C++11多线程之std::unique_lock

奇怪,几年前写的,不知道为何最近一年始终放在草稿箱了。拿出来晒晒。如有错漏之处,再修正吧。

原文如下:
http://en.cppreference.com/w/cpp/thread/unique_lock
http://en.cppreference.com/w/cpp/thread/unique_lock/unique_lock
http://en.cppreference.com/w/cpp/thread/unique_lock/~unique_lock
http://en.cppreference.com/w/cpp/thread/unique_lock/operator%3D
http://en.cppreference.com/w/cpp/thread/unique_lock/lock
http://en.cppreference.com/w/cpp/thread/unique_lock/try_lock
http://en.cppreference.com/w/cpp/thread/unique_lock/try_lock_for
http://en.cppreference.com/w/cpp/thread/unique_lock/try_lock_until
http://en.cppreference.com/w/cpp/thread/unique_lock/unlock
http://en.cppreference.com/w/cpp/thread/unique_lock/mutex
http://en.cppreference.com/w/cpp/thread/unique_lock/owns_lock
http://en.cppreference.com/w/cpp/thread/unique_lock/operator_bool

#std::unique_lock

定义在头文件 中。

template       (since C++11)
class unique_lock;

类 unique_lock 是一个一般性质的 mutex 属主的封装,提供延迟锁定(deferred locking),限时尝试(time-constrained attempts),递归锁定(recursive locking), 锁主的转换, 以及对条件变量的使用。

类 unique_lock 是 movable 的,但不是 copyable 的 – 它满足 MoveConstructible 和 MoveAssignable, 但不满足 CopyConstructible 和 CopyAssignable.

类 unique_lock 满足 BasicLockable 的要求。如果 Mutex 满足 Lockable 的要求,那么 unique_lock 也满足 Lockable 的要求(比如,可以被用于 std::lock);如果 Mutex 满足 TimedLockable 的要求,unique_lock 也满足 TimedLockable 的要求。

模板参数

Mutex - 用于锁定的 mutex 的类型。该类型必须满足 BasicLockable 的要求。

构造函数

(1) unique_lock();  
(2) unique_lock( unique_lock&& other );  
(3) explicit unique_lock( mutex_type& m );
(4) unique_lock( mutex_type& m, std::defer_lock_t t );
(5) unique_lock( mutex_type& m, std::try_to_lock_t t );
(6) unique_lock( mutex_type& m, std::adopt_lock_t t );
(7) template< class Rep, class Period >
    unique_lock(mutex_type& m, 
                const std::chrono::duration& timeout_duration);
(8) template< class Clock, class Duration >
    unique_lock(mutex_type& m, 
                const std::chrono::time_point& timeout_time);

(1) unique_lock();
构造一个没有关联 mutex 的 unique_lock

(2) unique_lock( unique_lock&& other );
Move构造函数,使用 other 的内容来构造 unique_lock. 使得other变成没有mutex关联的unique_lock.

(3) - (8) 构造一个以 m 为关联的mutex的unique_lock, 另外:

(3) explicit unique_lock( mutex_type& m );
通过调用 m.lock() 来锁定相关联的 mutex. 如果当前线程已经拥有了mutex,且不是递归的mutex,那么行为未定义。

(4) unique_lock( mutex_type& m, std::defer_lock_t t );
不锁定相关的mutex.

(5) unique_lock( mutex_type& m, std::try_to_lock_t t );
通过调用 m.try_lock() 来尝试锁定相关的mutex而不会阻塞。如果当前线程已经拥有mutex且不是递归mutex,则行为未定义。

(6) unique_lock( mutex_type& m, std::adopt_lock_t t );
假设线程已经拥有m.

(7)

template< class Rep, class Period >  
    unique_lock(mutex_type& m, const std::chrono::duration& timeout_duration);  

通过调用 m.try_lock_for(timeout_duration) 试图锁定相关联的 mutex. 一直阻塞直到超时或锁定成功。也可能阻塞得比time_duration的时间更长一些。

(8)

template< class Clock, class Duration >  
    unique_lock(mutex_type& m, const std::chrono::time_point& timeout_time);  

通过调用 m.try_lock_until(timeout_time) 来试图锁定相关联的 mutex. 一直阻塞直到指定的时间点到达或者锁定成功。可能会在指定的时间到达后仍阻塞一会儿。

示例程序:

#include 
#include  // std::cout
#include 
#include 
#include 
 
class Number;
std::ostream& operator<<(std::ostream& stream, const Number& number);
 
class Number {
 public:
  Number() : v_(1) {}
 
  // thread-safe update of 'a' and 'b'
  static void update(Number& a, Number& b, bool order) {
    // do not lock 'mutex_' of 'a' and 'b' sequentially,
    // two sequential lock may lead to deadlock,
    // that's why 'std::lock' exists (see below)
    GuardLock lock_a(a.mutex_, std::defer_lock);
    GuardLock lock_b(b.mutex_, std::defer_lock);
 
    // mutexes is not locked
    assert(!lock_a.owns_lock());
    assert(!lock_b.owns_lock());
 
    // unspecified series of calls...
    std::lock(lock_a, lock_b);
 
    // Result: 'a.mutex_' and 'b.mutex_' is in locked state
    // 'a' and 'b' can be modified safety
    assert(lock_a.owns_lock());
    assert(lock_b.owns_lock());
 
    if (order) {
      a.v_ += b.v_;
      b.v_ += a.v_;
 
      std::cout << a << b;
    }
    else {
      b.v_ += a.v_;
      a.v_ += b.v_;
 
      std::cout << b << a;
    }
 
    // 'lock_a' and 'lock_b' will be destroyed,
    // unlocking 'a.mutex_' and 'b.mutex_'
  }
 
  // not thread-safe; used before thread creation or in thread-safe 'update'
  std::ostream& print(std::ostream& stream) const {
    stream << v_ << " ";
    return stream;
  }
 
 private:
  using Mutex = std::mutex;
  using GuardLock = std::unique_lock;
 
  Mutex mutex_;
  int v_;
};
 
// not thread-safe; see 'Number::print'
std::ostream& operator<<(std::ostream& stream, const Number& number) {
  return number.print(stream);
}
 
int main() {
  Number a, b;
  std::cout << a << b;
 
  std::vector threads;
 
  for (unsigned i = 0; i < 4; ++i) {
    // without 'std::lock' deadlock may occur in this situation:
    //   thread #1 lock 'a.mutex_'
    //   thread #2 lock 'b.mutex_'
    //   thread #1 try to lock 'b.mutex_' and blocked (it's locked by #2)
    //   thread #2 try to lock 'a.mutex_' and blocked (it's locked by #1)
    //   ... deadlock
    threads.emplace_back(Number::update, std::ref(a), std::ref(b), true); // #1
    threads.emplace_back(Number::update, std::ref(b), std::ref(a), false); // #2
  }
 
  for (auto& i: threads) {
    i.join();
  }
 
  std::cout << '\n';
}

// Output: 
// 1 1 2 3 5 8 13 21 34 55 89 144 233 377 610 987 1597 2584

析构函数

销毁这个unique_lock对象。如果 *this 此时拥有一个相关联的 mutex 并且已经获得它,那么就会解锁该mutex.

赋值操作符 operator =

unique_lock& operator=( unique_lock&& other ); (since C++11)

Move赋值操作符。使用 other 中的内容来赋值给自己。
如果在此调用前,*this 已经拥有一个mutex并锁定了它,那么此调用会解锁该mutex.

std::unique_lock::lock 函数

void lock(); (since C++11)
锁定关联的mutex. 高效地调用 mutex()->lock().

异常:

  • 抛出mutex()->lock()所抛出的异常。
  • 如果没有相关联的 mutex, std::system_error就会抛出,所携带的错误码是std::errc::operation_not_permitted.
  • 如果mutex已经被本unique_lock锁定了(换句话说,owns_lock为true),那么std::system_error就会被抛出,错误码是std::errc::resource_deadlock_would_occur.

参见示例程序:

#include 
#include 
#include 
#include 
#include 
 
int main()
{
    int counter = 0;
    std::mutex counter_mutex;
    std::vector threads;
 
    auto worker_task = [&](int id) {
        // Note, this is just lock! See the document of the constructor. 
        // The reason is, it is RAAI (Resource Acquisition is Initialization)
        std::unique_lock lock(counter_mutex);  
        
        ++counter;
        std::cout << id << ", initial counter: " << counter << '\n';
        lock.unlock();
        
        // don't hold the lock while we simulate an expensive operation
        std::this_thread::sleep_for(std::chrono::seconds(1));
 
        lock.lock();
        ++counter;
        std::cout << id << ", final counter: " << counter << '\n';
    };
 
    for (int i = 0; i < 10; ++i) {
        // vector::push_back() cannot work due to std::thread is not copyable.
        threads.emplace_back(worker_task, i);
    }
 
    for (auto &thread : threads) thread.join();
}


/**
Possible Output:
0, initial counter: 1
1, initial counter: 2
2, initial counter: 3
3, initial counter: 4
4, initial counter: 5
5, initial counter: 6
6, initial counter: 7
7, initial counter: 8
8, initial counter: 9
9, initial counter: 10
1, final counter: 11
0, final counter: 12
3, final counter: 13
5, final counter: 14
2, final counter: 15
4, final counter: 16
7, final counter: 17
9, final counter: 18
6, final counter: 19
8, final counter: 20
**/

std::unique_lock::try_lock 函数

bool try_lock(); (since C++11)
试图锁定相关联的mutex,而不会阻塞。高效地调用 mutex()->try_lock().
如果没有相关联的mutex或者该mutex已经被该unique_lock锁定,那么 std::system_error 就会被抛出。

异常:

  • 任何 mutex()->try_lock() 抛出的异常都会被抛出 (Mutex类型不会被try_lock抛出,但一个自定义的Lockable类型可能会被抛出)。
  • 如果没有相关联的mutex,那么std::system_error就会被抛出,其错误码是std::errc::operation_not_permitted.
  • 如果mutex已经被本unique_lock锁定了,std::system_error也会被抛出,错误码是std::errc::resource_deadlock_would_occur.

std::unique_lock::try_lock_for 函数

template
bool try_lock_for( const std::chrono::duration& timeout_duration ); (since C++11)

试图锁定相关联的mutex. 会阻塞住,直到指定的时间段超时或者锁定成功。一旦成功锁定,就返回true,否则返回false. 高效地调用 mutex()->try_lock_for(timeout_duration).
一个稳定的时钟被用来衡量此时间段。本函数可能阻塞地比timeout_duration的时间更久一些,原因是系统调度或资源竞争所导致的延时。
如果没有相关联的mutex或该mutex已经被锁定,则std::system_error就会被抛出。

返回值:
获得mutex则返回true,否则false.

异常:

  • 任何被mutex()->try_lock_for(timeout_duration)抛出的异常都将被抛出;
  • 若没有相关联的mutex,则std::system_error就会被抛出,错误码是std::errc::operation_not_permitted;
  • 若mutex已经被锁的那个,则std::system_error就会被抛出,错误码是std::errc::resource_deadlock_would_occur.

std::unique_lock::unlock 函数

void unlock(); (since C++11)
解锁相关的mutex.
如果没有相关联的mutex或该mutex已经被锁定,则std::system_error就会被抛出。

异常:

  • 任何被mutex()->unlock()抛出的异常都会被抛出。
  • 如果没有相关联的mutex或者mutex并没有被锁定,则std::system_error就会被抛出,错误码是std::errc::operation_not_permitted.

std::unique_lock::mutex 函数

mutex_type* mutex() const; (since C++11)
返回一个指向所关联的mutex的指针,或者如果没有相关联的mutex的话就返回一个空指针。

std::unique_lock::owns_lock 函数

bool owns_lock() const; (since C++11)
检查 *this 是否已经锁住mutex.
是,则返回true;否则返回false.

std::unique_lock::operator bool 函数

explicit operator bool() const; (since C++11)

检查 *this 是否已锁定一个mutex. 高效地调用owns_lock().
是,则返回true;否则返回false.

示例程序

#include 
#include 
#include 
#include 
 
struct Box {
    explicit Box(int num) : num_things{num} {}
 
    int num_things;
    std::mutex m;
};
 
void transfer(Box &a, Box &b, int num)
{
    // don't actually take the locks yet
    std::unique_lock lock1(a.m, std::defer_lock);
    std::unique_lock lock2(b.m, std::defer_lock);
 
    // lock both unique_locks without deadlock
    std::lock(lock1, lock2);
 
    a.num_things -= num;
    b.num_things += num;
 
    // 'a.m' and 'b.m' mutexes unlocked outside of 'unique_lock'
}
 
int main()
{
    Box acc1(100);
    Box acc2(50);
 
    std::thread t1(transfer, std::ref(acc1), std::ref(acc2), 10);
    std::thread t2(transfer, std::ref(acc2), std::ref(acc1), 5);
 
    t1.join();
    t2.join();
    
    std::cout << "acc1: " << acc1.num_things << std::endl;
    std::cout << "acc2: " << acc2.num_things << std::endl;
}

/** Output: 
acc1: 95
acc2: 55
**/

(译注: 原示例程序没有任何输出且变量命名比较令人费解,故略作修改。)

(完)

你可能感兴趣的:(C++,多线程,C++11)