一、AtomicInteger
/** * @author Java和算法学习:周一 */public class AtomicInteger { //不必加volatile public AtomicInteger count = new AtomicInteger(0); //不必加synchronized public void m() { for (int i = 0; i < 1000; i++) { //count++ count.incrementAndGet(); } } public static void main(String[] args) { T01_AtomicInteger t = new T01_AtomicInteger(); List<Thread> threadList = new ArrayList<>(); for (int i = 0; i < 10; i++) { threadList.add(new Thread(t::m, "t" + i)); } threadList.forEach((o)->{ o.start(); }); threadList.forEach((o)->{ try { o.join(); } catch (InterruptedException e) { e.printStackTrace(); } }); System.out.println(t.count); }}
1、底层实现
CAS(Compare And Swap/Set)无锁优化、乐观锁
cas(V, Expected, NewValue)
if V == E
V = New
otherwise try again or fail
V:要改的值
Expected:冀望以后这个值是多少
NewValue:要设置的新值
比方要改的值是3(即最开始拿到的这个值是3),当执行CAS操作时,我冀望(Expected)这个值是3,是3我才批改这个值;如果当执行CAS时,不等于我冀望的值3,阐明这个值被其余线程改了(如果改为了4),那我就再试一次(try again)。此时我冀望这个值是4,如果执行CAS操作时,没有其余的线程再次批改这个值,即和我冀望的值4相等,那我再执行CAS操作,把值批改为新值。
(1) 如果在执行CAS操作的时候,在判断值是否为我冀望的值后,马上有其余线程把这个值改为了其余的值,那这种状况不是仍然有问题吗?
CAS操作是CPU原语反对的,也就是说CAS的操作是CPU指令级别的操作,不容许被扭转。
(2)ABA问题
(就好比 你的女朋友跟你复合时,两头又交了别的男朋友,她可能就曾经变了,不再是你原来的女朋友)
就是在我执行CAS操作的时候,这个值被其余的线程批改为了2,而后又改为了3,也就是两头通过了更改。如果是根底类型,能够不必管;如果要解决,须要加版本号,也就是这个值作任何一次的批改,版本号都加1,前面查看的时候和版本号一起查看。
(3)CAS底层是怎么做到的?
外面是通过sun.misc.Unsafe类来做的(大多数办法都是native的)。该类次要作用:间接操作JVM内存(native allocateMemory)、间接生成类实例(native allocateInstance)、间接操作变量(native getInt、native getObject)、以及CAS相干操作(native compareAndSwapObject)。该类只能通过反射或者getUnsafe失去该类的对象来应用(单例),不能间接应用。
2、SyncLong VS AtomicLong VS LongAdder
/** * @author Java和算法学习:周一 */public class AtomicSynclongLongAdder { private static AtomicLong count1 = new AtomicLong(0L); private static long count2 = 0L; private static LongAdder count3 = new LongAdder(); public static void main(String[] args) throws Exception { Thread[] threads = new Thread[1000]; //AtomicLong for (int i = 0; i < threads.length; i++) { threads[i] = new Thread(() -> { for (int j = 0; j < 100000; j++) { count1.incrementAndGet(); } }); } long start = System.currentTimeMillis(); for (Thread thread : threads) { thread.start(); } for (Thread thread : threads) { thread.join(); } long end = System.currentTimeMillis(); System.out.println("AtomicLong: " + count1.get() + " time: " + (end - start)); //long Object o = new Object(); for (int i = 0; i < threads.length; i++) { threads[i] = new Thread(() -> { for (int j = 0; j < 100000; j++) { synchronized (o) { count2++; } } }); } start = System.currentTimeMillis(); for (Thread thread : threads) { thread.start(); } for (Thread thread : threads) { thread.join(); } end = System.currentTimeMillis(); System.out.println("Long: " + count2 + " time: " + (end - start)); //LongAdder for (int i = 0; i < threads.length; i++) { threads[i] = new Thread(() -> { for (int j = 0; j < 100000; j++) { count3.increment(); } }); } start = System.currentTimeMillis(); for (Thread thread : threads) { thread.start(); } for (Thread thread : threads) { thread.join(); } end = System.currentTimeMillis(); System.out.println("LongAdder: " + count3.longValue() + " time: " + (end - start)); }}
LongAdder外部用的是分段锁(外部也是CAS实现)。它会把值放到一个数组里,比方数组大小为4,则每一个数组里锁250个线程,每一个数组都做运算,最初再把所有的数组后果求和。所以LongAdder在超高并发时,劣势特地显著。
二、ReentrantLock
1、reentrantlock能够代替synchronized,应用reentrantlock能够实现和synchronized同样的性能,然而必须得手动开释锁。应用synchronized锁定如果遇到异样,jvm会主动开释锁,然而ReentrantLock必须手动开释锁,因而常常在finally中开释锁。
2、应用reentrantlock能够进行尝试锁定(tryLock),这样无奈锁定或者在指定工夫内无奈锁定,线程能够决定是否持续期待。
3、reentrantlock能够指定为偏心锁。ReentrantLock lock=new ReentrantLock(true);示意new一个偏心锁。默认是非偏心锁。
1、偏心锁与非偏心锁
偏心锁:如果一个新来的线程,首先去判断锁的期待队列有无正在期待的线程,有则进入期待队列,等后面的先运行,无则间接去抢锁,这样的锁是偏心锁。先来后到。
非偏心锁:如果一个新来的线程,间接就去抢锁,而不去判断期待队列是否有线程在期待,这就是非偏心锁。synchronized都是非偏心锁。
新来的线程检不查看队列是偏心锁与非偏心锁的要害。
2、reentrantlock VS synchronized
1、reentrantlock能够代替synchronized
2、reentrantlock必须手动敞开锁,synchronized执行完结或异样时JVM会主动开释锁
3、reentrantlock是通过CAS实现的,synchronized实质是锁降级
4、reentrantlock能够通过tryLock来进行尝试锁定
5、reentrantlock能够在偏心锁与非偏心锁之间切换,synchronized都是非偏心锁
三、CountDownLatch(倒数 门闩)
/** * @author Java和算法学习:周一 */public class TestCountDownLatch { public static void usingCountDownLatch() { Thread[] threads = new Thread[10]; CountDownLatch countDownLatch = new CountDownLatch(threads.length); for (int i = 0; i < threads.length; i++) { threads[i] = new Thread(() -> { int count = 0; for (int j = 0; j < 10000; j++) { count++; } System.out.println(count); countDownLatch.countDown(); }); } for (Thread thread : threads) { thread.start(); } try { countDownLatch.await(); } catch (InterruptedException e) { e.printStackTrace(); } System.out.println("CountDownLatch end..."); } public static void main(String[] args) { usingCountDownLatch(); }}
countDownLatch初始大小定义为10,每减少一个线程门闩值减1(countDownLatch.countDown()), 始终期待着(countDownLatch.await()),直到门闩值为0才执行。
四、CyclicBarrier (循环 栅栏)
/** * @author Java和算法学习:周一 */public class TestCyclicBarrier { public static void main(String[] args) { CyclicBarrier barrier = new CyclicBarrier(20, ()->{ System.out.println("满人,发车"); }); for (int i = 0; i < 100; i++) { new Thread(()->{ try { barrier.await(); } catch (InterruptedException e) { e.printStackTrace(); } catch (BrokenBarrierException e) { e.printStackTrace(); } }).start(); } }}
CyclicBarrier大小定义为20,达到20个线程时才执行一次,否则始终期待着(barrier.await())。即以上程序会输入5次 满人,发车。
五、Phaser
/** * @author Java和算法学习:周一 */public class TestPhaser { private static MarriagePhaser marriagePhaser = new MarriagePhaser(); public static void sleep() { try { TimeUnit.MILLISECONDS.sleep(new Random().nextInt(1000)); } catch (InterruptedException e) { e.printStackTrace(); } } static class MarriagePhaser extends Phaser { @Override protected boolean onAdvance(int phase, int registeredParties) { switch (phase) { case 0: System.out.println("所有人到齐 " + registeredParties); System.out.println(); return false; case 1: System.out.println("所有人吃完 " + registeredParties); System.out.println(); return false; case 2: System.out.println("所有人来到 " + registeredParties); System.out.println(); return false; case 3: System.out.println("新郎新娘抱抱 " + registeredParties); return true; default: return true; } } } static class Person implements Runnable { String name; public Person(String name) { this.name = name; } private void arrive() { sleep(); System.out.println(name + " 达到现场"); marriagePhaser.arriveAndAwaitAdvance(); } private void eat() { sleep(); System.out.println(name + " 吃"); marriagePhaser.arriveAndAwaitAdvance(); } private void leave() { sleep(); System.out.println(name + " 来到"); marriagePhaser.arriveAndAwaitAdvance(); } private void hug() { if ("新郎".equals(name) || "新娘".equals(name)) { sleep(); System.out.println(name + " 拥抱"); marriagePhaser.arriveAndAwaitAdvance(); } else { marriagePhaser.arriveAndDeregister(); } } @Override public void run() { arrive(); eat(); leave(); hug(); } } public static void main(String[] args) { marriagePhaser.bulkRegister(7); for (int i = 0; i < 5; i++) { new Thread(new Person("person" + i)).start(); } new Thread(new Person("新郎")).start(); new Thread(new Person("新娘")).start(); }}
相似于栅栏组,分阶段的栅栏 。所有的线程都到了某个栅栏的时候,才会执行下一步的操作。
六、ReadWriteLock
Read的时候是共享锁
Write的时候是排它锁(互斥锁)
最开始如果是读线程拿到了锁,当第二个来的线程是读线程时,能够一起读;当第二个来的是写线程时则阻塞,不容许你写,等我读完再写。
最开始如果是写线程拿到了锁,不论第二个线程是读还是写,均阻塞,必须等我改完了其余线程能力读、其余线程能力写。
/** * @author Java和算法学习:周一 */public class TestReadWriteLock { private static int value; private static ReentrantLock lock = new ReentrantLock(); private static ReadWriteLock readWriteLock = new ReentrantReadWriteLock(); private static Lock readLock = readWriteLock.readLock(); private static Lock writeLock = readWriteLock.writeLock(); public static void read(Lock lock) { try { lock.lock(); Thread.sleep(1000); System.out.println("read..."); } catch (Exception e) { e.printStackTrace(); } finally { lock.unlock(); } } public static void write(Lock lock, int v) { try { lock.lock(); value = v; Thread.sleep(1000); System.out.println("write..." + value); } catch (Exception e) { e.printStackTrace(); } finally { lock.unlock(); } } public static void main(String[] args) throws InterruptedException { //读线程 8个 Thread[] tRead = new Thread[8]; for (int i = 0; i < 8; i++) { tRead[i] = new Thread(() -> { read(readLock); }); } long start = System.currentTimeMillis(); for (Thread t : tRead) { t.start(); } for (Thread t : tRead) { t.join(); } //写线程 2个 Thread[] tWrite = new Thread[2]; for (int i = 0; i < 2; i++) { tWrite[i] = new Thread(() -> { write(writeLock, new Random().nextInt(10)); }); } for (Thread t : tWrite) { t.start(); } for (Thread t : tWrite) { t.join(); } long end = System.currentTimeMillis(); System.out.println("total time: " + (end - start)); }}
如果应用Reentrantlock加锁,程序运行10s,也就是说Reentrantlock读写均是排它锁;
如果应用ReadWriteLock加锁,程序运行3s,也就是说ReadWriteLock的读是共享锁,写是排它锁。
七、Semaphore (信号灯)
初始时,给Semaphore赋默认值,该值示意最多能够容许多少个线程同时运行。即限流的意思。
Semaphore默认是非偏心锁,new Semaphore(1, true) new对象时设置true即示意是偏心锁。
/** * @author Java和算法学习:周一 */public class TestSemaphore { public static void main(String[] args) { Semaphore s = new Semaphore(1); new Thread(()->{ try { s.acquire(); System.out.println("t1..."); TimeUnit.SECONDS.sleep(1); System.out.println("t1..."); } catch (InterruptedException e) { e.printStackTrace(); } finally { s.release(); } }, "t1").start(); new Thread(()->{ try { s.acquire(); System.out.println("t2..."); TimeUnit.SECONDS.sleep(1); System.out.println("t2..."); } catch (InterruptedException e) { e.printStackTrace(); } finally { s.release(); } }, "t2").start(); }}
八、Exchanger
/** * @author Java和算法学习:周一 */public class TestExchanger { public static void main(String[] args) { Exchanger<String> exchanger = new Exchanger<>(); new Thread(() -> { String s = "T1"; try { s = exchanger.exchange(s); } catch (InterruptedException e) { e.printStackTrace(); } System.out.println(Thread.currentThread().getName() + " " + s); }, "t1").start(); new Thread(() -> { String s = "T2"; try { s = exchanger.exchange(s); } catch (InterruptedException e) { e.printStackTrace(); } System.out.println(Thread.currentThread().getName() + " " + s); }, "t2").start(); }}
exchange()办法是阻塞的,有一个线程调用了exchange()办法,它会始终阻塞直到第二个线程调用了exchange()办法;而后第一个线程才会往下执行。