Skip to main content

锁和同步原语

  • 在同一个线程中的多个协程，彼此之间没有可见性问题。例如他们可能会修改线程内部的某个变量，修改动作本身不需要使用atomic，不需要关注memory order。
  • 但同步原语（sync primitives）仍然是需要的，如在一个长的时间段内使用锁保护变量不被其他的协程修改，因为锁的持有者可能会让出CPU。
  • 所有的协程同步原语都是支持跨线程使用的（也包括之前介绍的thread_interrupt唤醒操作）。

Namespace​

photon::

Headers​

<photon/thread/thread.h>

API​

mutex​

class mutex {
public:
    int lock(uint64_t timeout = -1);    // threads are guaranteed to get the lock
    int try_lock();                     // in FIFO order, when there's contention
    void unlock();
}
note

The timeout type is uint64_t. -1UL means forever.

spinlock​

class spinlock {
public:
    int lock();
    int try_lock();
    void unlock();
};

scoped_lock​

using scoped_lock = locker<mutex>;

condition_variable​

class condition_variable {
public:
    int wait(mutex* m, uint64_t timeout = -1);
    int wait(mutex& m, uint64_t timeout = -1);

    int wait(spinlock* m, uint64_t timeout = -1);
    int wait(spinlock& m, uint64_t timeout = -1);

    int wait(scoped_lock& lock, uint64_t timeout = -1);
    // 针对不需要跨vCPU使用的场景，不加锁可以提升性能
    int wait_no_lock(uint64_t timeout = -1);
    
    thread* notify_one();
    int notify_all();
};

semaphore​

class semaphore {
public:
    explicit semaphore(uint64_t count = 0);
    int wait(uint64_t count, uint64_t timeout = -1);
    int signal(uint64_t count);
    uint64_t count() const;
};

rwlock​

class rwlock {
public:
    int lock(int mode, uint64_t timeout = -1);  // mode: RLOCK / WLOCK
    int unlock();
};