpty(aProofOfLock);
|
|
|
|
|
}
|
