多線程與高併發（八）：ThreadPoolExecutor源碼解析， SingleThreadPool，CachedPool，FixedThreadPool，ForkJoinPoll 等

線程池

今天我們來看看JDK給我們提供的默認的線程池的實現。

ThreadPoolExecutor：我們通常所說的線程池。多個線程共享同一個任務隊列。

• SingleThreadPool
• CachedPool
• FixedThreadPool
• ScheduledPool

ForkJoinPoll：先將任務分解，最後再彙總。每個線程有自己的任務隊列。

• WorkStealingPool

SingleThreadPool

SingleThreadPool 這個線程池裏面只有一個線程。這樣可以保證 我們扔進去的任務是被順序執行的。

問：爲什麼會有單線程的線程池？
單線程的線程池是有任務隊列的；線程池能幫你提供線程生命週期的管理。

package com.mashibing.juc.c_026_01_ThreadPool;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class T07_SingleThreadPool {
    public static void main(String[] args) {
        ExecutorService service = Executors.newSingleThreadExecutor();
        for (int i = 0; i < 5; i++) {
            final int j = i;
            service.execute(() -> {
                System.out.println(j + " " + Thread.currentThread().getName());
            });
        }
    }
}

CachedPool

CachedPool 核心線程數爲0，最大線程數是Integer.MAX_VALUE

SynchronousQueue是一個特殊的隊列：

SynchronousQueue 是一個阻塞隊列，其中每個插入操作必須等待另一個線程執行相應的刪除操作，反之亦然。同步隊列沒有任何內部容量，甚至容量都不是1。您無法查看同步隊列，因爲只有當您試圖刪除某個元素時，它纔會出現;你不能插入元素(使用任何方法)，除非另一個線程試圖刪除它;你不能迭代，因爲沒有東西可以迭代。隊列的頭是第一個排隊插入線程試圖添加到隊列中的元素;如果沒有這樣的排隊線程，那麼就沒有可刪除的元素，poll()將返回null。對於其他集合方法(例如包含)，SynchronousQueue充當空集合。此隊列不允許空元素。

CachedPool 這個線程池，當任務到來時，如果有線程空閒，我就用現有的線程；如果所有線程忙，就啓動一個新線程。

阿里不推薦使用這種線程池！

package com.mashibing.juc.c_026_01_ThreadPool;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

public class T08_CachedPool {
    public static void main(String[] args) throws InterruptedException {
        ExecutorService service = Executors.newCachedThreadPool();
        System.out.println(service);
        for (int i = 0; i < 2; i++) {
            service.execute(() -> {
                try {
                    TimeUnit.MILLISECONDS.sleep(500);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
                System.out.println(Thread.currentThread().getName());
            });
        }
        System.out.println(service);
        TimeUnit.SECONDS.sleep(80);
        System.out.println(service);
    }
}

運行結果

java.util.concurrent.ThreadPoolExecutor@6aaa5eb0[Running, pool size = 0, active threads = 0, queued tasks = 0, completed tasks = 0]
java.util.concurrent.ThreadPoolExecutor@6aaa5eb0[Running, pool size = 2, active threads = 2, queued tasks = 0, completed tasks = 0]
pool-1-thread-1
pool-1-thread-2
java.util.concurrent.ThreadPoolExecutor@6aaa5eb0[Running, pool size = 0, active threads = 0, queued tasks = 0, completed tasks = 2]

Process finished with exit code 0

FixedThreadPool

FixedThreadPool 是一個固定線程數的線程池。
它的核心線程數和最大線程數是固定的！

固定線程數的好處是：
適合做一些並行的計算，比如你要找1-200000之內所有的質數，你將這個大任務拆成4個小線程，共同去運行。
利用線程池進行並行的計算，肯定比串行計算要更快。

併發和並行的區別：
併發指任務提交，並行指任務執行；
並行是多個CPU同時處理，併發是多個任務同時過來；
並行是併發的子集

/**
 * 線程池的概念
 * nasa
 */
package com.mashibing.juc.c_026_01_ThreadPool;

import java.util.ArrayList;
import java.util.List;
import java.util.concurrent.Callable;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Future;

public class T09_FixedThreadPool {
    public static void main(String[] args) throws InterruptedException, ExecutionException {
        long start = System.currentTimeMillis();
        getPrime(1, 200000);
        long end = System.currentTimeMillis();
        System.out.println(end - start);

        final int cpuCoreNum = 4;

        ExecutorService service = Executors.newFixedThreadPool(cpuCoreNum);

        MyTask t1 = new MyTask(1, 80000); //1-5 5-10 10-15 15-20
        MyTask t2 = new MyTask(80001, 130000);
        MyTask t3 = new MyTask(130001, 170000);
        MyTask t4 = new MyTask(170001, 200000);

        Future<List<Integer>> f1 = service.submit(t1);
        Future<List<Integer>> f2 = service.submit(t2);
        Future<List<Integer>> f3 = service.submit(t3);
        Future<List<Integer>> f4 = service.submit(t4);

        start = System.currentTimeMillis();
        f1.get();
        f2.get();
        f3.get();
        f4.get();
        end = System.currentTimeMillis();
        System.out.println(end - start);
    }

    static class MyTask implements Callable<List<Integer>> {
        int startPos, endPos;

        MyTask(int s, int e) {
            this.startPos = s;
            this.endPos = e;
        }

        @Override
        public List<Integer> call() throws Exception {
            List<Integer> r = getPrime(startPos, endPos);
            return r;
        }

    }

    static boolean isPrime(int num) {
        for (int i = 2; i <= num / 2; i++) {
            if (num % i == 0) return false;
        }
        return true;
    }

    static List<Integer> getPrime(int start, int end) {
        List<Integer> results = new ArrayList<>();
        for (int i = start; i <= end; i++) {
            if (isPrime(i)) results.add(i);
        }
        return results;
    }
}

串行和並行計算耗時比較：

8860
3039

CachedThreadPool 和 FixedThreadPool 的選用

如果任務來的速度忽快忽慢，但是我要保證任務來的時候有人來做這個任務，那麼我們可以使用 CachedThreadPool，保證任務不會堆積。

如果任務來的比較平穩，我們大概估算出一個需要的線程數量，這個值完全能夠處理所有的任務，就可以使用FixedThreadPool

阿里是這兩種都不用，自己估算，進行精確定義。

ScheduledPool

專門用來執行定時任務的一個線程池。

package com.mashibing.juc.c_026_01_ThreadPool;

import java.util.Random;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;

public class T10_ScheduledPool {
    public static void main(String[] args) {
        ScheduledExecutorService service = Executors.newScheduledThreadPool(4);
        service.scheduleAtFixedRate(() -> {
            try {
                TimeUnit.MILLISECONDS.sleep(new Random().nextInt(1000));
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            System.out.println(Thread.currentThread().getName());
        }, 0, 500, TimeUnit.MILLISECONDS);
    }
}

自定義拒絕策略

package com.mashibing.juc.c_026_01_ThreadPool;

import java.util.concurrent.*;

public class T14_MyRejectedHandler {
    public static void main(String[] args) {
        ExecutorService service = new ThreadPoolExecutor(4, 4,
                0, TimeUnit.SECONDS, new ArrayBlockingQueue<>(6),
                Executors.defaultThreadFactory(),
                new MyHandler());
    }

    static class MyHandler implements RejectedExecutionHandler {

        @Override
        public void rejectedExecution(Runnable r, ThreadPoolExecutor executor) {
            //log("r rejected")
            //save r kafka mysql redis
            //try 3 times
            if (executor.getQueue().size() < 10000) {
                //try put again();
            }
        }
    }
}

ThreadPoolExecutor源碼解析

1、常用變量的解釋

// 1. `ctl`，可以看做一個int類型的數字，高3位表示線程池狀態，低29位表示worker數量
private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0));
// 2. `COUNT_BITS`，`Integer.SIZE`爲32，所以`COUNT_BITS`爲29
private static final int COUNT_BITS = Integer.SIZE - 3;
// 3. `CAPACITY`，線程池允許的最大線程數。1左移29位，然後減1，即爲 2^29 - 1
private static final int CAPACITY   = (1 << COUNT_BITS) - 1;

// runState is stored in the high-order bits
// 4. 線程池有5種狀態，按大小排序如下：RUNNING < SHUTDOWN < STOP < TIDYING < TERMINATED
private static final int RUNNING    = -1 << COUNT_BITS;
private static final int SHUTDOWN   =  0 << COUNT_BITS;
private static final int STOP       =  1 << COUNT_BITS;
private static final int TIDYING    =  2 << COUNT_BITS;
private static final int TERMINATED =  3 << COUNT_BITS;

// Packing and unpacking ctl
// 5. `runStateOf()`，獲取線程池狀態，通過按位與操作，低29位將全部變成0
private static int runStateOf(int c)     { return c & ~CAPACITY; }
// 6. `workerCountOf()`，獲取線程池worker數量，通過按位與操作，高3位將全部變成0
private static int workerCountOf(int c)  { return c & CAPACITY; }
// 7. `ctlOf()`，根據線程池狀態和線程池worker數量，生成ctl值
private static int ctlOf(int rs, int wc) { return rs | wc; }

/*
 * Bit field accessors that don't require unpacking ctl.
 * These depend on the bit layout and on workerCount being never negative.
 */
// 8. `runStateLessThan()`，線程池狀態小於xx
private static boolean runStateLessThan(int c, int s) {
    return c < s;
}
// 9. `runStateAtLeast()`，線程池狀態大於等於xx
private static boolean runStateAtLeast(int c, int s) {
    return c >= s;
}

2、構造方法

public ThreadPoolExecutor(int corePoolSize,
                          int maximumPoolSize,
                          long keepAliveTime,
                          TimeUnit unit,
                          BlockingQueue<Runnable> workQueue,
                          ThreadFactory threadFactory,
                          RejectedExecutionHandler handler) {
    // 基本類型參數校驗
    if (corePoolSize < 0 ||
        maximumPoolSize <= 0 ||
        maximumPoolSize < corePoolSize ||
        keepAliveTime < 0)
        throw new IllegalArgumentException();
    // 空指針校驗
    if (workQueue == null || threadFactory == null || handler == null)
        throw new NullPointerException();
    this.corePoolSize = corePoolSize;
    this.maximumPoolSize = maximumPoolSize;
    this.workQueue = workQueue;
    // 根據傳入參數`unit`和`keepAliveTime`，將存活時間轉換爲納秒存到變量`keepAliveTime `中
    this.keepAliveTime = unit.toNanos(keepAliveTime);
    this.threadFactory = threadFactory;
    this.handler = handler;
}

3、提交執行task的過程

public void execute(Runnable command) {
    if (command == null)
        throw new NullPointerException();
    /*
     * Proceed in 3 steps:
     *
     * 1. If fewer than corePoolSize threads are running, try to
     * start a new thread with the given command as its first
     * task.  The call to addWorker atomically checks runState and
     * workerCount, and so prevents false alarms that would add
     * threads when it shouldn't, by returning false.
     *
     * 2. If a task can be successfully queued, then we still need
     * to double-check whether we should have added a thread
     * (because existing ones died since last checking) or that
     * the pool shut down since entry into this method. So we
     * recheck state and if necessary roll back the enqueuing if
     * stopped, or start a new thread if there are none.
     *
     * 3. If we cannot queue task, then we try to add a new
     * thread.  If it fails, we know we are shut down or saturated
     * and so reject the task.
     */
    int c = ctl.get();
    // worker數量比核心線程數小，直接創建worker執行任務
    if (workerCountOf(c) < corePoolSize) {
        if (addWorker(command, true))
            return;
        c = ctl.get();
    }
    // worker數量超過核心線程數，任務直接進入隊列
    if (isRunning(c) && workQueue.offer(command)) {
        int recheck = ctl.get();
        // 線程池狀態不是RUNNING狀態，說明執行過shutdown命令，需要對新加入的任務執行reject()操作。
        // 這兒爲什麼需要recheck，是因爲任務入隊列前後，線程池的狀態可能會發生變化。
        if (! isRunning(recheck) && remove(command))
            reject(command);
        // 這兒爲什麼需要判斷0值，主要是在線程池構造方法中，核心線程數允許爲0
        else if (workerCountOf(recheck) == 0)
            addWorker(null, false);
    }
    // 如果線程池不是運行狀態，或者任務進入隊列失敗，則嘗試創建worker執行任務。
    // 這兒有3點需要注意：
    // 1. 線程池不是運行狀態時，addWorker內部會判斷線程池狀態
    // 2. addWorker第2個參數表示是否創建核心線程
    // 3. addWorker返回false，則說明任務執行失敗，需要執行reject操作
    else if (!addWorker(command, false))
        reject(command);
}

4、addworker源碼解析

private boolean addWorker(Runnable firstTask, boolean core) {
    retry:
    // 外層自旋
    for (;;) {
        int c = ctl.get();
        int rs = runStateOf(c);

        // 這個條件寫得比較難懂，我對其進行了調整，和下面的條件等價
        // (rs > SHUTDOWN) || 
        // (rs == SHUTDOWN && firstTask != null) || 
        // (rs == SHUTDOWN && workQueue.isEmpty())
        // 1. 線程池狀態大於SHUTDOWN時，直接返回false
        // 2. 線程池狀態等於SHUTDOWN，且firstTask不爲null，直接返回false
        // 3. 線程池狀態等於SHUTDOWN，且隊列爲空，直接返回false
        // Check if queue empty only if necessary.
        if (rs >= SHUTDOWN &&
            ! (rs == SHUTDOWN &&
               firstTask == null &&
               ! workQueue.isEmpty()))
            return false;

        // 內層自旋
        for (;;) {
            int wc = workerCountOf(c);
            // worker數量超過容量，直接返回false
            if (wc >= CAPACITY ||
                wc >= (core ? corePoolSize : maximumPoolSize))
                return false;
            // 使用CAS的方式增加worker數量。
            // 若增加成功，則直接跳出外層循環進入到第二部分
            if (compareAndIncrementWorkerCount(c))
                break retry;
            c = ctl.get();  // Re-read ctl
            // 線程池狀態發生變化，對外層循環進行自旋
            if (runStateOf(c) != rs)
                continue retry;
            // 其他情況，直接內層循環進行自旋即可
            // 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;
            // worker的添加必須是串行的，因此需要加鎖
            mainLock.lock();
            try {
                // Recheck while holding lock.
                // Back out on ThreadFactory failure or if
                // shut down before lock acquired.
                // 這兒需要重新檢查線程池狀態
                int rs = runStateOf(ctl.get());

                if (rs < SHUTDOWN ||
                    (rs == SHUTDOWN && firstTask == null)) {
                    // worker已經調用過了start()方法，則不再創建worker
                    if (t.isAlive()) // precheck that t is startable
                        throw new IllegalThreadStateException();
                    // worker創建並添加到workers成功
                    workers.add(w);
                    // 更新`largestPoolSize`變量
                    int s = workers.size();
                    if (s > largestPoolSize)
                        largestPoolSize = s;
                    workerAdded = true;
                }
            } finally {
                mainLock.unlock();
            }
            // 啓動worker線程
            if (workerAdded) {
                t.start();
                workerStarted = true;
            }
        }
    } finally {
        // worker線程啓動失敗，說明線程池狀態發生了變化（關閉操作被執行），需要進行shutdown相關操作
        if (! workerStarted)
            addWorkerFailed(w);
    }
    return workerStarted;
}

5、線程池worker任務單元

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;
        // 這兒是Worker的關鍵所在，使用了線程工廠創建了一個線程。傳入的參數爲當前worker
        this.thread = getThreadFactory().newThread(this);
    }

    /** Delegates main run loop to outer runWorker  */
    public void run() {
        runWorker(this);
    }

    // 省略代碼...
}

6、核心線程執行邏輯-runworker

final void runWorker(Worker w) {
    Thread wt = Thread.currentThread();
    Runnable task = w.firstTask;
    w.firstTask = null;
    // 調用unlock()是爲了讓外部可以中斷
    w.unlock(); // allow interrupts
    // 這個變量用於判斷是否進入過自旋（while循環）
    boolean completedAbruptly = true;
    try {
        // 這兒是自旋
        // 1. 如果firstTask不爲null，則執行firstTask；
        // 2. 如果firstTask爲null，則調用getTask()從隊列獲取任務。
        // 3. 阻塞隊列的特性就是：當隊列爲空時，當前線程會被阻塞等待
        while (task != null || (task = getTask()) != null) {
            // 這兒對worker進行加鎖，是爲了達到下面的目的
            // 1. 降低鎖範圍，提升性能
            // 2. 保證每個worker執行的任務是串行的
            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
            // 如果線程池正在停止，則對當前線程進行中斷操作
            if ((runStateAtLeast(ctl.get(), STOP) ||
                 (Thread.interrupted() &&
                  runStateAtLeast(ctl.get(), STOP))) &&
                !wt.isInterrupted())
                wt.interrupt();
            // 執行任務，且在執行前後通過`beforeExecute()`和`afterExecute()`來擴展其功能。
            // 這兩個方法在當前類裏面爲空實現。
            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 {
                // 幫助gc
                task = null;
                // 已完成任務數加一 
                w.completedTasks++;
                w.unlock();
            }
        }
        completedAbruptly = false;
    } finally {
        // 自旋操作被退出，說明線程池正在結束
        processWorkerExit(w, completedAbruptly);
    }
}

WorkStealingPool

原來的線程池：有一個線程的集合，去一個任務隊列裏面取任務，取出任務之後執行。下圖：

WorkStealingPool 偷任務的線程池：每一個線程都有自己獨立的任務隊列，如果某一個線程執行完自己的任務之後，要去別的線程那裏偷任務，分擔別的線程的任務。

WorkStealingPool 本質上還是一個 ForkJoinPool

/**
 *
 */
package com.mashibing.juc.c_026_01_ThreadPool;

import java.io.IOException;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

public class T11_WorkStealingPool {
    public static void main(String[] args) throws IOException {
        ExecutorService service = Executors.newWorkStealingPool();
        System.out.println(Runtime.getRuntime().availableProcessors());

        service.execute(new R(1000));
        service.execute(new R(2000));
        service.execute(new R(2000));
        service.execute(new R(2000)); //daemon
        service.execute(new R(2000));

        //由於產生的是精靈線程（守護線程、後臺線程），主線程不阻塞的話，看不到輸出
        System.in.read();
    }

    static class R implements Runnable {
        int time;
        R(int t) {
            this.time = t;
        }

        @Override
        public void run() {
            try {
                TimeUnit.MILLISECONDS.sleep(time);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            System.out.println(time + " " + Thread.currentThread().getName());
        }
    }
}

ForkJoinPool

把大任務切分成一個個小任務去運行，執行完之後進行彙總。
可以有返回值或無返回值。

package com.mashibing.juc.c_026_01_ThreadPool;

import java.io.IOException;
import java.util.Arrays;
import java.util.Random;
import java.util.concurrent.ForkJoinPool;
import java.util.concurrent.RecursiveAction;
import java.util.concurrent.RecursiveTask;

public class T12_ForkJoinPool {
    static int[] nums = new int[1000000];
    static final int MAX_NUM = 50000;
    static Random r = new Random();

    static {
        for (int i = 0; i < nums.length; i++) {
            nums[i] = r.nextInt(100);
        }
        System.out.println("stream api---" + Arrays.stream(nums).sum()); //stream api，單線程的計算方式
    }


    static class AddTask extends RecursiveAction {
        int start, end;

        AddTask(int s, int e) {
            start = s;
            end = e;
        }

        @Override
        protected void compute() {
            if (end - start <= MAX_NUM) {
                long sum = 0L;
                for (int i = start; i < end; i++) sum += nums[i];
                System.out.println("from:" + start + " to:" + end + " = " + sum);
            } else {

                int middle = start + (end - start) / 2;

                AddTask subTask1 = new AddTask(start, middle);
                AddTask subTask2 = new AddTask(middle, end);
                subTask1.fork();
                subTask2.fork();
            }
        }
    }

	// 帶有返回值的任務拆分
    static class AddTaskReturn extends RecursiveTask<Long> {
        private static final long serialVersionUID = 1L;
        int start, end;

        AddTaskReturn(int s, int e) {
            start = s;
            end = e;
        }

        @Override
        protected Long compute() {
            if (end - start <= MAX_NUM) {
                long sum = 0L;
                for (int i = start; i < end; i++) sum += nums[i];
                return sum;
            }
            int middle = start + (end - start) / 2;
            AddTaskReturn subTask1 = new AddTaskReturn(start, middle);
            AddTaskReturn subTask2 = new AddTaskReturn(middle, end);
            subTask1.fork();
            subTask2.fork();
            return subTask1.join() + subTask2.join();
        }

    }

    public static void main(String[] args) throws IOException {
		/*ForkJoinPool fjp = new ForkJoinPool();
		AddTask task = new AddTask(0, nums.length);
		fjp.execute(task);*/

        T12_ForkJoinPool temp = new T12_ForkJoinPool();
        ForkJoinPool fjp = new ForkJoinPool();
        AddTaskReturn task = new AddTaskReturn(0, nums.length);
        fjp.execute(task);
        long result = task.join();
        System.out.println(result);
        //System.in.read();
    }
}

流式API的ForkJoinPool的算法實現

流式API的底層也是使用 ForkJoinPool 來實現的。nums.parallelStream().forEach這種並行流處理起來效率會更高一些。

package com.mashibing.juc.c_026_01_ThreadPool;

import java.util.ArrayList;
import java.util.List;
import java.util.Random;

public class T13_ParallelStreamAPI {
    public static void main(String[] args) {
        List<Integer> nums = new ArrayList<>();
        Random r = new Random();
        for (int i = 0; i < 10000; i++) nums.add(1000000 + r.nextInt(1000000));

        //System.out.println(nums);

        long start = System.currentTimeMillis();
        nums.forEach(v -> isPrime(v));
        long end = System.currentTimeMillis();
        System.out.println(end - start);

        //使用parallel stream api

        start = System.currentTimeMillis();
        nums.parallelStream().forEach(T13_ParallelStreamAPI::isPrime);
        end = System.currentTimeMillis();

        System.out.println(end - start);
    }

    static boolean isPrime(int num) {
        for (int i = 2; i <= num / 2; i++) {
            if (num % i == 0) return false;
        }
        return true;
    }
}

OK，到此爲止，我們的線程池講完了！