CountDownLatch
并发类CountDownLatch是什么?
CountDownLatch:一个或多个线程等待其他线程完成操作。
什么情况下使用?
某一个动作需要等待其它线程完成后才会触发。
举个栗子,一个班上50个人,考完试之后需要计算全班同学的总成绩,这种情况使用CountDownLatch并发类就最合适了,每一个人的成绩等于一个线程,需要等待50个线程执行完之后才能执行最后一个动作(计算总成绩)。
如何使用并发类CountDownLatch?
代码示例:
public class CountDownLatchDemo implements Runnable {
private static final int threadCount = 3;
private static CountDownLatch countDownLatch = new CountDownLatch(threadCount);
private static double sum = 0D;
public String name;
public CountDownLatchDemo() {
}
public CountDownLatchDemo(String name) {
this.name = name;
}
public static void main(String[] args) throws InterruptedException {
ExecutorService executorService = Executors.newFixedThreadPool(threadCount);
executorService.execute(new CountDownLatchDemo("A同学"));
executorService.execute(new CountDownLatchDemo("B同学"));
executorService.execute(new CountDownLatchDemo("C同学"));
countDownLatch.await();
System.out.println("所有人的总成绩 " + sum);
}
@Override
public void run() {
double grade = Math.round(Math.random() * 100);
System.out.println("计算 " + name + " 成绩,成绩是" + grade);
sum += grade;
countDownLatch.countDown();
}
}
输出结果:
计算 C同学 成绩,成绩是26.0
计算 B同学 成绩,成绩是18.0
计算 A同学 成绩,成绩是53.0
所有人的总成绩 97.0
使用该并发类并不难,熟悉几个主要方法就可以。
- CountDownLatch(int count): 参数count为计数值
- await(): 调用await()方法的线程会被挂起,它会等待直到count值为0才继续执行
- await(long timeout,TimeUnit unit): 和await()类似,只不过等待一定的时间后count值还没变为0的话就会继续执行
- countDown(): 将count值减1
- getCount():获取count值
好了,知道大概怎么使用这个并发类,接下来开始撸源码!
public class CountDownLatch {
//内部使用Sync继承AQS
private static final class Sync extends AbstractQueuedSynchronizer {
private static final long serialVersionUID = 4982264981922014374L;
Sync(int count) {
//设置计算值
setState(count);
}
//获取计算值
int getCount() {
return getState();
}
//尝试获取共享锁
protected int tryAcquireShared(int acquires) {
return (getState() == 0) ? 1 : -1;
}
//尝试释放锁
protected boolean tryReleaseShared(int releases) {
for (; ; ) {
int c = getState();
if (c == 0)
return false;
int nextc = c - 1;
if (compareAndSetState(c, nextc))
return nextc == 0;
}
}
}
private final Sync sync;
//CountDownLatch唯一的构造器,参数count值为阻塞线程数
public CountDownLatch(int count) {
if (count < 0) throw new IllegalArgumentException("count < 0");
this.sync = new Sync(count);
}
//阻塞线程,直到count值为0
public void await() throws InterruptedException {
sync.acquireSharedInterruptibly(1);
}
//限定阻塞时间,超出限定时间后,count值还没为0都会执行之后的线程
public boolean await(long timeout, TimeUnit unit)
throws InterruptedException {
return sync.tryAcquireSharedNanos(1, unit.toNanos(timeout));
}
//释放锁操作,每调用一次,count值减一
public void countDown() {
sync.releaseShared(1);
}
//获取count值
public long getCount() {
return sync.getCount();
}
public String toString() {
return super.toString() + "[Count = " + sync.getCount() + "]";
}
}
大概流程就是,先调用CountDownLatch构造器设置计算值(count),当主线程(main方法)执行到await()方法的时候就会阻塞主线程,然后执行线程池内的任务,通过执行countDown()方法,将count值不断的减1,直到count值为0的时候就会释放锁,继续执行主线程。
CyclicBarrier
CyclicBarrier:一组线程相互等待,达到一个共同点再继续执行。
比较困惑的一点是,CyclicBarrier和CountDownLatch两个并发类有什么区别?从概念上和例子上来看,貌似都差不多呀。
参考多篇大佬博客,总结出以下个人观点:
源码上区别:
CountDownLatch基于AQS实现的;设置了等待线程数后无法重置;
CyclicBarrier基于Condition和ReentrantLock锁实现;设置等待线程数后可以调用reset()重置;
概念上区别:
CountDownLatch:一个或多个线程等待其他线程执行完成后再执行。(关键词:“其他线程执行完成后”)
例子:线程S需要等待线程A、B、C都执行完成后才会执行。
CyclicBarrier:多个线程相互等待,到达阻塞点(屏障)后等待其他线程,直到所有线程都达到阻塞点后,阻塞解除并唤醒所有等待线程,所有的线程都继续往下执行。(关键词:“阻塞点”)
举个栗子:小A、小B,小C三位朋友约定在广州塔见面之后去喝早茶,分别从三处不同的地方赶往广州塔碰面。但由于小C迟到,小A,小B等待小C,直到小C都到达广州塔后再一同去喝早茶。使用并发类CyclicBarrier解释就是,线程A,线程B被阻塞,直到线程C执行到阻塞点后阻塞解除,唤醒所有线程。线程A、B、C都继续往下执行。
代码示例:
public class CyclicBarrierDemo implements Runnable {
public static int count = 3;
public String name;
public CyclicBarrierDemo(String name) {
this.name = name;
}
public CyclicBarrierDemo() {
}
public static CyclicBarrier cyclicBarrier = new CyclicBarrier(count, new Runnable() {
@Override
public void run() {
System.out.println("大家启程去茶楼,喝茶去噜...");
}
});
public static void main(String[] args) {
//创建线程池
ExecutorService executor = Executors.newFixedThreadPool(count);
try {
//执行线程
executor.execute(new CyclicBarrierDemo("小A"));
executor.execute(new CyclicBarrierDemo("小B"));
executor.execute(new CyclicBarrierDemo("小C"));
} catch (Exception e) {
e.printStackTrace();
} finally {
executor.shutdown();
}
}
@Override
public void run() {
try {
if ("小C".equals(name)) {
System.out.println(name + "路上塞车");
Thread.sleep(1000);
}
System.out.println(this.name + "到达广州塔... ");
cyclicBarrier.await();
System.out.println(this.name + "到达茶楼 ");
} catch (InterruptedException e) {
e.printStackTrace();
} catch (BrokenBarrierException e) {
e.printStackTrace();
}
}
}
运行结果:
小B到达广州塔...
小C路上塞车
小A到达广州塔...
小C到达广州塔...
大家启程去茶楼,喝茶去噜...
小C到达茶楼
小B到达茶楼
小A到达茶楼
接下来,膜拜Doug Lea写的源码
public class CyclicBarrier {
//Generation是CyclicBarrier的一个私有内部类,只有一个成员变量broken来标识当前的barrier是否已“损坏”:
private static class Generation {
boolean broken = false;
}
private final ReentrantLock lock = new ReentrantLock();
private final Condition trip = lock.newCondition(); //通过lock得到的一个状态变量
private final int parties; //表示总的等待线程的数量。
private final Runnable barrierCommand; //当屏障(barrier)撤销时,需要执行的操作
private Generation generation = new Generation(); //当前的Generation。每当屏障失效或者开闸之后都会自动替换掉。从而实现重置的功能。
private int count;//实际中仍在等待的线程数,每当有一个线程到达屏障点,count值就会减一;当一次新的运算开始后,count的值被重置为parties。
//唤醒等待线程,设置下一个Generation
private void nextGeneration() {
// 唤醒所有等待线程
trip.signalAll();
// 重置进入屏障的线程数量
count = parties;
// 新生一代
generation = new Generation();
}
//将当前屏障置为破坏状态、重置count、并唤醒所有被阻塞的线程。
private void breakBarrier() {
generation.broken = true;
count = parties;
trip.signalAll();
}
private int dowait(boolean timed, long nanos)
throws InterruptedException, BrokenBarrierException,
TimeoutException {
final ReentrantLock lock = this.lock;
lock.lock(); //获取锁
try {
//保存此时的generation
final Generation g = generation;
//判断屏障是否被破坏
if (g.broken)
throw new BrokenBarrierException();
//判断线程是否被中断
if (Thread.interrupted()) {
breakBarrier();
throw new InterruptedException();
}
//正在等待进入屏障的线程数量减一
int index = --count;
//判断等待进入屏障的线程数量是否为0
if (index == 0) {
// 运行的动作标识
boolean ranAction = false;
try {
// 保存运行动作
final Runnable command = barrierCommand;
// 如果动作不为空则运行
if (command != null)
command.run();
// 设置运行的动作状态
ranAction = true;
// 进入下一代,唤醒所有线程;重置count和generation值.
nextGeneration();
return 0;
} finally {
if (!ranAction)
breakBarrier();
}
}
// index不为0,无限循环
for (; ; ) {
try {
// 如果没有设置定时就一直等待
if (!timed)
trip.await();
else if (nanos > 0L) // 如果设置了定时
nanos = trip.awaitNanos(nanos); // 等待定时的时间后,重新唤醒
} catch (InterruptedException ie) {
if (g == generation && !g.broken) {
breakBarrier();
throw ie;
} else {
Thread.currentThread().interrupt();
}
}
// 取消阻塞后,重新判断状态
if (g.broken)
throw new BrokenBarrierException();
if (g != generation)
return index;
if (timed && nanos <= 0L) {
breakBarrier();
throw new TimeoutException();
}
}
} finally {
lock.unlock();
}
}
public CyclicBarrier(int parties, Runnable barrierAction) {
if (parties <= 0) throw new IllegalArgumentException();
//表示必须同时到达屏障(barrier)的线程个数
this.parties = parties;
//表示处于等待状态的线程个数
this.count = parties;
//表示当等待线程数为0时,会执行的动作
this.barrierCommand = barrierAction;
}
public CyclicBarrier(int parties) {
this(parties, null);
}
//返回参与相互等待的线程数
public int getParties() {
return parties;
}
public int await() throws InterruptedException, BrokenBarrierException {
try {
return dowait(false, 0L);
} catch (TimeoutException toe) {
throw new Error(toe);
}
}
public int await(long timeout, TimeUnit unit)
throws InterruptedException,
BrokenBarrierException,
TimeoutException {
return dowait(true, unit.toNanos(timeout));
}
// 判断此屏障是否处于中断状态。
public boolean isBroken() {
final ReentrantLock lock = this.lock;
lock.lock();
try {
return generation.broken;
} finally {
lock.unlock();
}
}
//将屏障重置为其初始状态。
public void reset() {
final ReentrantLock lock = this.lock;
lock.lock();
try {
breakBarrier();
nextGeneration();
} finally {
lock.unlock();
}
}
//返回当前在屏障处等待的参与者数目,此方法主要用于调试和断言。
public int getNumberWaiting() {
final ReentrantLock lock = this.lock;
lock.lock();
try {
return parties - count;
} finally {
lock.unlock();
}
}
}
其中,使用CyclicBarrier并发类的核心就是dowait方法,结合源码分析,得出以下的流程图:
CyclicBarrier-dowait-flow.pngSemaphor
并发类Semaphor是什么?
Semaphore(信号量)是用来控制同时访问特定资源的线程数量,它通过协调各个线程,以保证合理的使用公共资源。
什么情况下使用Semaphor?
用于限制获取某种资源的线程数量。
怎样使用Semaphor?
public class SemaphoreDemo {
private static final int permits = 4;
private static Semaphore semaphore = new Semaphore(permits);
public static void main(String[] args) {
ExecutorService executor = Executors.newCachedThreadPool();
for (int i = 0; i < 12; i++) {
final int index = i;
executor.execute(() -> {
try {
semaphore.acquire();
show(index);
semaphore.release();
} catch (InterruptedException e) {
e.printStackTrace();
}
});
}
}
public static void show(int index) throws InterruptedException {
Thread.sleep(1000);
System.out.println(new Date().toString() + " --- " + Thread.currentThread().getName());
}
}
运行结果:
Tue Feb 12 10:56:30 CST 2019 --- pool-1-thread-1
Tue Feb 12 10:56:30 CST 2019 --- pool-1-thread-2
Tue Feb 12 10:56:30 CST 2019 --- pool-1-thread-6
Tue Feb 12 10:56:30 CST 2019 --- pool-1-thread-5
Tue Feb 12 10:56:31 CST 2019 --- pool-1-thread-7
Tue Feb 12 10:56:31 CST 2019 --- pool-1-thread-4
Tue Feb 12 10:56:31 CST 2019 --- pool-1-thread-8
Tue Feb 12 10:56:31 CST 2019 --- pool-1-thread-3
Tue Feb 12 10:56:32 CST 2019 --- pool-1-thread-9
Tue Feb 12 10:56:32 CST 2019 --- pool-1-thread-12
Tue Feb 12 10:56:32 CST 2019 --- pool-1-thread-10
Tue Feb 12 10:56:32 CST 2019 --- pool-1-thread-11
由运行结果可以看出,每隔一秒就有4个线程执行,可以说明,使用并发类Semaphor可以限制并发线程数量。
使用并发类Semaphor十分简单,只需要掌握几个常用的方法。
- Semaphore(int permits):创建给定的许可数和使用默认非公平策略。
- Semaphore(int permits,boolean fair):创建给定的许可数和根据fair判断使用非公平或公平策略。
- acquire():从信号量获取一个许可,在获取到一个许可前一直将线程阻塞,或者线程被中断。
- acquire(int n):从信号量获取给定数目的许可,在提供这些许可前一直将线程阻塞,或者线程已被中断。
- release():释放一个许可,将其返回给信号量。
- release(int n):释放给定数目的许可,将其返回到信号量。
- availablePermits():返回此信号量中当前可用的许可数。
接下来,开始撸源码!
public class Semaphore implements java.io.Serializable {
private static final long serialVersionUID = -3222578661600680210L;
private final Sync sync;
abstract static class Sync extends AbstractQueuedSynchronizer {
private static final long serialVersionUID = 1192457210091910933L;
Sync(int permits) {
setState(permits);
}
//获取当前许可数
final int getPermits() {
return getState();
}
final int nonfairTryAcquireShared(int acquires) {
for (; ; ) {
//获取当前许可数
int available = getState();
//剩余许可数
int remaining = available - acquires;
//如果剩余许可数小于0或者通过CAS操作成功设置剩余许可数为当前许可数则返回剩余许可数
if (remaining < 0 ||
compareAndSetState(available, remaining))
return remaining;
}
}
//尝试释放共享锁
protected final boolean tryReleaseShared(int releases) {
for (; ; ) {
int current = getState();
int next = current + releases;
if (next < current) // overflow
throw new Error("Maximum permit count exceeded");
if (compareAndSetState(current, next))
return true;
}
}
//通过CAS自旋减少许可数
final void reducePermits(int reductions) {
for (; ; ) {
int current = getState();
int next = current - reductions;
if (next > current) // underflow
throw new Error("Permit count underflow");
if (compareAndSetState(current, next))
return;
}
}
//通过CAS自旋清空许可数
final int drainPermits() {
for (; ; ) {
int current = getState();
if (current == 0 || compareAndSetState(current, 0))
return current;
}
}
}
static final class NonfairSync extends Sync {
private static final long serialVersionUID = -2694183684443567898L;
NonfairSync(int permits) {
super(permits);
}
protected int tryAcquireShared(int acquires) {
return nonfairTryAcquireShared(acquires);
}
}
static final class FairSync extends Sync {
private static final long serialVersionUID = 2014338818796000944L;
FairSync(int permits) {
super(permits);
}
protected int tryAcquireShared(int acquires) {
for (; ; ) {
//获取共享锁与非公平模式的唯一不同就是,公平模式下会判断是否有线程在等待队列中
if (hasQueuedPredecessors())
return -1;
int available = getState();
int remaining = available - acquires;
if (remaining < 0 ||
compareAndSetState(available, remaining))
return remaining;
}
}
}
//创建给定的许可数和使用默认非公平模式。
public Semaphore(int permits) {
sync = new Semaphore.NonfairSync(permits);
}
//创建给定的许可数和根据fair判断使用非公平或公平模式。
public Semaphore(int permits, boolean fair) {
sync = fair ? new Semaphore.FairSync(permits) : new Semaphore.NonfairSync(permits);
}
//从信号量获取一个许可,在获取到一个许可前一直将线程阻塞,或者线程被中断
public void acquire() throws InterruptedException {
sync.acquireSharedInterruptibly(1);
}
//从信号量获取一个许可,不会中断
public void acquireUninterruptibly() {
sync.acquireShared(1);
}
//尝试获取一个许可
public boolean tryAcquire() {
return sync.nonfairTryAcquireShared(1) >= 0;
}
//尝试在给定时间内获取一个许可
public boolean tryAcquire(long timeout, TimeUnit unit)
throws InterruptedException {
return sync.tryAcquireSharedNanos(1, unit.toNanos(timeout));
}
//尝试获取给定数量的许可数
public boolean tryAcquire(int permits) {
if (permits < 0) throw new IllegalArgumentException();
return sync.nonfairTryAcquireShared(permits) >= 0;
}
//尝试在规定时间内获取给定的许可数
public boolean tryAcquire(int permits, long timeout, TimeUnit unit)
throws InterruptedException {
if (permits < 0) throw new IllegalArgumentException();
return sync.tryAcquireSharedNanos(permits, unit.toNanos(timeout));
}
//释放一个许可
public void release() {
sync.releaseShared(1);
}
//释放给定数量的许可
public void acquire(int permits) throws InterruptedException {
if (permits < 0) throw new IllegalArgumentException();
sync.acquireSharedInterruptibly(permits);
}
//从信号量获取给定许可,不会中断
public void acquireUninterruptibly(int permits) {
if (permits < 0) throw new IllegalArgumentException();
sync.acquireShared(permits);
}
//释放给定数量的许可数
public void release(int permits) {
if (permits < 0) throw new IllegalArgumentException();
sync.releaseShared(permits);
}
//获取当前可用的许可数
public int availablePermits() {
return sync.getPermits();
}
//清空许可数
public int drainPermits() {
return sync.drainPermits();
}
//减少许可数
protected void reducePermits(int reduction) {
if (reduction < 0) throw new IllegalArgumentException();
sync.reducePermits(reduction);
}
//是否使用公平模式
public boolean isFair() {
return sync instanceof Semaphore.FairSync;
}
//判断是否有等待线程
public final boolean hasQueuedThreads() {
return sync.hasQueuedThreads();
}
//获取等待队列线程的数量
public final int getQueueLength() {
return sync.getQueueLength();
}
//获取等待队列线程集合
protected Collection<Thread> getQueuedThreads() {
return sync.getQueuedThreads();
}
public String toString() {
return super.toString() + "[Permits = " + sync.getPermits() + "]";
}
}
结合源码,画了一张使用并发类Semaphor非公平模式下调用acquire()方法的流程图:
Semaphor流程图参考资料:
https://blog.csdn.net/caoxiaohong1005/article/details/80000062
https://www.cnblogs.com/go2sea/p/5625536.html
https://blog.csdn.net/qq_19431333/article/details/70212663
《Java并发编程的艺术》
网友评论