Java线程池ThreadPoolExecutor的execute()原理
从线程提交任务的submit方法说起
众所周知，Java的线程池底层是使用工作队列+阻塞队列来实现的。
当工作队列（workQueue）任务数大于线程池的最大线程数时，任务阻塞，被放入一个阻塞队列（blockingQueue）中。
那么，在任务提交到线程开始工作这段时间里发生了什么呢？
下面是submit方法的源码，对Runnable接口的对象进行类封装，最后通过execute方法执行任务。
1
/**
2
* @throws RejectedExecutionException {@inheritDoc}
3
* @throws NullPointerException {@inheritDoc}
4
*/
5
public Future submit(Runnable task) {
6
if (task == null) throw new NullPointerException();
7
RunnableFuture<Void> ftask = newTaskFor(task, null);
8
execute(ftask);
9
return ftask;
10
}
我们来看看execute方法是怎么执行任务的（ThreadPoolExecutor.java）
1
/**
2
* Executes the given task sometime in the future. The task
3
* may execute in a new thread or in an existing pooled thread.
4
*
5
* If the task cannot be submitted for execution, either because this
6
* executor has been shutdown or because its capacity has been reached,
7
* the task is handled by the current {@code RejectedExecutionHandler}.
8
*
9
* @param command the task to execute
10
* @throws RejectedExecutionException at discretion of
11
* {@code RejectedExecutionHandler}, if the task
12
* cannot be accepted for execution
13
* @throws NullPointerException if {@code command} is null
14
*/
15
public void execute(Runnable command) {
16
if (command == null)
17
throw new NullPointerException();
18
/*
19
* Proceed in 3 steps:
20
*
21
* 1. If fewer than corePoolSize threads are running, try to
22
* start a new thread with the given command as its first
23
* task. The call to addWorker atomically checks runState and
24
* workerCount, and so prevents false alarms that would add
25
* threads when it shouldn't, by returning false.
26
*
27
* 2. If a task can be successfully queued, then we still need
28
* to double-check whether we should have added a thread
29
* (because existing ones died since last checking) or that
30
* the pool shut down since entry into this method. So we
31
* recheck state and if necessary roll back the enqueuing if
32
* stopped, or start a new thread if there are none.
33
*
34
* 3. If we cannot queue task, then we try to add a new
35
* thread. If it fails, we know we are shut down or saturated
36
* and so reject the task.
37
*/
38
int c = ctl.get();//计算工作线程的数量
39
if (workerCountOf(c) < corePoolSize) {//比较核心线程数和工作线程数的大小
40
if (addWorker(command, true))//工作线程数少，说明有线程空闲，不需要另外开启线程，直接往工作队列里添加线程即可
41
return;
42
c = ctl.get();
43
}
44
if (isRunning(c) && workQueue.offer(command)) {
45
int recheck = ctl.get();
46
if (! isRunning(recheck) && remove(command))//双重检查是否应该新建线程
47
reject(command);
48
else if (workerCountOf(recheck) == 0)
49
addWorker(null, false);//可以创建新线程
50
}
51
else if (!addWorker(command, false))
52
reject(command);//队列已满，拒绝创建新线程
53
}
那么是怎么创建新线程的呢？一起奥秘都在addWorker这个方法里：
1
/**
2
* Checks if a new worker can be added with respect to current
3
* pool state and the given bound (either core or maximum). If so,
4
* the worker count is adjusted accordingly, and, if possible, a
5
* new worker is created and started, running firstTask as its
6
* first task. This method returns false if the pool is stopped or
7
* eligible to shut down. It also returns false if the thread
8
* factory fails to create a thread when asked. If the thread
9
* creation fails, either due to the thread factory returning
10
* null, or due to an exception (typically OutOfMemoryError in
11
* Thread.start()), we roll back cleanly.
12
*
13
* @param firstTask the task the new thread should run first (or
14
* null if none). Workers are created with an initial first task
15
* (in method execute()) to bypass queuing when there are fewer
16
* than corePoolSize threads (in which case we always start one),
17
* or when the queue is full (in which case we must bypass queue).
18
* Initially idle threads are usually created via
19
* prestartCoreThread or to replace other dying workers.
20
*
21
* @param core if true use corePoolSize as bound, else
22
* maximumPoolSize. (A boolean indicator is used here rather than a
23
* value to ensure reads of fresh values after checking other pool
24
* state).
25
* @return true if successful
26
*/
27
private boolean addWorker(Runnable firstTask, boolean core) {
28
retry:
29
for (;;) {//死循环里判断线程池状态
30
int c = ctl.get();
31
int rs = runStateOf(c);
32
33
// Check if queue empty only if necessary.
34
if (rs >= SHUTDOWN &&
35
! (rs == SHUTDOWN &&
36
firstTask == null &&
37
! workQueue.isEmpty()))
38
return false;
39
//内层死循环，workCount数量加一
40
for (;;) {
41
int wc = workerCountOf(c);
42
if (wc >= CAPACITY ||
43
wc >= (core ? corePoolSize : maximumPoolSize))
44
return false;
45
if (compareAndIncrementWorkerCount(c))
46
break retry;
47
c = ctl.get(); // Re-read ctl
48
if (runStateOf(c) != rs)
49
continue retry;
50
// else CAS failed due to workerCount change; retry inner loop
51
}
52
}
53
54
boolean workerStarted = false;
55
boolean workerAdded = false;
56
Worker w = null;
57
try {
58
w = new Worker(firstTask);
59
final Thread t = w.thread;
60
if (t != null) {
61
final ReentrantLock mainLock = this.mainLock;
62
mainLock.lock();
63
try {
64
// Recheck while holding lock.
65
// Back out on ThreadFactory failure or if
66
// shut down before lock acquired.
67
int rs = runStateOf(ctl.get());
68
69
if (rs < SHUTDOWN ||
70
(rs == SHUTDOWN && firstTask == null)) {
71
if (t.isAlive()) // precheck that t is startable
72
throw new IllegalThreadStateException();
73
workers.add(w);
74
int s = workers.size();
75
if (s > largestPoolSize)
76
largestPoolSize = s;
77
workerAdded = true;
78
}
79
} finally {
80
mainLock.unlock();
81
}
82
//如果往HashSet中添加worker成功，启动线程
83
if (workerAdded) {
84
t.start();
85
workerStarted = true;
86
}
87
}
88
} finally {
89
if (! workerStarted)
90
addWorkerFailed(w);
91
}
92
return workerStarted;
93
}
94
参考资料：http://www.cnblogs.com/trust-freedom/p/6681948.html