Java并发包中CyclicBarrier的工作原理、使用示例

1. CyclicBarrier的介绍与源码分析

CyclicBarrier 的字面意思是可循环（Cyclic）使用的屏障（Barrier）。它要做的事情是，让一组线程到达一个屏障（也可以叫同步点）时被阻塞，直到最后一个线程到达屏障时，屏障才会开门，所有被屏障拦截的线程才会继续干活。线程进入屏障通过CyclicBarrier的await()方法。

CyclicBarrier默认的构造方法是CyclicBarrier(int parties)，其参数表示屏障拦截的线程数量，每个线程调用await方法告诉CyclicBarrier我已经到达了屏障，然后当前线程被阻塞。

CyclicBarrier还提供一个更高级的构造函数CyclicBarrier(int parties, Runnable barrierAction)，用于在线程到达屏障时，优先执行barrierAction这个Runnable对象，方便处理更复杂的业务场景。

构造函数

public CyclicBarrier(int parties) {
    this(parties, null);
}
public int getParties() {
    return parties;
}

实现原理：在CyclicBarrier的内部定义了一个Lock对象，每当一个线程调用CyclicBarrier的await方法时，将剩余拦截的线程数减1，然后判断剩余拦截数是否为0，如果不是，进入Lock对象的条件队列等待。如果是，执行barrierAction对象的Runnable方法，然后将锁的条件队列中的所有线程放入锁等待队列中，这些线程会依次的获取锁、释放锁，接着先从await方法返回，再从CyclicBarrier的await方法中返回。

await源码

public int await() throws InterruptedException, BrokenBarrierException {
    try {
        return dowait(false, 0L);
    } catch (TimeoutException toe) {
        throw new Error(toe); // cannot happen
    }
}

dowait源码

private int dowait(boolean timed, long nanos)
    throws InterruptedException, BrokenBarrierException,
           TimeoutException {
    final ReentrantLock lock = this.lock;
    lock.lock();
    try {
        final Generation g = generation;

        if (g.broken)
            throw new BrokenBarrierException();

        if (Thread.interrupted()) {
            breakBarrier();
            throw new InterruptedException();
        }

        int index = --count;
        if (index == 0) {  // tripped
            boolean ranAction = false;
            try {
                final Runnable command = barrierCommand;
                if (command != null)
                    command.run();
                ranAction = true;
                nextGeneration();
                return 0;
            } finally {
                if (!ranAction)
                    breakBarrier();
            }
        }

        // loop until tripped, broken, interrupted, or timed out
        for (;;) {
            try {
                if (!timed)
                    trip.await();
                else if (nanos > 0L)
                    nanos = trip.awaitNanos(nanos);
            } catch (InterruptedException ie) {
                if (g == generation && ! g.broken) {
                    breakBarrier();
                    throw ie;
                } else {
                    // We're about to finish waiting even if we had not
                    // been interrupted, so this interrupt is deemed to
                    // "belong" to subsequent execution.
                    Thread.currentThread().interrupt();
                }
            }

            if (g.broken)
                throw new BrokenBarrierException();

            if (g != generation)
                return index;

            if (timed && nanos <= 0L) {
                breakBarrier();
                throw new TimeoutException();
            }
        }
    } finally {
        lock.unlock();
    }
}

当最后一个线程到达屏障点，也就是执行dowait方法时，会在return 0 返回之前调用finally块中的breakBarrier方法。

breakBarrier源代码

private void breakBarrier() {
    generation.broken = true;
    count = parties;
    trip.signalAll();
}

CyclicBarrier主要用于一组线程之间的相互等待，而CountDownLatch一般用于一组线程等待另一组些线程。实际上可以通过CountDownLatch的countDown()和await()来实现CyclicBarrier的功能。即 CountDownLatch中的countDown()+await() = CyclicBarrier中的await()。注意：在一个线程中先调用countDown()，然后调用await()。

其它方法：CycliBarrier对象可以重复使用，重用之前应当调用CyclicBarrier对象的reset方法。

reset源码

public void reset() {
    final ReentrantLock lock = this.lock;
    lock.lock();
    try {
        breakBarrier();   // break the current generation
        nextGeneration(); // start a new generation
    } finally {
        lock.unlock();
    }
}

2. 使用示例

package javalearning;

import java.util.Random;
import java.util.concurrent.BrokenBarrierException;
import java.util.concurrent.CyclicBarrier;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class CyclicBarrierDemo {
	private CyclicBarrier cb = new CyclicBarrier(4);
	private Random rnd = new Random();

	class TaskDemo implements Runnable{
		private String id;
		TaskDemo(String id){
			this.id = id;
		}
		@Override
		public void run(){
			try {
				Thread.sleep(rnd.nextInt(1000));
				System.out.println("Thread " + id + " will wait");
				cb.await();
				System.out.println("-------Thread " + id + " is over");
			} catch (InterruptedException e) {
			} catch (BrokenBarrierException e) {
			}
		}
	}

	public static void main(String[] args){
		CyclicBarrierDemo cbd = new CyclicBarrierDemo();
		ExecutorService es = Executors.newCachedThreadPool();
		es.submit(cbd.new TaskDemo("a"));
		es.submit(cbd.new TaskDemo("b"));
		es.submit(cbd.new TaskDemo("c"));
		es.submit(cbd.new TaskDemo("d"));
		es.shutdown();
	}
}

在这个示例中，我们创建了四个线程a、b、c、d，这四个线程提交给了线程池。四个线程不同时间到达cb.await()语句，当四个线程都输出“Thread x will wait”以后才会输出“Thread x is over”。

运行结果

Thread d will wait

Thread a will wait

Thread c will wait

Thread b will wait

-------Thread b is over

-------Thread d is over

-------Thread a is over

-------Thread c is over

3. 参考内容

[1] http://ifeve.com/concurrency-cyclicbarrier/