多线程应用中,经常会遇到这种场景:后面的处理,依赖前面的N个线程的处理结果,必须等前面的线程执行完毕后,后面的代码才允许执行。
在我不知道CyclicBarrier之前,最容易想到的就是放置一个公用的static变量,假如有10个线程,每个线程处理完上去累加下结果,然后后面用一个死循环(或类似线程阻塞的方法),去数这个结果,达到10个,说明大家都爽完了,可以进行后续的事情了,这个想法虽然土鳖,但是基本上跟语言无关,几乎所有主流编程语言都支持。
package yjmyzz.test; public class ThreadLockTest { public static int flag = 0;//公用变量 public static void main(String[] args) throws Exception { ThreadLockTest testObj = new ThreadLockTest(); final int threadNum = 10; for (int i = 0; i < threadNum; i++) { new Thread(new MyRunable(i, testObj)).start(); } while (true) { if (testObj.flag >= threadNum) { System.out.println("-----------\n所有thread执行完成!"); break; } Thread.sleep(10); } } static class MyRunable implements Runnable { int _i = 0; ThreadLockTest _test; public MyRunable(int i, ThreadLockTest test) { this._i = i; this._test = test; } @Override public void run() { try { Thread.sleep((long) (Math.random() * 10)); System.out.println("thread " + _i + " done"); //利用synchronized获得同步锁 synchronized (_test) { _test.flag += 1; } System.out.println("thread " + _i + " => " + _test.flag);//测试用 } catch (Exception e) { e.printStackTrace(); } } } }
输出结果:
thread 0 done
thread 0 => 1
thread 9 done
thread 9 => 2
thread 1 done
thread 1 => 3
thread 3 done
thread 3 => 4
thread 7 done
thread 7 => 5
thread 6 done
thread 6 => 6
thread 2 done
thread 2 => 7
thread 4 done
thread 4 => 8
thread 8 done
thread 8 => 9
thread 5 done
thread 5 => 10
-----------
所有thread执行完成!
除了这个方法,还可以借助FutureTask,达到类似的效果,其get方法会阻塞线程,等到该异步处理完成。缺点就是,FutureTask调用的是Callable,必须要有返回值,所以就算你不想要返回值,也得返回点啥
package yjmyzz.test; import java.util.concurrent.Callable; import java.util.concurrent.ExecutionException; import java.util.concurrent.FutureTask; public class FutureTaskTest { public static void main(String[] args) throws ExecutionException, InterruptedException { FutureTask<String>[] tasks = new FutureTask[10]; for (int i = 0; i < tasks.length; i++) { final int j = i; tasks[i] = new FutureTask<String>(new Callable<String>() { @Override public String call() throws Exception { Thread.sleep((long) (Math.random() * 100)); return "task" + j + " done"; } }); new Thread(tasks[i]).start(); } for (int i = 0; i < tasks.length; i++) { System.out.println(tasks[i].get());//依次等待所有task执行完毕 } System.out.println("-----------\n所有task执行完成!"); } }
执行结果:
task0 done
task1 done
task2 done
task3 done
task4 done
task5 done
task6 done
task7 done
task8 done
task9 done
-----------
所有task执行完成!
此外,Thread的Join方法也可以实现类似的效果,主要代码如下:
public static void main(String[] args) throws Exception { final int threadNum = 10; Thread[] threads = new Thread[threadNum]; for (int i = 0; i < threadNum; i++) { threads[i] = new Thread(new MyRunable(i)); threads[i].start(); } for (int i = 0; i < threadNum; i++) { threads[i].join(); } System.out.println("-----------\n所有thread执行完成!"); }
当然,这个需求最“正统”的解法应该是使用CyclicBarrier,它可以设置一个所谓的“屏障点”(或称集合点),好比在一项团队活动中,每个人都是一个线程,但是规定某一项任务开始前,所有人必须先到达集合点,集合完成后,才能继续后面的任务。
package yjmyzz.test; import java.util.concurrent.CyclicBarrier; public class ThreadTest { public static void main(String[] args) throws Exception { final int threadNum = 10; CyclicBarrier cb = new CyclicBarrier(threadNum + 1);//注意:10个子线程 + 1个主线程 for (int i = 0; i < threadNum; i++) { new Thread(new MyRunable(cb, i)).start(); } cb.await(); System.out.println("-----------\n所有thread执行完成!"); } static class MyRunable implements Runnable { CyclicBarrier _cb; int _i = 0; public MyRunable(CyclicBarrier cb, int i) { this._cb = cb; this._i = i; } @Override public void run() { try { Thread.sleep((long) (Math.random() * 100)); System.out.println("thread " + _i + " done,正在等候其它线程完成..."); _cb.await(); } catch (Exception e) { e.printStackTrace(); } } } }
thread 9 done,正在等候其它线程完成...
thread 5 done,正在等候其它线程完成...
thread 0 done,正在等候其它线程完成...
thread 6 done,正在等候其它线程完成...
thread 4 done,正在等候其它线程完成...
thread 2 done,正在等候其它线程完成...
thread 3 done,正在等候其它线程完成...
thread 8 done,正在等候其它线程完成...
thread 7 done,正在等候其它线程完成...
thread 1 done,正在等候其它线程完成...
-----------
所有thread执行完成!
参考文章:
http://ifeve.com/concurrency-cyclicbarrier/
http://ifeve.com/thread-synchronization-utilities-5/
http://ifeve.com/semaphore-countdownlatch-cyclicbarrier-phaser-exchanger-in-java/
http://ifeve.com/thread-management-7/