服务器之家

服务器之家 > 正文

java 中ThreadPoolExecutor原理分析

时间:2020-09-03 13:42     来源/作者:刘正阳

java 中ThreadPoolExecutor原理分析

线程池简介

Java线程池是开发中常用的工具,当我们有异步、并行的任务要处理时,经常会用到线程池,或者在实现一个服务器时,也需要使用线程池来接收连接处理请求。

线程池使用

JDK中提供的线程池实现位于java.util.concurrent.ThreadPoolExecutor。在使用时,通常使用ExecutorService接口,它提供了submit,invokeAll,shutdown等通用的方法。

在线程池配置方面,Executors类中提供了一些静态方法能够提供一些常用场景的线程池,如newFixedThreadPool,newCachedThreadPool,newSingleThreadExecutor等,这些方法最终都是调用到了ThreadPoolExecutor的构造函数。

ThreadPoolExecutor的包含所有参数的构造函数是

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
/**
   * @param corePoolSize the number of threads to keep in the pool, even
   *    if they are idle, unless {@code allowCoreThreadTimeOut} is set
   * @param maximumPoolSize the maximum number of threads to allow in the
   *    pool
   * @param keepAliveTime when the number of threads is greater than
   *    the core, this is the maximum time that excess idle threads
   *    will wait for new tasks before terminating.
   * @param unit the time unit for the {@code keepAliveTime} argument
   * @param workQueue the queue to use for holding tasks before they are
   *    executed. This queue will hold only the {@code Runnable}
   *    tasks submitted by the {@code execute} method.
   * @param threadFactory the factory to use when the executor
   *    creates a new thread
   * @param handler the handler to use when execution is blocked
   *    because the thread bounds and queue capacities are reached
  public ThreadPoolExecutor(int corePoolSize,
               int maximumPoolSize,
               long keepAliveTime,
               TimeUnit unit,
               BlockingQueue<Runnable> workQueue,
               ThreadFactory threadFactory,
               RejectedExecutionHandler handler) {
    if (corePoolSize < 0 ||
      maximumPoolSize <= 0 ||
      maximumPoolSize < corePoolSize ||
      keepAliveTime < 0)
      throw new IllegalArgumentException();
    if (workQueue == null || threadFactory == null || handler == null)
      throw new NullPointerException();
    this.corePoolSize = corePoolSize;
    this.maximumPoolSize = maximumPoolSize;
    this.workQueue = workQueue;
    this.keepAliveTime = unit.toNanos(keepAliveTime);
    this.threadFactory = threadFactory;
    this.handler = handler;
  }
  • corePoolSize设置线程池的核心线程数,当添加新任务时,如果线程池中的线程数小于corePoolSize,则不管当前是否有线程闲置,都会创建一个新的线程来执行任务。
  • maximunPoolSize是线程池中允许的最大的线程数
  • workQueue用于存放排队的任务
  • keepAliveTime是大于corePoolSize的线程闲置的超时时间
  • handler用于在任务逸出、线程池关闭时的任务处理 ,线程池的线程增长策略为,当前线程数小于corePoolSize时,新增线程,当线程数=corePoolSize且corePoolSize时,只有在workQueue不能存放新的任务时创建新线程,超出的线程在闲置keepAliveTime后销毁。

实现(基于JDK1.8)

ThreadPoolExecutor中保存的状态有

当前线程池状态, 包括RUNNING,SHUTDOWN,STOP,TIDYING,TERMINATED。

当前有效的运行线程的数量。

将这两个状态放到一个int变量中,前三位作为线程池状态,后29位作为线程数量。

例如0b11100000000000000000000000000001, 表示RUNNING, 一个线程。

通过HashSet来存储工作者集合,访问该HashSet前必须先获取保护状态的mainLock:ReentrantLock

submit、execute

execute的执行方式为,首先检查当前worker数量,如果小于corePoolSize,则尝试add一个core Worker。线程池在维护线程数量以及状态检查上做了大量检测。

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
public void execute(Runnable command) {
    int c = ctl.get();
    // 如果当期数量小于corePoolSize
    if (workerCountOf(c) < corePoolSize) {
      // 尝试增加worker
      if (addWorker(command, true))
        return;
      c = ctl.get();
    }
    // 如果线程池正在运行并且成功添加到工作队列中
    if (isRunning(c) && workQueue.offer(command)) {
      // 再次检查状态,如果已经关闭则执行拒绝处理
      int recheck = ctl.get();
      if (! isRunning(recheck) && remove(command))
        reject(command);
      // 如果工作线程都down了
      else if (workerCountOf(recheck) == 0)
        addWorker(null, false);
    }
    else if (!addWorker(command, false))
      reject(command);
  }

addWorker方法实现

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
private boolean addWorker(Runnable firstTask, boolean core) {
    retry:
    for (;;) {
      int c = ctl.get();
      int rs = runStateOf(c);
      // Check if queue empty only if necessary.
      if (rs >= SHUTDOWN &&
        ! (rs == SHUTDOWN &&
          firstTask == null &&
          ! workQueue.isEmpty()))
        return false;
      for (;;) {
        int wc = workerCountOf(c);
        if (wc >= CAPACITY ||
          wc >= (core ? corePoolSize : maximumPoolSize))
          return false;
        if (compareAndIncrementWorkerCount(c))
          break retry;
        c = ctl.get(); // Re-read ctl
        if (runStateOf(c) != rs)
          continue retry;
        // else CAS failed due to workerCount change; retry inner loop
      }
    }
    boolean workerStarted = false;
    boolean workerAdded = false;
    Worker w = null;
    try {
      w = new Worker(firstTask);
      final Thread t = w.thread;
      if (t != null) {
        final ReentrantLock mainLock = this.mainLock;
        mainLock.lock();
        try {
          // Recheck while holding lock.
          // Back out on ThreadFactory failure or if
          // shut down before lock acquired.
          int rs = runStateOf(ctl.get());
          if (rs < SHUTDOWN ||
            (rs == SHUTDOWN && firstTask == null)) {
            if (t.isAlive()) // precheck that t is startable
              throw new IllegalThreadStateException();
            workers.add(w);
            int s = workers.size();
            if (s > largestPoolSize)
              largestPoolSize = s;
            workerAdded = true;
          }
        } finally {
          mainLock.unlock();
        }
        if (workerAdded) {
          // 如果添加成功,则启动该线程,执行Worker的run方法,Worker的run方法执行外部的runWorker(Worker)
          t.start();
          workerStarted = true;
        }
      }
    } finally {
      if (! workerStarted)
        addWorkerFailed(w);
    }
    return workerStarted;
  }

Worker类继承了AbstractQueuedSynchronizer获得了同步等待这样的功能。

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
private final class Worker
    extends AbstractQueuedSynchronizer
    implements Runnable
  {
    /**
     * This class will never be serialized, but we provide a
     * serialVersionUID to suppress a javac warning.
     */
    private static final long serialVersionUID = 6138294804551838833L;
    /** Thread this worker is running in. Null if factory fails. */
    final Thread thread;
    /** Initial task to run. Possibly null. */
    Runnable firstTask;
    /** Per-thread task counter */
    volatile long completedTasks;
    /**
     * Creates with given first task and thread from ThreadFactory.
     * @param firstTask the first task (null if none)
     */
    Worker(Runnable firstTask) {
      setState(-1); // inhibit interrupts until runWorker
      this.firstTask = firstTask;
      this.thread = getThreadFactory().newThread(this);
    }
    /** Delegates main run loop to outer runWorker */
    public void run() {
      runWorker(this);
    }
    // Lock methods
    //
    // The value 0 represents the unlocked state.
    // The value 1 represents the locked state.
    protected boolean isHeldExclusively() {
      return getState() != 0;
    }
    protected boolean tryAcquire(int unused) {
      if (compareAndSetState(0, 1)) {
        setExclusiveOwnerThread(Thread.currentThread());
        return true;
      }
      return false;
    }
    protected boolean tryRelease(int unused) {
      setExclusiveOwnerThread(null);
      setState(0);
      return true;
    }
    public void lock()    { acquire(1); }
    public boolean tryLock() { return tryAcquire(1); }
    public void unlock()   { release(1); }
    public boolean isLocked() { return isHeldExclusively(); }
    void interruptIfStarted() {
      Thread t;
      if (getState() >= 0 && (t = thread) != null && !t.isInterrupted()) {
        try {
          t.interrupt();
        } catch (SecurityException ignore) {
        }
      }
    }

runWorker(Worker)是Worker的轮询执行逻辑,不断地从工作队列中获取任务并执行它们。Worker每次执行任务前需要进行lock,防止在执行任务时被interrupt。

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
final void runWorker(Worker w) {
    Thread wt = Thread.currentThread();
    Runnable task = w.firstTask;
    w.firstTask = null;
    w.unlock(); // allow interrupts
    boolean completedAbruptly = true;
    try {
      while (task != null || (task = getTask()) != null) {
        w.lock();
        // If pool is stopping, ensure thread is interrupted;
        // if not, ensure thread is not interrupted. This
        // requires a recheck in second case to deal with
        // shutdownNow race while clearing interrupt
        if ((runStateAtLeast(ctl.get(), STOP) ||
           (Thread.interrupted() &&
           runStateAtLeast(ctl.get(), STOP))) &&
          !wt.isInterrupted())
          wt.interrupt();
        try {
          beforeExecute(wt, task);
          Throwable thrown = null;
          try {
            task.run();
          } catch (RuntimeException x) {
            thrown = x; throw x;
          } catch (Error x) {
            thrown = x; throw x;
          } catch (Throwable x) {
            thrown = x; throw new Error(x);
          } finally {
            afterExecute(task, thrown);
          }
        } finally {
          task = null;
          w.completedTasks++;
          w.unlock();
        }
      }
      completedAbruptly = false;
    } finally {
      processWorkerExit(w, completedAbruptly);
    }
  }

ThreadPoolExecutor的submit方法中将Callable包装成FutureTask后交给execute方法。

FutureTask

FutureTask继承于Runnable和Future,FutureTask定义的几个状态为
NEW, 尚未执行
COMPLETING, 正在执行
NORMAL, 正常执行完成得到结果
EXCEPTIONAL, 执行抛出异常
CANCELLED, 执行被取消
INTERRUPTING,执行正在被中断
INTERRUPTED, 已经中断。

其中关键的get方法

?
1
2
3
4
5
6
public V get() throws InterruptedException, ExecutionException {
    int s = state;
    if (s <= COMPLETING)
      s = awaitDone(false, 0L);
    return report(s);
  }

先获取当前状态,如果还未执行完成并且正常,则进入等待结果流程。在awaitDone不断循环获取当前状态,如果没有结果,则将自己通过CAS的方式添加到等待链表的头部,如果设置了超时,则LockSupport.parkNanos到指定的时间。

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
static final class WaitNode {
    volatile Thread thread;
    volatile WaitNode next;
    WaitNode() { thread = Thread.currentThread(); }
  }
private int awaitDone(boolean timed, long nanos)
    throws InterruptedException {
    final long deadline = timed ? System.nanoTime() + nanos : 0L;
    WaitNode q = null;
    boolean queued = false;
    for (;;) {
      if (Thread.interrupted()) {
        removeWaiter(q);
        throw new InterruptedException();
      }
      int s = state;
      if (s > COMPLETING) {
        if (q != null)
          q.thread = null;
        return s;
      }
      else if (s == COMPLETING) // cannot time out yet
        Thread.yield();
      else if (q == null)
        q = new WaitNode();
      else if (!queued)
        queued = UNSAFE.compareAndSwapObject(this, waitersOffset,
                           q.next = waiters, q);
      else if (timed) {
        nanos = deadline - System.nanoTime();
        if (nanos <= 0L) {
          removeWaiter(q);
          return state;
        }
        LockSupport.parkNanos(this, nanos);
      }
      else
        LockSupport.park(this);
    }
  }

FutureTask的run方法是执行任务并设置结果的位置,首先判断当前状态是否为NEW并且将当前线程设置为执行线程,然后调用Callable的call获取结果后设置结果修改FutureTask状态。

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
public void run() {
    if (state != NEW ||
      !UNSAFE.compareAndSwapObject(this, runnerOffset,
                     null, Thread.currentThread()))
      return;
    try {
      Callable<V> c = callable;
      if (c != null && state == NEW) {
        V result;
        boolean ran;
        try {
          result = c.call();
          ran = true;
        } catch (Throwable ex) {
          result = null;
          ran = false;
          setException(ex);
        }
        if (ran)
          set(result);
      }
    } finally {
      // runner must be non-null until state is settled to
      // prevent concurrent calls to run()
      runner = null;
      // state must be re-read after nulling runner to prevent
      // leaked interrupts
      int s = state;
      if (s >= INTERRUPTING)
        handlePossibleCancellationInterrupt(s);
    }
  }

感谢阅读,希望能帮助到大家,谢谢大家对本站的支持!

原文链接:https://liuzhengyang.github.io/2017/03/27/threadpoolexecutor/

相关文章

热门资讯

2020微信伤感网名听哭了 让对方看到心疼的伤感网名大全
2020微信伤感网名听哭了 让对方看到心疼的伤感网名大全 2019-12-26
Intellij idea2020永久破解,亲测可用!!!
Intellij idea2020永久破解,亲测可用!!! 2020-07-29
歪歪漫画vip账号共享2020_yy漫画免费账号密码共享
歪歪漫画vip账号共享2020_yy漫画免费账号密码共享 2020-04-07
最新idea2020注册码永久激活(激活到2100年)
最新idea2020注册码永久激活(激活到2100年) 2020-07-29
iPhone12什么时候上市 iPhone12手机真实图片 苹果iphone12多少钱
iPhone12什么时候上市 iPhone12手机真实图片 苹果iphone12多少钱 2020-06-03
返回顶部