[core] Use an actor model for tile worker concurrency

src/mbgl/actor/actor.hpp

@@ -0,0 +1,76 @@
+#pragma once
+
+#include <mbgl/actor/mailbox.hpp>
+#include <mbgl/actor/message.hpp>
+#include <mbgl/actor/actor_ref.hpp>
+#include <mbgl/util/noncopyable.hpp>
+
+#include <memory>
+
+namespace mbgl {
+
+/*
+    An `Actor<O>` is an owning reference to an asynchronous object of type `O`: an "actor".
+    Communication with an actor happens via message passing: you send a message to the object
+    (using `invoke`), passing a pointer to the member function to call and arguments which
+    are then forwarded to the actor.
+
+    The actor receives messages sent to it asynchronously, in a manner defined its `Scheduler`.
+    To store incoming messages before their receipt, each actor has a `Mailbox`, which acts as
+    a FIFO queue. Messages sent from actor S to actor R are guaranteed to be processed in the
+    order sent. However, relative order of messages sent by two *different* actors S1 and S2
+    to R is *not* guaranteed (and can't be: S1 and S2 may be acting asynchronously with respect
+    to each other).
+
+    Construction and destruction of an actor is currently synchronous: the corresponding `O`
+    object is constructed synchronously by the `Actor` constructor, and destructed synchronously
+    by the `~Actor` destructor, after ensuring that the `O` is not currently receiving an
+    asynchronous message. (Construction and destruction may change to be asynchronous in the
+    future.)
+
+    An `Actor<O>` can be converted to an `ActorRef<O>`, a non-owning value object representing
+    a (weak) reference to the actor. Messages can be sent via the `Ref` as well.
+
+    It's safe -- and encouraged -- to pass `Ref`s between actors via messages. This is how two-way
+    communication and other forms of collaboration between multiple actors is accomplished.
+
+    It's safe for a `Ref` to outlive its `Actor` -- the reference is "weak", and does not extend
+    the lifetime of the owning Actor, and sending a message to a `Ref` whose `Actor` has died is
+    a no-op. (In the future, a dead-letters queue or log may be implemented.)
+
+    Please don't send messages that contain shared pointers or references. That subverts the
+    purpose of the actor model: prohibiting direct concurrent access to shared state.
+*/
+
+template <class Object>
+class Actor : public util::noncopyable {
+public:
+    template <class... Args>
+    Actor(Scheduler& scheduler, Args&&... args_)
+        : mailbox(std::make_shared<Mailbox>(scheduler)),
+          object(self(), std::forward<Args>(args_)...) {
+    }
+
+    ~Actor() {
+        mailbox->close();
+    }
+
+    template <typename Fn, class... Args>
+    void invoke(Fn fn, Args&&... args) {
+        mailbox->push(actor::makeMessage(object, fn, std::forward<Args>(args)...));
+    }
+
+    ActorRef<std::decay_t<Object>> self() {
+        return ActorRef<std::decay_t<Object>>(object, mailbox);
+    }
+
+    operator ActorRef<std::decay_t<Object>>() {
+        return self();
+    }
+
+private:
+    std::shared_ptr<Mailbox> mailbox;
+    Object object;
+};
+
+} // namespace mbgl

src/mbgl/actor/actor_ref.hpp

@@ -0,0 +1,43 @@
+#pragma once
+
+#include <mbgl/actor/mailbox.hpp>
+#include <mbgl/actor/message.hpp>
+
+#include <memory>
+
+namespace mbgl {
+
+/*
+    An `ActorRef<O>` is a *non*-owning, weak reference to an actor of type `O`. You can send it
+    messages just like an `Actor<O>`. It's a value object: safe to copy and pass between actors
+    via messages.
+
+    An `ActorRef<O>` does not extend the lifetime of the corresponding `Actor<O>`. That's determined
+    entirely by whichever object owns the `Actor<O>` -- the actor's "supervisor".
+
+    It's safe for a `Ref` to outlive its `Actor` -- the reference is "weak", and does not extend
+    the lifetime of the owning Actor, and sending a message to a `Ref` whose `Actor` has died is
+    a no-op. (In the future, a dead-letters queue or log may be implemented.)
+*/
+
+template <class Object>
+class ActorRef {
+public:
+    ActorRef(Object& object_, std::weak_ptr<Mailbox> weakMailbox_)
+        : object(object_),
+          weakMailbox(std::move(weakMailbox_)) {
+    }
+
+    template <typename Fn, class... Args>
+    void invoke(Fn fn, Args&&... args) {
+        if (auto mailbox = weakMailbox.lock()) {
+            mailbox->push(actor::makeMessage(object, fn, std::forward<Args>(args)...));
+        }
+    }
+
+private:
+    Object& object;
+    std::weak_ptr<Mailbox> weakMailbox;
+};
+
+} // namespace mbgl

src/mbgl/actor/mailbox.cpp

@@ -0,0 +1,55 @@
+#include <mbgl/actor/mailbox.hpp>
+#include <mbgl/actor/message.hpp>
+#include <mbgl/actor/scheduler.hpp>
+
+#include <cassert>
+
+namespace mbgl {
+
+Mailbox::Mailbox(Scheduler& scheduler_)
+    : scheduler(scheduler_) {
+}
+
+void Mailbox::push(std::unique_ptr<Message> message) {
+    assert(!closing);
+
+    std::lock_guard<std::mutex> queueLock(queueMutex);
+    bool wasEmpty = queue.empty();
+    queue.push(std::move(message));
+    if (wasEmpty) {
+        scheduler.schedule(shared_from_this());
+    }
+}
+
+void Mailbox::close() {
+    // Block until the scheduler is guaranteed not to be executing receive().
+    std::lock_guard<std::mutex> closingLock(closingMutex);
+    closing = true;
+}
+
+void Mailbox::receive() {
+    std::lock_guard<std::mutex> closingLock(closingMutex);
+
+    if (closing) {
+        return;
+    }
+
+    std::unique_ptr<Message> message;
+    bool wasEmpty;
+
+    {
+        std::lock_guard<std::mutex> queueLock(queueMutex);
+        assert(!queue.empty());
+        message = std::move(queue.front());
+        queue.pop();
+        wasEmpty = queue.empty();
+    }
+
+    (*message)();
+
+    if (!wasEmpty) {
+        scheduler.schedule(shared_from_this());
+    }
+}
+
+} // namespace mbgl

src/mbgl/actor/mailbox.hpp

@@ -0,0 +1,31 @@
+#pragma once
+
+#include <memory>
+#include <mutex>
+#include <queue>
+
+namespace mbgl {
+
+class Scheduler;
+class Message;
+
+class Mailbox : public std::enable_shared_from_this<Mailbox> {
+public:
+    Mailbox(Scheduler&);
+
+    void push(std::unique_ptr<Message>);
+
+    void close();
+    void receive();
+
+private:
+    Scheduler& scheduler;
+
+    std::mutex closingMutex;
+    bool closing { false };
+
+    std::mutex queueMutex;
+    std::queue<std::unique_ptr<Message>> queue;
+};
+
+} // namespace mbgl

src/mbgl/actor/message.hpp

@@ -0,0 +1,48 @@
+#pragma once
+
+#include <utility>
+
+namespace mbgl {
+
+// A movable type-erasing function wrapper. This allows to store arbitrary invokable
+// things (like std::function<>, or the result of a movable-only std::bind()) in the queue.
+// Source: http://stackoverflow.com/a/29642072/331379
+class Message {
+public:
+    virtual ~Message() = default;
+    virtual void operator()() = 0;
+};
+
+template <class Object, class MemberFn, class ArgsTuple>
+class MessageImpl : public Message {
+public:
+    MessageImpl(Object& object_, MemberFn memberFn_, ArgsTuple argsTuple_)
+      : object(object_),
+        memberFn(memberFn_),
+        argsTuple(std::move(argsTuple_)) {
+    }
+
+    void operator()() override {
+        invoke(std::make_index_sequence<std::tuple_size<ArgsTuple>::value>());
+    }
+
+    template <std::size_t... I>
+    void invoke(std::index_sequence<I...>) {
+        (object.*memberFn)(std::move(std::get<I>(argsTuple))...);
+    }
+
+    Object& object;
+    MemberFn memberFn;
+    ArgsTuple argsTuple;
+};
+
+namespace actor {
+
+template <class Object, class MemberFn, class... Args>
+std::unique_ptr<Message> makeMessage(Object& object, MemberFn memberFn, Args&&... args) {
+    auto tuple = std::make_tuple(std::forward<Args>(args)...);
+    return std::make_unique<MessageImpl<Object, MemberFn, decltype(tuple)>>(object, memberFn, std::move(tuple));
+}
+
+} // namespace actor
+} // namespace mbgl

src/mbgl/actor/scheduler.hpp

@@ -0,0 +1,38 @@
+#pragma once
+
+#include <memory>
+
+namespace mbgl {
+
+class Mailbox;
+
+/*
+    A `Scheduler` is responsible for coordinating the processing of messages by
+    one or more actors via their mailboxes. It's an abstract interface. Currently,
+    the following concrete implementations exist:
+
+    * `ThreadPool` can coordinate an unlimited number of actors over any number of
+      threads via a pool, preserving the following behaviors:
+
+      - Messages from each individual mailbox are processed in order
+      - Only a single message from a mailbox is processed at a time; there is no
+        concurrency within a mailbox
+
+      Subject to these constraints, processing can happen on whatever thread in the
+      pool is available.
+
+    * `RunLoop` is a `Scheduler` that is typically used to create a mailbox and
+      `ActorRef` for an object that lives on the main thread and is not itself wrapped
+      as an `Actor`:
+
+        auto mailbox = std::make_shared<Mailbox>(*util::RunLoop::Get());
+        Actor<Worker> worker(threadPool, ActorRef<Foo>(*this, mailbox));
+*/
+
+class Scheduler {
+public:
+    virtual ~Scheduler() = default;
+    virtual void schedule(std::weak_ptr<Mailbox>) = 0;
+};
+
+} // namespace mbgl

src/mbgl/actor/thread_pool.cpp

@@ -0,0 +1,48 @@
+#include <mbgl/actor/thread_pool.hpp>
+#include <mbgl/actor/mailbox.hpp>
+
+namespace mbgl {
+
+ThreadPool::ThreadPool(std::size_t count) {
+    threads.reserve(count);
+    for (std::size_t i = 0; i < count; ++i) {
+        threads.emplace_back([this] () {
+            while (true) {
+                std::unique_lock<std::mutex> lock(mutex);
+
+                cv.wait(lock, [this] {
+                    return !queue.empty() || terminate.load();
+                });
+
+                if (terminate.load()) {
+                    return;
+                }
+
+                auto mailbox = queue.front();
+                queue.pop();
+                lock.unlock();
+
+                if (auto locked = mailbox.lock()) {
+                    locked->receive();
+                }
+            }
+        });
+    }
+}
+
+ThreadPool::~ThreadPool() {
+    terminate.store(true);
+    cv.notify_all();
+
+    for (auto& thread : threads) {
+        thread.join();
+    }
+}
+
+void ThreadPool::schedule(std::weak_ptr<Mailbox> mailbox) {
+    std::lock_guard<std::mutex> lock(mutex);
+    queue.push(mailbox);
+    cv.notify_one();
+}
+
+} // namespace mbgl

src/mbgl/actor/thread_pool.hpp

@@ -0,0 +1,28 @@
+#pragma once
+
+#include <mbgl/actor/scheduler.hpp>
+
+#include <atomic>
+#include <condition_variable>
+#include <mutex>
+#include <queue>
+#include <thread>
+
+namespace mbgl {
+
+class ThreadPool : public Scheduler {
+public:
+    ThreadPool(std::size_t count);
+    ~ThreadPool() override;
+
+    void schedule(std::weak_ptr<Mailbox>) override;
+
+private:
+    std::vector<std::thread> threads;
+    std::queue<std::weak_ptr<Mailbox>> queue;
+    std::mutex mutex;
+    std::condition_variable cv;
+    std::atomic_bool terminate { false };
+};
+
+} // namespace mbgl