这个问题,在网络各大论坛上面,很多人做出了自己的解答,给出的答案大多是否定的。
first example:
the second:
the third(被我看到逻辑不太对,所以拿来自己测试):
先说一下,为什么它逻辑不对,因为他每次唤醒一次,主线程就要争用锁一次。那么主线程和被唤醒的线程在争用锁呢,怎么可能得出不受干扰的结果呢?
the last(黑马程序员):
我想应该是他们没有理解从waitSet里面取出线程唤醒和唤醒后争用到锁是两回事。而他们只看锁争用的结果表象的话,就误认为,线程唤醒是随机的。
下面是我的测试代码:
public void test() throws InterruptedException {
Object ref = new Object();
IntStream.rangeClosed(1, 10).forEach(i ->
{
new Thread(String.valueOf(i)) {
@Override
public void run() {
synchronized (ref) {
try {
System.out.println(Thread.currentThread().getName() + " will come to wait set.");
ref.wait();
System.err.println(Thread.currentThread().getName() + " will leave to wait set.");
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
}.start();
try {
Thread.sleep(1000);
}catch (Exception e){
e.printStackTrace();
}
});
Thread.sleep(3000);
synchronized (ref) {
ref.notify();
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
执行了很多次,得到的结果,我贴到下面:
1 will come to wait set.
2 will come to wait set.
3 will come to wait set.
4 will come to wait set.
5 will come to wait set.
6 will come to wait set.
7 will come to wait set.
8 will come to wait set.
9 will come to wait set.
10 will come to wait set.
唤醒了第一个
1 will leave to wait set.
从结果来看,调用obj.notify()唤醒的就是第一个进入entrySet的线程。
下面我将唤醒逻辑改成将所有的waiting线程都唤醒是什么状况:
// 修改唤醒的逻辑代码
synchronized (ref) {
for (int i = 0; i < 10; i++) {
ref.notify();
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
System.out.println("唤醒了第一个");
}
运行得到的结果:
1 will come to wait set.
2 will come to wait set.
3 will come to wait set.
4 will come to wait set.
5 will come to wait set.
6 will come to wait set.
7 will come to wait set.
8 will come to wait set.
9 will come to wait set.
10 will come to wait set.
1 will leave to wait set.
10 will leave to wait set.
9 will leave to wait set.
8 will leave to wait set.
7 will leave to wait set.
6 will leave to wait set.
5 will leave to wait set.
4 will leave to wait set.
3 will leave to wait set.
2 will leave to wait set.
不管执行多少次,运行得到的结果总是同一个。说明就算全部唤醒,最先被唤醒是永远是第一个进入waitSet的线程。但是后面的线程的为什么开始倒数了呢?我们来看下notify()的策略,和锁释放后,线程争用锁的策略:
notify() 根据Policy 做不同的操作
Policy==0 :放入到entrylist队列的排头位置
Policy==1 :放入到entrylist队列的末尾位置
Policy==2 :判断entrylist是否为空,为空就放入entrylist中,否则放入cxq队列排头位置(默认
策略)
Policy==3 :判断cxq是否为空,如果为空,直接放入头部,否则放入cxq队列末尾位置
当同步代码块执行完毕,唤醒是什么样子的呢?根据Qmode进行选择
根据QMode策略唤醒:
QMode=2,取cxq头部节点直接唤醒
QMode=3,如果cxq非空,把cxq队列放置到entrylist的尾部(顺序跟cxq一致)
QMode=4,如果cxq非空,把cxq队列放置到entrylist的头部(顺序跟cxq相反)
QMode=0,啥都不做,继续往下走(QMode默认是0)默认是0
Qmode=0的判断逻辑就是先判断entrylist是否为空,如果不为空,则取出第一个唤醒,如
果为空,将cxq里面线程按照头尾顺序,全量放入EntryList,再从EntryList中获取第一个唤醒。
在默认策略policy == 2的情况下,waitSet取出的线程放到entryList(待入队列)之后,entryList不为空,那么剩下的线程都会进入到cxq(contention List争用队列),在默认策略下,cxq变成了FILO的,也就是说它此时的逻辑是个栈(从排头入,从排头取),先入后出,所以后面的线程拿到锁的顺序是倒序的。
当我将notify()改成notifyAll()之后:
synchronized (ref) {
for (int i = 0; i < 1; i++) {
ref.notifyAll();
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
运行的结果:
1 will come to wait set.
2 will come to wait set.
3 will come to wait set.
4 will come to wait set.
5 will come to wait set.
6 will come to wait set.
7 will come to wait set.
8 will come to wait set.
9 will come to wait set.
10 will come to wait set.
10 will leave to wait set.
9 will leave to wait set.
8 will leave to wait set.
7 will leave to wait set.
6 will leave to wait set.
5 will leave to wait set.
4 will leave to wait set.
3 will leave to wait set.
2 will leave to wait set.
1 will leave to wait set.
说明notifyAll()的策略跟notify的不一样,因为唤醒的线程比较多,不会给任何一个待唤醒的线程优先权而进入EntryList,所以它跟线程正常的争用锁机制一样,不会管entryList是否为空,直接一股脑全部从排头放入cxq了,然后这段代码释放锁之后,cxq里面线程按照头尾顺序,全量放入EntryList,再从EntryList中获取第一个唤醒。产生了如上结果。
如果你觉得我的博客内容对你有帮助,请关注我的公众号,更多的学习资料,源码解析,设计思想,方案推陈,都可以与你分享,共同进步。