CAS和Atomic原子操作

CAS

指的是针对一个变量，首先比较它的内存值与某个期望值是否相同，如果相同，就给它赋一个新值。可以看做是乐观锁的一种实现方式，java中的原子类操作就是通过CAS来完成，它是一种无锁的算法，在不使用锁(发生线程阻塞)的情况下实现多线程之间的同步。

实现方法

CAS的操作来源于Unsafe类，具体有三个方法进行操作：

//针对对象的比较更新
public final native boolean compareAndSwapObject(Object var1, long var2, Object var4, Object var5);
//针对int类型的比较更新
public final native boolean compareAndSwapInt(Object var1, long var2, int var4, int var5);
//针对Long类型的比较更新
public final native boolean compareAndSwapLong(Object var1, long var2, long var4, long var6);

主要有四个参数，分别是对象实例、偏移量、旧的值、新的值

工具类

    /**
     * 获取 Unsafe 对象
     * @return
     */
    public static Unsafe getUnsafe() {
        try {
            Field field = Unsafe.class.getDeclaredField("theUnsafe");
            field.setAccessible(true);
            return (Unsafe) field.get(null);
        } catch (Exception e) {
            e.printStackTrace();
        }
        return null;
    }

    /**
     * 获取字段的内存偏移量
     * @param unsafe
     * @param clazz
     * @param fieldName
     * @return
     */
    public static long getFieldOffset(Unsafe unsafe, Class clazz, String fieldName) {
        try {
            return unsafe.objectFieldOffset(clazz.getDeclaredField(fieldName));
        } catch (NoSuchFieldException e) {
            throw new Error(e);
        }
    }

 

测试

public static void main(String[] args) {
        Entity entity = new Entity();

        Unsafe unsafe = UnsafeFactory.getUnsafe();

        long offset = UnsafeFactory.getFieldOffset(unsafe, Entity.class, "x");
        System.out.println(offset);
        boolean successful;

        // 4个参数分别是：对象实例、字段的内存偏移量、字段期望值、字段更新值
        successful = unsafe.compareAndSwapInt(entity, offset, 0, 3);
        System.out.println(successful + "\t" + entity.x);

        successful = unsafe.compareAndSwapInt(entity, offset, 3, 5);
        System.out.println(successful + "\t" + entity.x);

        successful = unsafe.compareAndSwapInt(entity, offset, 3, 8);
        System.out.println(successful + "\t" + entity.x);

    }
}

class Entity{
    int x;
    int y;
}

结果

CAS优点

它是一种无锁的操作，相比较syschronized和lock来说，不会发生线程的上下文切换，不会有程序从用户态到内核态的转变，很高效地解决了原子操作，在性能上更优。

CAS缺陷

• ABA问题
• 自旋CAS长时间不成功，会给CPU带来很大的开销
• 只能针对一个共享变量的原子操作

ABA问题

共享变量的值被另一线程操作，由A-->B-->A,但对第一个拿到值为A的线程来说，变化是无感的。

@Slf4j
public class ABATest {

    public static void main(String[] args) {
        AtomicInteger atomicInteger = new AtomicInteger(1);

        new Thread(()->{
            int value = atomicInteger.get();
            log.debug("Thread1 read value: " + value);

            // 阻塞1s
            LockSupport.parkNanos(1000000000L);

            // Thread1通过CAS修改value值为3
            if (atomicInteger.compareAndSet(value, 3)) {
                log.debug("Thread1 update from " + value + " to 3");
            } else {
                log.debug("Thread1 update fail!");
            }
        },"Thread1").start();

        new Thread(()->{
            int value = atomicInteger.get();
            log.debug("Thread2 read value: " + value);
            // Thread2通过CAS修改value值为2
            if (atomicInteger.compareAndSet(value, 2)) {
                log.debug("Thread2 update from " + value + " to 2");

                // do something
                value = atomicInteger.get();
                log.debug("Thread2 read value: " + value);
                // Thread2通过CAS修改value值为1
                if (atomicInteger.compareAndSet(value, 1)) {
                    log.debug("Thread2 update from " + value + " to 1");
                }
            }
        },"Thread2").start();
    }
}

执行结果

解决方式，

通过添加版本号来做更新，即在判断值相等的前提下，再判断版本号是否一致，一致才能更新，否则更新失败，其具体实现类为AtomicStampedReference。

package java.util.concurrent.atomic;

/**
 * An {@code AtomicStampedReference} maintains an object reference
 * along with an integer "stamp", that can be updated atomically.
 *
 * <p>Implementation note: This implementation maintains stamped
 * references by creating internal objects representing "boxed"
 * [reference, integer] pairs.
 *
 * @since 1.5
 * @author Doug Lea
 * @param <V> The type of object referred to by this reference
 */
public class AtomicStampedReference<V> {

    private static class Pair<T> {
        final T reference;
        //版本号
        final int stamp;
        private Pair(T reference, int stamp) {
            this.reference = reference;
            this.stamp = stamp;
        }
        static <T> Pair<T> of(T reference, int stamp) {
            return new Pair<T>(reference, stamp);
        }
    }

    private volatile Pair<V> pair;

    /**
     * Creates a new {@code AtomicStampedReference} with the given
     * initial values.
     *
     * @param initialRef the initial reference
     * @param initialStamp the initial stamp
     */
    public AtomicStampedReference(V initialRef, int initialStamp) {
        pair = Pair.of(initialRef, initialStamp);
    }

    /**
     * Returns the current value of the reference.
     *
     * @return the current value of the reference
     */
    public V getReference() {
        return pair.reference;
    }

    /**
     * Returns the current value of the stamp.
     *
     * @return the current value of the stamp
     */
    public int getStamp() {
        return pair.stamp;
    }

    /**
     * Returns the current values of both the reference and the stamp.
     * Typical usage is {@code int[1] holder; ref = v.get(holder); }.
     *
     * @param stampHolder an array of size of at least one.  On return,
     * {@code stampholder[0]} will hold the value of the stamp.
     * @return the current value of the reference
     */
    public V get(int[] stampHolder) {
        Pair<V> pair = this.pair;
        stampHolder[0] = pair.stamp;
        return pair.reference;
    }

    /**
     * Atomically sets the value of both the reference and stamp
     * to the given update values if the
     * current reference is {@code ==} to the expected reference
     * and the current stamp is equal to the expected stamp.
     *
     * <p><a href="package-summary.html#weakCompareAndSet">May fail
     * spuriously and does not provide ordering guarantees</a>, so is
     * only rarely an appropriate alternative to {@code compareAndSet}.
     *
     * @param expectedReference the expected value of the reference
     * @param newReference the new value for the reference
     * @param expectedStamp the expected value of the stamp
     * @param newStamp the new value for the stamp
     * @return {@code true} if successful
     */
    public boolean weakCompareAndSet(V   expectedReference,
                                     V   newReference,
                                     int expectedStamp,
                                     int newStamp) {
        return compareAndSet(expectedReference, newReference,
                             expectedStamp, newStamp);
    }

    /**
     * Atomically sets the value of both the reference and stamp
     * to the given update values if the
     * current reference is {@code ==} to the expected reference
     * and the current stamp is equal to the expected stamp.
     *
     * @param expectedReference the expected value of the reference
     * @param newReference the new value for the reference
     * @param expectedStamp the expected value of the stamp
     * @param newStamp the new value for the stamp
     * @return {@code true} if successful
     */
    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)));
    }

    /**
     * Unconditionally sets the value of both the reference and stamp.
     *
     * @param newReference the new value for the reference
     * @param newStamp the new value for the stamp
     */
    public void set(V newReference, int newStamp) {
        Pair<V> current = pair;
        if (newReference != current.reference || newStamp != current.stamp)
            this.pair = Pair.of(newReference, newStamp);
    }

    /**
     * Atomically sets the value of the stamp to the given update value
     * if the current reference is {@code ==} to the expected
     * reference.  Any given invocation of this operation may fail
     * (return {@code false}) spuriously, but repeated invocation
     * when the current value holds the expected value and no other
     * thread is also attempting to set the value will eventually
     * succeed.
     *
     * @param expectedReference the expected value of the reference
     * @param newStamp the new value for the stamp
     * @return {@code true} if successful
     */
    public boolean attemptStamp(V expectedReference, int newStamp) {
        Pair<V> current = pair;
        return
            expectedReference == current.reference &&
            (newStamp == current.stamp ||
             casPair(current, Pair.of(expectedReference, newStamp)));
    }

    // Unsafe mechanics

    private static final sun.misc.Unsafe UNSAFE = sun.misc.Unsafe.getUnsafe();
    private static final long pairOffset =
        objectFieldOffset(UNSAFE, "pair", AtomicStampedReference.class);

    private boolean casPair(Pair<V> cmp, Pair<V> val) {
        return UNSAFE.compareAndSwapObject(this, pairOffset, cmp, val);
    }

    static long objectFieldOffset(sun.misc.Unsafe UNSAFE,
                                  String field, Class<?> klazz) {
        try {
            return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
        } catch (NoSuchFieldException e) {
            // Convert Exception to corresponding Error
            NoSuchFieldError error = new NoSuchFieldError(field);
            error.initCause(e);
            throw error;
        }
    }
}

public static void main(String[] args) {
        // 定义AtomicStampedReference    Pair.reference值为1, Pair.stamp为1
        AtomicStampedReference atomicStampedReference = new AtomicStampedReference(1,1);

        new Thread(()->{
            int[] stampHolder = new int[1];
            int value = (int) atomicStampedReference.get(stampHolder);
            int stamp = stampHolder[0];
            log.debug("Thread1 read value: " + value + ", stamp: " + stamp);

            // 阻塞1s
            LockSupport.parkNanos(1000000000L);
            // Thread1通过CAS修改value值为3   stamp是版本，每次修改可以通过+1保证版本唯一性
            if (atomicStampedReference.compareAndSet(value, 3,stamp,stamp+1)) {
                log.debug("Thread1 update from " + value + " to 3");
            } else {
                log.debug("Thread1 update fail!");
            }
        },"Thread1").start();

        new Thread(()->{
            int[] stampHolder = new int[1];
            int value = (int)atomicStampedReference.get(stampHolder);
            int stamp = stampHolder[0];
            log.debug("Thread2 read value: " + value+ ", stamp: " + stamp);
            // Thread2通过CAS修改value值为2
            if (atomicStampedReference.compareAndSet(value, 2,stamp,stamp+1)) {
                log.debug("Thread2 update from " + value + " to 2");

                // do something

                value = (int) atomicStampedReference.get(stampHolder);
                stamp = stampHolder[0];
                log.debug("Thread2 read value: " + value+ ", stamp: " + stamp);
                // Thread2通过CAS修改value值为1
                if (atomicStampedReference.compareAndSet(value, 1,stamp,stamp+1)) {
                    log.debug("Thread2 update from " + value + " to 1");
                }
            }
        },"Thread2").start();
    }

补充：使用AtomicMarkableReference也可以解决ABA问题，它和AtomicStampReference的区别在于，它只判断共享变量的值是否有过更改，若有，则更新失败

长时间自旋失败

    public final long getAndAddLong(Object var1, long var2, long var4) {
        long var6;
        do {
            var6 = this.getLongVolatile(var1, var2);
        } while(!this.compareAndSwapLong(var1, var2, var6, var6 + var4));

        return var6;
    }

上述方法会用自旋的方式不断更新目标值，直到更新成功。

在并发量较低的环境下，线程冲突的概率比较小，自旋的次数不会很多。但是，高并发环境下， N个线程同时进行自旋操作，会出现大量失败并不断自旋的情况，此时AtomicLong的自旋会成为瓶颈。

解决方法

使用LongAdder

public static void main(String[] args) {
        testAtomicLongVSLongAdder(10, 10000);
        System.out.println("==================");
        testAtomicLongVSLongAdder(10, 200000);
        System.out.println("==================");
        testAtomicLongVSLongAdder(100, 200000);
    }

    static void testAtomicLongVSLongAdder(final int threadCount, final int times) {
        try {
            long start = System.currentTimeMillis();
            testLongAdder(threadCount, times);
            long end = System.currentTimeMillis() - start;
            System.out.println("条件>>>>>>线程数:" + threadCount + ", 单线程操作计数" + times);
            System.out.println("结果>>>>>>LongAdder方式增加计数" + (threadCount * times) + "次,共计耗时:" + end);

            long start2 = System.currentTimeMillis();
            testAtomicLong(threadCount, times);
            long end2 = System.currentTimeMillis() - start2;
            System.out.println("条件>>>>>>线程数:" + threadCount + ", 单线程操作计数" + times);
            System.out.println("结果>>>>>>AtomicLong方式增加计数" + (threadCount * times) + "次,共计耗时:" + end2);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
    }

    static void testAtomicLong(final int threadCount, final int times) throws InterruptedException {
        CountDownLatch countDownLatch = new CountDownLatch(threadCount);
        AtomicLong atomicLong = new AtomicLong();
        for (int i = 0; i < threadCount; i++) {
            new Thread(new Runnable() {
                @Override
                public void run() {
                    for (int j = 0; j < times; j++) {
                        atomicLong.incrementAndGet();
                    }
                    countDownLatch.countDown();
                }
            }, "my-thread" + i).start();
        }
        countDownLatch.await();
    }

    static void testLongAdder(final int threadCount, final int times) throws InterruptedException {
        CountDownLatch countDownLatch = new CountDownLatch(threadCount);
        LongAdder longAdder = new LongAdder();
        for (int i = 0; i < threadCount; i++) {
            new Thread(new Runnable() {
                @Override
                public void run() {
                    for (int j = 0; j < times; j++) {
                        longAdder.add(1);
                    }
                    countDownLatch.countDown();
                }
            }, "my-thread" + i).start();
        }

        countDownLatch.await();
    }

从结果来看，使用LongAddder在效率上要比Atomic高很多

LongAddder原理

AtomicLong中有个来自Striped64的base变量和Cell[]数组，base保存着实际的long值，高并发环境下，多个线程竞争的就是这个base。LongAdder的基本思路就是分段治之，将value值分散到一个Cell数组中，不同线程会命中到数组的不同槽中，各个线程只对自己槽中的那个值进行CAS操作，以减少竞争，如果要获取真正的long值，只要将各个槽中的变量值累加返回。

add方法

public void add(long x) {
        Cell[] as; long b, v; int m; Cell a;
        if ((as = cells) != null || !casBase(b = base, b + x)) {
            boolean uncontended = true;
            if (as == null || (m = as.length - 1) < 0 ||
                (a = as[getProbe() & m]) == null ||
                !(uncontended = a.cas(v = a.value, v + x)))
                longAccumulate(x, null, uncontended);
        }
    }

多个共享变量原子操作

使用对象操作，将多个共享变量包装为对象，使用AtomicReference

Atomic

在并发编程的过程中，因为多线程的问题很可能会出现并发安全问题；因此，大部分时候通过Syschrnozied来加锁以达到线程安全，但它采用的是悲观锁(有访问即阻塞)的策略，在效率上不是很高。因此，在JUC下的atomic包下的一系列针对基本类型变量、数组元素、引用类型以及对象中的字段类型的采用了乐观锁的原子操作类就称为了一种较为高效的解决方案。

列举如下

• 基本类型：AtomicInteger、AtomicLong、AtomicBoolean；
• 引用类型：AtomicReference、AtomicStampedRerence、AtomicMarkableReference；
• 数组类型：AtomicIntegerArray、AtomicLongArray、AtomicReferenceArray
• 对象属性原子修改器：AtomicIntegerFieldUpdater、AtomicLongFieldUpdater、AtomicReferenceFieldUpdater
• 原子类型累加器（jdk1.8增加的类）：DoubleAccumulator、DoubleAdder、LongAccumulator、LongAdder、Striped64