• 谈论高并发(三十)解析java.util.concurrent各种组件(十二) 认识CyclicBarrier栅栏


    这次谈话CyclicBarrier栅栏,如可以从它的名字可以看出,它是可重复使用。

    它的功能和CountDownLatch类别似,也让一组线程等待,然后开始往下跑起来。但也有在两者之间有一些差别

    1. 不同的对象等。CountDownLatch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,仅仅有线程都到齐了才往下运行

    2. 使用方式不同,这个也是由等待的对象不同引起的,CountDownLatch须要调用await()来让线程等待。调用countDown()来改动状态,直到触发状态为0的事件。而CyclicBarrier仅仅须要调用await()让线程等待,当调用await()方法的线程数满足条件。就自己主动唤醒全部线程往下运行

    3. CyclicBarrier能够自己主动循环使用,当一次拦截被打开后,会自己主动创建下一个拦截。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等待的线程假设被中断或者超时取消,不会影响其它线程。而CyclicBarrier採用all-or-none的机制,要么所有不通过,要么所有都通过。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,那么其它所有在这个CyclicBarrier等待的线程都被唤醒,通过栅栏往下运行

    6. CyclicBarrier支持线程所有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运行。而CountDownLatch不支持回调机制


    以下看看CyclicBarrier的源码,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。假设失效,能够重置栅栏的状态。

    当栅栏被打破时,就设置当前generation的broker为true表示失效,并唤醒全部等待的线程,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()唤醒全部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();

    维护了一个Runnable引用来支持回调功能

    /* 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;

    能够看到CyclicBarrier的实例属性都没有使用volatile变量。那它怎么保证状态的可见性呢?CyclicBarrier使用了加显式锁的方式。我们知道显式锁和内置锁一样,都保证了可见性,有序性和原子性。

    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经常使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));
        }
    

    来看看dowait方法

    1. 必须先获取锁,保证了可见性,有序性,原子性

    2. 推断当前栅栏的状态,假设已经失效,抛出BrokerBarrierException异常

    3. 假设线程被中断。那么让栅栏失效,会唤醒全部等待线程往下运行

    4. 运行一次dowait就对count减一,用index记录下当前线程运行是的count值作为索引

    5. 假设index == 0表示是最后到达的线程,能够打开栅栏了。首先假设有回调。就运行回调。然后重置栅栏状态,使之能够循环使用,返回0

    6. 假设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来唤醒全部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();
            }
        }

    以下使用一个測试用例来測试CyclicBarrier的功能

    1. 创建一个5个容量的CyclicBarrier,并设置回调

    2. 执行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();
    		}
    	}
    }
    

    測试结果:

    1. 能够看到5个线程在等待。直到满5个线程到达之后打开栅栏,这5个线程往下运行,并运行回调

    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运行。并运行回调

    3. 栅栏又被循环使用,可是仅仅有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

    版权声明:本文博客原创文章。博客,未经同意,不得转载。

  • 相关阅读:
    打开fiddler之后,python写的爬虫无法正常运行
    Fiddler抓包https失败踩坑
    python 报错 name is not defined
    HttpWebRequest爬虫问题汇总
    Navicat Premium 16 for Mac 最新版
    CentOS升级polkit版本,解决 Linux Polkit 存在权限提升的漏洞 (CVE20214034)
    Nginx Proxy Manager
    docker tar.gz
    C# HttpClientHandle
    nginx joplin
  • 原文地址:https://www.cnblogs.com/hrhguanli/p/4671329.html
Copyright © 2020-2023  润新知