Lock和synchronized的性能的比较

尽管synchronized在语法上已经足够简单了，在JDK 5之前只能借助此实现，但是由于是独占锁，
性能却不高，因此JDK 5以后就开始借助于JNI来完成更高级的锁实现。JDK 5中的锁是接口java.util.concurrent.locks.Lock。另外java.util.concurrent.locks.ReadWriteLock提供了一对可供读写并发的锁今天我们来比较一下，Lock和synchronized的性能；
//Lock 实现的AtomicInteger

package juc.automic;

import java.util.concurrent.locks.Lock;
import java.util.concurrent.locks.ReentrantLock;
/**
 * Lock 实现的AtomicInteger
 * @author donald
 * 2017年2月28日
 * 下午7:52:34
 */
public class AtomicIntegerWithLock {
	private int value;
	private Lock lock = new ReentrantLock();

	public AtomicIntegerWithLock() {
		super();
	}

	public AtomicIntegerWithLock(int value) {
		this.value = value;
	}

	public final int get() {
		lock.lock();
		try {
			return value;
		} finally {
			lock.unlock();
		}
	}

	public final void set(int newValue) {
		lock.lock();
		try {
			value = newValue;
		} finally {
			lock.unlock();
		}
	}

	public final int getAndSet(int newValue) {
		lock.lock();
		try {
			int ret = value;
			value = newValue;
			return ret;
		} finally {
			lock.unlock();
		}
	}

	public final boolean compareAndSet(int expect, int update) {
		lock.lock();
		try {
			if (value == expect) {
				value = update;
				return true;
			}
			return false;
		} finally {
			lock.unlock();
		}
	}

	public final int getAndIncrement() {
		lock.lock();
		try {
			return value++;
		} finally {
			lock.unlock();
		}
	}

	public final int getAndDecrement() {
		lock.lock();
		try {
			return value--;
		} finally {
			lock.unlock();
		}
	}

	public final int incrementAndGet() {
		lock.lock();
		try {
			return ++value;
		} finally {
			lock.unlock();
		}
	}

	public final int decrementAndGet() {
		lock.lock();
		try {
			return --value;
		} finally {
			lock.unlock();
		}
	}

	public String toString() {
		return Integer.toString(get());
	}

}

//测试主类

package juc.automic;


/**
 * Lock和synchronized的性能比较测试
 * @author donald
 * 2017年2月28日
 * 下午7:53:04
 */
public class LockCompareSynchronize {
	static int staticValue = 0;

	public static void main(String[] args) throws Exception {
		final int max = 10;
		final int loopCount = 100000;
		long costTime = 0;
		for (int m = 0; m < max; m++) {
			long start0 = System.nanoTime();
			final AtomicIntegerWithLock value1 = new AtomicIntegerWithLock(0);
			Thread[] ts = new Thread[max];
			for (int i = 0; i < max; i++) {
				ts[i] = new Thread() {
					public void run() {
						for (int i = 0; i < loopCount; i++) {
							value1.incrementAndGet();
						}
					}
				};
			}
			for (Thread t : ts) {
				t.start();
			}
			for (Thread t : ts) {
				t.join();
			}
			long end0 = System.nanoTime();
			costTime += (end0 - start0);
		}
		System.out.println("Lock所耗时间: " + (costTime));
		System.out.println();
		costTime = 0;

		final Object lock = new Object();
		for (int m = 0; m < max; m++) {
			staticValue = 0;
			long start1 = System.nanoTime();
			Thread[] ts = new Thread[max];
			for (int i = 0; i < max; i++) {
				ts[i] = new Thread() {
					public void run() {
						for (int i = 0; i < loopCount; i++) {
							synchronized (lock) {
								++staticValue;
							}
						}
					}
				};
			}
			for (Thread t : ts) {
				t.start();
			}
			for (Thread t : ts) {
				t.join();
			}
			long end1 = System.nanoTime();
			costTime += (end1 - start1);
		}
		System.out.println("synchronized所耗时间: " + (costTime));
	}

}

控制台输出：

Lock所耗时间: 343165627
synchronized所耗时间: 436721381

尽管上面的例子不是非常正式的测试案例，但上面的例子在于说明，Lock的性能比synchronized的要好得多。如果可以的话总是使用Lock替代synchronized是一个明智的选择。

再来看一下ReentrantLock
// ReentrantLock

public class ReentrantLock implements Lock, java.io.Serializable {
    private static final long serialVersionUID = 7373984872572414699L;
    //锁机制
    /** Synchronizer providing all implementation mechanics */
    private final Sync sync;
    abstract static class Sync extends AbstractQueuedSynchronizer {
     }
}

//创建锁

  public ReentrantLock() {
        sync = new NonfairSync();
    }

//非公平锁，ReentrantLock的内部类NonfairSync

/**
     * Sync object for non-fair locks
     */
    static final class NonfairSync extends Sync {
        private static final long serialVersionUID = 7316153563782823691L;

        /**
         * Performs lock.  Try immediate barge, backing up to normal
         * acquire on failure.
         */
        final void lock() {
	    //获取锁
            if (compareAndSetState(0, 1))
	        //设置锁持有者线程
                setExclusiveOwnerThread(Thread.currentThread());
            else
	        //获取失败，添加等待队列
                acquire(1);
        }
	//尝试获取锁
        protected final boolean tryAcquire(int acquires) {
            return nonfairTryAcquire(acquires);
        }
    }

来看lock中的方法compareAndSetState
//AbstractQueuedSynchronizer

public abstract class AbstractQueuedSynchronizer
    extends AbstractOwnableSynchronizer
    implements java.io.Serializable {
  protected final boolean compareAndSetState(int expect, int update) {
        // See below for intrinsics setup to support this
	//通过unsafe实现
        return unsafe.compareAndSwapInt(this, stateOffset, expect, update);
    }
}

//创建公平锁：

 /**
     * Creates an instance of {@code ReentrantLock} with the
     * given fairness policy.
     *
     * @param fair {@code true} if this lock should use a fair ordering policy
     */
    public ReentrantLock(boolean fair) {
        sync = fair ? new FairSync() : new NonfairSync();
    } 

//公平锁

static final class FairSync extends Sync {
        private static final long serialVersionUID = -3000897897090466540L;

        final void lock() {
            acquire(1);
        }

        /**
         * Fair version of tryAcquire.  Don't grant access unless
         * recursive call or no waiters or is first.
         */
        protected final boolean tryAcquire(int acquires) {
            final Thread current = Thread.currentThread();
            int c = getState();
            if (c == 0) {
                if (!hasQueuedPredecessors() &&
                    compareAndSetState(0, acquires)) {
                    setExclusiveOwnerThread(current);
                    return true;
                }
            }
            else if (current == getExclusiveOwnerThread()) {
                int nextc = c + acquires;
                if (nextc < 0)
                    throw new Error("Maximum lock count exceeded");
                setState(nextc);
                return true;
            }
            return false;
        }
    }

//AbstractOwnableSynchronizer

public abstract class AbstractOwnableSynchronizer
    implements java.io.Serializable {

    /** Use serial ID even though all fields transient. */
    private static final long serialVersionUID = 3737899427754241961L;

    /**
     * Empty constructor for use by subclasses.
     */
    protected AbstractOwnableSynchronizer() { }

    /**
     * The current owner of exclusive mode synchronization.
     */
    private transient Thread exclusiveOwnerThread;

    /**
     * Sets the thread that currently owns exclusive access. A
     * <tt>null</tt> argument indicates that no thread owns access.
     * This method does not otherwise impose any synchronization or
     * <tt>volatile</tt> field accesses.
     */
    protected final void setExclusiveOwnerThread(Thread t) {
        exclusiveOwnerThread = t;
    }

    /**
     * Returns the thread last set by
     * <tt>setExclusiveOwnerThread</tt>, or <tt>null</tt> if never
     * set.  This method does not otherwise impose any synchronization
     * or <tt>volatile</tt> field accesses.
     * @return the owner thread
     */
    protected final Thread getExclusiveOwnerThread() {
        return exclusiveOwnerThread;
    }
}

[color=green]在公平的锁上，线程按照他们发出请求的顺序获取锁，但在非公平锁上，则允许‘插队’：
当一个线程请求非公平锁时，如果在发出请求的同时该锁变成可用状态，那么这个线程会跳过队列中所有的等待线程而获得锁。
非公平的ReentrantLock 并不提倡 插队行为，但是无法防止某个线程在合适的时候进行插队。
在公平的锁中，如果有另一个线程持有锁或者有其他线程在等待队列中等待这个所，
那么新发出的请求的线程将被放入到队列中。而非公平锁上，只有当锁被某个线程持有时，新发出请求的线程才会被放入队列中。

非公平锁性能高于公平锁性能的原因：
在恢复一个被挂起的线程与该线程真正运行之间存在着严重的延迟。
假设线程A持有一个锁，并且线程B请求这个锁。由于锁被A持有，因此B将被挂起。当A释放锁时，B将被唤醒，因此B会再次尝试获取这个锁。与此同时，如果线程C也请求这个锁，那么C很可能会在B被完全唤醒之前获得、使用以及释放这个锁。这样就是一种双赢的局面：B获得锁的时刻并没有推迟，C更早的获得了锁，并且吞吐量也提高了。当持有锁的时间相对较长或者请求锁的平均时间间隔较长，应该使用公平锁。在这些情况下，插队带来的吞吐量提升（当锁处于可用状态时，线程却还处于被唤醒的过程中）可能不会出现。[/color]