这篇讲讲CyclicBarrier栅栏,从它的名字可以看出,它是可循环使用的。它的功能和CountDownLatch类似,也是让一组线程等待,然后一起开始往下执行。但是两者还是有几个区别
1. 等待的对象不同。CountDownLatch的一组线程等待的是一个事件,或者说是一个计数器归0的事件。而CyclicBarrier等待的对象是线程,只有线程都到齐了才往下执行
2. 使用方式不同,这个也是由等待的对象不同引起的,CountDownLatch需要调用await()来让线程等待,调用countDown()来修改状态,直到触发状态为0的事件。而CyclicBarrier只需要调用await()让线程等待,当调用await()方法的线程数满足条件,就自动唤醒所有线程往下执行
3. CyclicBarrier可以自动循环使用,当一次拦截被打开后,会自动创建下一个拦截。CountDownLatch的计数器归0后不能再次使用
4. 底层实现不同,CountDownLatch使用AQS来实现底层同步,CyclicBarrier基于更上层的ReetrantLock + Condition条件队列实现
5. 失效机制不同,在CountDownLatch等待的线程如果被中断或者超时取消,不会影响其他线程。而CyclicBarrier采用all-or-none的机制,要么全部不通过,要么全部都通过,也就是说一旦在CyclicBarrier等待的线程有一个被中断或者超时取消,那么其他所有在这个CyclicBarrier等待的线程都被唤醒,通过栅栏往下执行
6. CyclicBarrier支持线程全部通过之后的回调功能,通过传入一个Runnable对象,由最后一个到达的线程来执行。而CountDownLatch不支持回调机制
下面看看CyclicBarrier的源代码,它有一个内部类Generation来处理循环使用的问题,维护了一个broker状态表示当前的栅栏是否失效。如果失效,可以重置栅栏的状态。当栅栏被打破时,就设置当前generation的broker为true表示失效,并唤醒所有等待的线程,即all-or-none机制
private static class Generation { boolean broken = false; } private void nextGeneration() { // signal completion of last generation trip.signalAll(); // set up next generation count = parties; generation = new Generation(); } private void breakBarrier() { generation.broken = true; count = parties; trip.signalAll(); }
维护了一个ReentrantLock来作同步,并创建了一个相关的条件队列Condition,使用Condition的await()方法让线程在同一个条件队列等待,使用Condition.signalAll()唤醒所有在通过一条件队列等待的线程。
/** The lock for guarding barrier entry */ private final ReentrantLock lock = new ReentrantLock(); /** Condition to wait on until tripped */ private final Condition trip = lock.newCondition();
/* The command to run when tripped */ private final Runnable barrierCommand; public CyclicBarrier(int parties, Runnable barrierAction) { if (parties <= 0) throw new IllegalArgumentException(); this.parties = parties; this.count = parties; this.barrierCommand = barrierAction; }
维护了一个count来计数,当await()方法被调用一次, count就减1,直到count为0打开栅栏。
private int count;
1. 进入锁相当于读volatile,会清空CPU缓存,强制从内存读取
2. 离开锁相当于写volatile,会把CPU写缓冲区的数据强制刷新到内存
CyclicBarrier常用支持普通的等待和限时的等待。最后都是落到了dowait()方法。
public int await() throws InterruptedException, BrokenBarrierException { try { return dowait(false, 0L); } catch (TimeoutException toe) { throw new Error(toe); // cannot happen; } } public int await(long timeout, TimeUnit unit) throws InterruptedException, BrokenBarrierException, TimeoutException { return dowait(true, unit.toNanos(timeout)); }
1. 必须先获取锁,保证了可见性,有序性,原子性
2. 判断当前栅栏的状态,如果已经失效,抛出BrokerBarrierException异常
3. 如果线程被中断,那么让栅栏失效,会唤醒所有等待线程往下执行
4. 执行一次dowait就对count减一,用index记录下当前线程执行是的count值作为索引
5. 如果index == 0表示是最后到达的线程,可以打开栅栏了。首先如果有回调,就执行回调。然后重置栅栏状态,使之可以循环使用,返回0
6. 如果index不为0,表示不是最后到达的线程,就轮询等待,这里支持了限时操作,使用了Condition条件队列的await()机制。直到超时或者栅栏被正常失效。栅栏失效后会使用Condition来唤醒所有在同一个条件队列等待的线程。
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(); } }
1. 创建一个5个容量的CyclicBarrier,并设置回调
2. 运行12个线程
package com.lock.test; import java.util.concurrent.CyclicBarrier; public class CyclicBarrierUsecase { private CyclicBarrier barrier = new CyclicBarrier(5, new Runnable(){ @Override public void run() { System.out.println("Callback is running"); } }); public void race() throws Exception{ System.out.println("Thread " + Thread.currentThread().getName() + " is waiting the resource"); barrier.await(); System.out.println("Thread " + Thread.currentThread().getName() + " got the resource"); } public static void main(String[] args){ final CyclicBarrierUsecase usecase = new CyclicBarrierUsecase(); for(int i = 0; i < 12; i++){ Thread t = new Thread(new Runnable(){ @Override public void run() { try { usecase.race(); } catch (Exception e) { // TODO Auto-generated catch block e.printStackTrace(); } } }, String.valueOf(i)); t.start(); } } }
1. 可以看到5个线程在等待,直到满5个线程到达之后打开栅栏,这5个线程往下执行,并执行回调
2. 栅栏被循环使用了,又有5个线程等待,直到满5个线程到达又打开栅栏往下执行,并执行回调
3. 栅栏又被循环使用,但是只有2个线程,不满5个,就一直等待
Thread 0 is waiting the resource Thread 4 is waiting the resource Thread 5 is waiting the resource Thread 3 is waiting the resource Thread 2 is waiting the resource Callback is running Thread 1 is waiting the resource Thread 0 got the resource Thread 2 got the resource Thread 6 is waiting the resource Thread 7 is waiting the resource Thread 4 got the resource Thread 9 is waiting the resource Thread 8 is waiting the resource Thread 3 got the resource Thread 5 got the resource Callback is running Thread 8 got the resource Thread 1 got the resource Thread 7 got the resource Thread 6 got the resource Thread 10 is waiting the resource Thread 11 is waiting the resource Thread 9 got the resource