原子性-Atomic类常见的使用 1、(信号量)Semaphore+(闭锁)CountDownLatch+(常用源自类)AtomicIntegerpackage com.yanxizhu.demo.concurrency.atomic; import com.yanxizhu.demo.concurrency.annotation.ThreadSafety; import lombok.extern.slf4j.Slf4j; import java.util.concurrent.CountDownLatch; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; import java.util.concurrent.Semaphore; import java.util.concurrent.atomic.AtomicInteger; /** * @description: 线程安全 * @author: <a href="mailto:batis@foxmail.com">清风</a> * @date: 2022/4/19 20:05 * @version: 1.0 */ @Slf4j @ThreadSafety public class DemoSemaphoreAndCountDownLatchAndAtomic { //用户数量 private static final int clientsTotal = 5000; //并发数量 private static final int concurrencyTotal = 200; //累加总和 private static AtomicInteger count = new AtomicInteger(0); public static void main(String[] args) throws InterruptedException { ExecutorService executorService = Executors.newCachedThreadPool(); //信号量 final Semaphore semaphore = new Semaphore(clientsTotal); //闭锁 final CountDownLatch countDownLatch = new CountDownLatch(concurrencyTotal); for (int i = 0; i < clientsTotal; i++) { executorService.execute(()->{ try { semaphore.acquire(); add(); semaphore.release(); } catch (InterruptedException e) { log.error("出现错误:【{}】", e.getMessage()); } countDownLatch.countDown(); }); } countDownLatch.await(); executorService.shutdown(); log.info("累加结果为count=【{}】",count.get()); } /** * 累加 */ private static void add(){ //说明:incrementAndGet、getAndIncrement类似,i++与++i count.incrementAndGet(); count.getAndIncrement(); } }运行结果：累加结果为count=【10000】2、AtomicReferencepackage com.yanxizhu.demo.concurrency.atomic; import com.yanxizhu.demo.concurrency.annotation.ThreadSafety; import lombok.extern.slf4j.Slf4j; import java.util.concurrent.atomic.AtomicReference; /** * @description: AtomicReference * @author: <a href="mailto:batis@foxmail.com">清风</a> * @date: 2022/4/20 21:09 * @version: 1.0 */ @Slf4j @ThreadSafety public class DemoAtomicReference { public static AtomicReference<Integer> count = new AtomicReference<>(0); public static void main(String[] args) { count.compareAndSet(0, 2); count.compareAndSet(0, 1); count.compareAndSet(1, 3); count.compareAndSet(2, 4); count.compareAndSet(3, 5); log.info("count结果为【{}】",count.get()); } }运行结果：count结果为【4】3、AtomicIntegerFieldUpdaterpackage com.yanxizhu.demo.concurrency.atomic; import lombok.Getter; import lombok.extern.slf4j.Slf4j; import java.util.concurrent.atomic.AtomicIntegerFieldUpdater; /** * @description: AtomicIntegerFieldUpdater,原子性修改 * @author: <a href="mailto:batis@foxmail.com">清风</a> * @date: 2022/4/20 21:17 * @version: 1.0 */ @Slf4j public class DemoAtomicIntegerFieldUpdater { private static AtomicIntegerFieldUpdater<DemoAtomicIntegerFieldUpdater> updater = AtomicIntegerFieldUpdater.newUpdater(DemoAtomicIntegerFieldUpdater.class, "count"); @Getter public volatile int count = 100; private static DemoAtomicIntegerFieldUpdater demoAtomicIntegerFieldUpdater = new DemoAtomicIntegerFieldUpdater(); public static void main(String[] args) { if(updater.compareAndSet(demoAtomicIntegerFieldUpdater, 100, 200)) { log.info("updater succcess1,{}", demoAtomicIntegerFieldUpdater.getCount()); } if(updater.compareAndSet(demoAtomicIntegerFieldUpdater, 100, 200)) { log.info("updater succcess2,{}", demoAtomicIntegerFieldUpdater.getCount()); }else { log.info("updater fail,{}", demoAtomicIntegerFieldUpdater.getCount()); } } }运行结果：updater succcess1,200 updater fail,2004、AtomicStampedReference主要用于解决CAS中ABA问题，主要是AtomicStampedReference类中compareAndSet方法多了一个stamp的比较。stamp是每次更新来维护的。 public boolean compareAndSet(V expectedReference, V newReference, int expectedStamp, int newStamp) { Pair<V> current = pair; return expectedReference == current.reference && expectedStamp == current.stamp && ((newReference == current.reference && newStamp == current.stamp) || casPair(current, Pair.of(newReference, newStamp))); } 5、LongAdderJDK1.8后新增了LongAdder。AtomicLong：是基于 CAS 方式自旋更新的；LongAdder： 是把 value 分成若干cell，并发量低的时候，直接 CAS 更新值，成功即结束。并发量高的情况，CAS更新某个cell值和需要时对cell数据扩容，成功结束；更新失败自旋 CAS 更新 cell值。取值的时候，调用 sum() 方法进行每个cell累加。AtomicLong： 包含有原子性的读、写结合的api；LongAdder ：没有原子性的读、写结合的api，能保证结果最终一致性。低并发：低并发场景AtomicLong 和 LongAdder 性能相似。高并发：高并发场景 LongAdder 性能优于 AtomicLong。6、AtomicLongArray主要用于原子修改数组 // 下标， 期望值 更新值 public final boolean compareAndSet(int i, long expectedValue, long newValue) { return AA.compareAndSet(array, i, expectedValue, newValue); }通过传入下标、期望值、要更新的值进行修改。7、AtomicBoolean public final boolean compareAndSet(boolean expectedValue, boolean newValue) { return VALUE.compareAndSet(this, (expectedValue ? 1 : 0), (newValue ? 1 : 0)); }package com.yanxizhu.demo.concurrency.atomic; import lombok.extern.slf4j.Slf4j; import java.util.concurrent.CountDownLatch; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; import java.util.concurrent.Semaphore; import java.util.concurrent.atomic.AtomicBoolean; /** * @description: 结果，最后只会执行一次 * @author: <a href="mailto:batis@foxmail.com">清风</a> * @date: 2022/4/20 22:05 * @version: 1.0 */ @Slf4j public class DemoAtomicBoolean { private static AtomicBoolean isHappen = new AtomicBoolean(false); //用户数量 private static final int clientsTotal = 5000; //并发数量 private static final int concurrencyTotal = 200; public static void main(String[] args) throws InterruptedException { ExecutorService executorService = Executors.newCachedThreadPool(); //信号量 final Semaphore semaphore = new Semaphore(clientsTotal); //闭锁 final CountDownLatch countDownLatch = new CountDownLatch(concurrencyTotal); for (int i = 0; i < clientsTotal; i++) { executorService.execute(()->{ try { semaphore.acquire(); test(); semaphore.release(); } catch (InterruptedException e) { log.error("出现错误:【{}】", e.getMessage()); } countDownLatch.countDown(); }); } countDownLatch.await(); executorService.shutdown(); log.info("isHappen结果为count=【{}】",isHappen.get()); } public static void test() { //从false变成true是一次原子操作，只会执行一次 if(isHappen.compareAndSet(false, true)) { log.info("execute .. {}", isHappen); } } }运行结果：execute .. true isHappen结果为count=【true】

AtomicStampedReference解决CAS中ABA问题 解决CAS的ABA问题CAS虽然高效的实现了原子性操作，但是也存在一些缺点，主要表现在以下三个方面。什么是ABA问题比如：线程1从主存中读取值A，另一个线程2也从主存中读取值A，此时线程2将主存值修改成了B，然后线程2又将主存值修改成了A，这时候线程1进行CAS操作发现内存中仍然是A，然后线程1操作成功，这是不正确的。如何解决ABA问题通过AtomicStampedReference的compareAndSet方法进行处理，这里的compareAndSet比一般的多了一个stamp的比较。源码如下： public boolean compareAndSet(V expectedReference, V newReference, int expectedStamp, int newStamp) { Pair<V> current = pair; return expectedReference == current.reference && expectedStamp == current.stamp && ((newReference == current.reference && newStamp == current.stamp) || casPair(current, Pair.of(newReference, newStamp))); }多了一个stamp的比较，stamp是每次更新来维护的。