多线程高并发编程(8) -- Fork/Join源码分析

一.概念

Fork/Join就是将一个大任务分解(fork)成许多个独立的小任务，然后多线程并行去处理这些小任务，每个小任务处理完得到结果再进行合并(join)得到最终的结果。

流程：任务继承RecursiveTask，重写compute方法，使用ForkJoinPool的submit提交任务，任务在某个线程中运行，工作任务中的compute方法的代码开始对任务进行分析，如果符合条件就进行任务拆分，拆分成多个子任务，每个子任务进行数据的计算或操作，得到结果返回给上一层任务开启线程进行合并，最终通过get获取整体处理结果。【只能将任务1个切分为两个，不能切分为3个或其他数量】

ForkJoinTask：代表fork/join里面的任务类型，一般用它的两个子类RecursiveTask（任务有返回值）和RecursiveAction（任务没有返回值），任务的处理逻辑包括任务的切分都是在重写compute方法里面进行处理。只有ForkJoinTask任务可以被拆分运行和合并运行。【可查看上篇Future源码分析的类图结构】【ForkJoinTask使用了模板模式进行设计，将ForkJoinTask的执行相关代码进行隐藏，通过提供抽象类（即子类RecursiveTask、RecursiveAction）暴露用户的实际业务处理。】 RecursiveTask：在进行exec之后会使用一个result的变量进行接受返回的结果；

public abstract class RecursiveTask<V> extends ForkJoinTask<V> {V result;protected abstract V compute();public final V getRawResult() {return result;}protected final void setRawResult(V value) {result = value;}protected final boolean exec() {result = compute();return true;}}

RecursiveAction：在进行exec之后没有返回结果；

public abstract class RecursiveAction extends ForkJoinTask<Void> {protected abstract void compute();public final Void getRawResult() { return null; }protected final void setRawResult(Void mustBeNull) { }protected final boolean exec() {compute();return true;}}

ForkJoinPool：fork/join框架的管理者，最原始的任务都要交给它来处理。它负责控制整个fork/join有多少个工作线程，工作线程的创建、机会都是由它来控制。它还负责workQueue队列的创建和分配，每当创建一个工作线程，它负责分配对应的workQueue，然后它把接到的活都交给工作线程去处理。是整个fork/join的容器。 ForkJoinPool.WorkQueue：双端队列，负责存储接收的任务；ForkJoinWorkerThread：fork/join里面真正干活的”工人“，它继承了Thread，所以本质是一个线程。它有一个ForkJoinPool.WorkQueue的队列存放着它要干的活，接活之前它要向ForkJoinPool注册（registerWorker），拿到相应的workQueue，然后就从workQueue里面拿任务出来处理。它是依附于ForkJoinPool而存活，如果ForkJoinPool销毁了，它也会跟着结束。【每一个ForkJoinWorkerThread线程都具有一个独立的任务等待队列workQueue。】 当使用ForkJoinPool进行submit任务提交时，创建1个workQueue将任务放进去，然后进行fork任务切分，如果切分后的任务放的进去之前的workQueue就放进去，不行就随机选取workQueue放进去，如果还放不了就创建一个新的workQueue放进去；

public class ForkJoinWorkerThread extends Thread {final ForkJoinPool pool;final ForkJoinPool.WorkQueue workQueue;protected ForkJoinWorkerThread(ForkJoinPool pool) {super("aForkJoinWorkerThread");this.pool = pool;this.workQueue = pool.registerWorker(this);//向ForkJoinPool执行池注册当前工作线程，ForkJoinPool为其分配一个工作队列}}

二.用法

以前1+2+3+...+100这样的处理可以用for循环处理，现在使用fork/join来处理：从下面结果可以看到，大任务被不断的拆分成小任务，然后添加到工作线程的队列中，每个小任务都会被工作线程从队列中取出进行运行，然后每个小任务的结果的合并也由工作线程执行，然后不断的汇总成最终结果。【task通过ForkJoinPool来执行，分割的子任务添加到当前工作线程的队列中，进入队列的头部，当一个工作线程中没有任务时，会从其他工作线程的队列尾部获取一个任务。（工作窃取：当前工作线程对应的队列中没有任务了，从其他工作线程对应的队列中取出任务进行操作，然后将操作结果返还给对应队列的线程。）】

public class MyFrokJoinTask extends RecursiveTask<Integer> {private int begin;private int end;public MyFrokJoinTask(int begin, int end) {this.begin = begin;this.end = end;}public static void main(String[] args) throws Exception {ForkJoinPool pool = new ForkJoinPool();ForkJoinTask<Integer> result = pool.submit(new MyFrokJoinTask(1, 100));//提交任务System.out.println("计算的值："+result.get());//得到最终的结果}@Overrideprotected Integer compute() {int sum = 0;if (end - begin <= 2) {for (int i = begin; i <= end; i++) {sum += i;System.out.println("i:"+i);}} else {MyFrokJoinTask d1 = new MyFrokJoinTask(begin, (begin + end) / 2);MyFrokJoinTask d2 = new MyFrokJoinTask((begin + end) / 2+1, end);d1.fork();//任务拆分d2.fork();//任务拆分Integer a = d1.join();//每个任务的结果Integer b = d2.join();//每个任务的结果sum = a + b;//汇总任务结果System.out.println("sum:" + sum + ",a:" + a + ",b:" + b);}System.out.println("name:"+Thread.currentThread().getName());return sum;}}//=========结果============i:1i:2name:ForkJoinPool-1-worker-1i:3i:4name:ForkJoinPool-1-worker-1sum:10,a:3,b:7name:ForkJoinPool-1-worker-1i:5i:6i:7name:ForkJoinPool-1-worker-1sum:28,a:10,b:18name:ForkJoinPool-1-worker-1..............................sum:91,a:28,b:63sum:99,a:45,b:54name:ForkJoinPool-1-worker-3name:ForkJoinPool-1-worker-1i:23i:24i:25name:ForkJoinPool-1-worker-2sum:135,a:63,b:72name:ForkJoinPool-1-worker-2sum:234,a:99,b:135name:ForkJoinPool-1-worker-3sum:325,a:91,b:234name:ForkJoinPool-1-worker-1sum:1275,a:325,b:950name:ForkJoinPool-1-worker-1sum:5050,a:1275,b:3775name:ForkJoinPool-1-worker-1计算的值：5050

三.分析

ForkJoinPool

ForkJoinPool forkJoinPool = new ForkJoinPool();//Runtime.getRuntime().availableProcessors()当前操作系统可以使用的CPU内核数量public ForkJoinPool() {this(Math.min(MAX_CAP, Runtime.getRuntime().availableProcessors()),defaultForkJoinWorkerThreadFactory, null, false);}//this调用到下面这段代码public ForkJoinPool(int parallelism,ForkJoinWorkerThreadFactory factory,UncaughtExceptionHandler handler,boolean asyncMode) {this(checkParallelism(parallelism), //并行度checkFactory(factory), //工作线程创建工厂handler, //异常处理handlerasyncMode ? FIFO_QUEUE : LIFO_QUEUE, //任务队列出队模式 异步：先进先出，同步：后进先出"ForkJoinPool-" + nextPoolId() + "-worker-");checkPermission();}//上面的this最终调用到下面这段代码private ForkJoinPool(int parallelism,ForkJoinWorkerThreadFactory factory,UncaughtExceptionHandler handler,int mode,String workerNamePrefix) {this.workerNamePrefix = workerNamePrefix;this.factory = factory;this.ueh = handler;this.config = (parallelism & SMASK) | mode;long np = (long)(-parallelism); // offset ctl countsthis.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);}

parallelism：可并行数量，fork/join框架将依据这个并行数量的设定，决定框架内并行执行的线程数量。并行的每一个任务都会有一个线程进行处理；factory：当fork/join创建一个新的线程时，同样会用到线程创建工厂。它实现了ForkJoinWorkerThreadFactory接口，使用默认的的接口实现类DefaultForkJoinWorkerThreadFactory来实现newThread方法创建一个新的工作线程；

public static interface ForkJoinWorkerThreadFactory {/*** Returns a new worker thread operating in the given pool.*/public ForkJoinWorkerThread newThread(ForkJoinPool pool);}static final class DefaultForkJoinWorkerThreadFactoryimplements ForkJoinWorkerThreadFactory {public final ForkJoinWorkerThread newThread(ForkJoinPool pool) {return new ForkJoinWorkerThread(pool);}}

handler：异常捕获处理器。当执行的任务出现异常，并从任务中被抛出时，就会被handler捕获；asyncMode：fork/join为每一个独立的工作线程准备了对应的待执行任务队列，这个任务队列是使用数组进行组合的双向队列。即可以使用先进先出的工作模式，也可以使用后进先出的工作模式；

Fork()和Join()

fork/join框架中提供的fork()和join()是最重要的两个方法，它们和parallelism(”可并行任务数量“)配合工作，可以导致拆分的子任务T1.1、T1.2甚至TX在fork/join中不同的运行效果(上面1+2....+100的每次运行的子任务都是不同的)。即TX子任务或等待其他已存在的线程运行关联的子任务(sum操作)，或在运行TX的线程中”递归“执行其他任务(将1-50进行拆分后的子任务递归运行），或启动一个新的线程执行子任务(运行1-50另一边拆分的任务，即50-100的子任务）。

fork()用于将新创建的子任务放入当前线程的workQueue队列中，fork/join框架将根据当前正在并发执行ForkJoinTask任务的ForkJoinWorkerThread线程状态，决定是让这个任务在队列中等待，还是创建一个新的ForkJoinWorkedThread线程运行它，又或者是唤起其他正在等待任务的ForkJoinWorkerThread线程运行它。

join()用于让当前线程阻塞，直到对应的子任务完成运行并返回执行结果。或者，如果这个子任务存在于当前线程的任务等待队列workQueue中，则取出这个子任务进行”递归“执行，其目的是尽快得到当前子任务的运行结果，然后继续执行。

提交任务：

sumbit的第一次提交：ForkJoinPool.submit(ForkJoinTask<T> task) -> externalPush(task)->externalSubmit(task)

submit：

public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {if (task == null)throw new NullPointerException();externalPush(task);return task;}public <T> ForkJoinTask<T> submit(Callable<T> task) {ForkJoinTask<T> job = new ForkJoinTask.AdaptedCallable<T>(task);externalPush(job);return job;}public <T> ForkJoinTask<T> submit(Runnable task, T result) {ForkJoinTask<T> job = new ForkJoinTask.AdaptedRunnable<T>(task, result);externalPush(job);return job;}public ForkJoinTask<?> submit(Runnable task) {if (task == null)throw new NullPointerException();ForkJoinTask<?> job;if (task instanceof ForkJoinTask<?>) // avoid re-wrapjob = (ForkJoinTask<?>) task;elsejob = new ForkJoinTask.AdaptedRunnableAction(task);externalPush(job);return job;}

externalPush：将任务添加到随机选取的队列中或新创建的队列中；

final void externalPush(ForkJoinTask<?> task) {WorkQueue[] ws; WorkQueue q; int m;int r = ThreadLocalRandom.getProbe();//当前线程的一个随机数int rs = runState;//当前容器的状态//如果随机选取的队列还有空位置可以存放、队列加锁锁定成功，任务就放入队列中if ((ws = workQueues) != null && (m = (ws.length - 1)) >= 0 &&(q = ws[m & r & SQMASK]) != null && r != 0 && rs > 0 &&pareAndSwapInt(q, QLOCK, 0, 1)) {ForkJoinTask<?>[] a; int am, n, s;if ((a = q.array) != null &&(am = a.length - 1) > (n = (s = q.top) - q.base)) {int j = ((am & s) << ASHIFT) + ABASE;U.putOrderedObject(a, j, task);//任务加入队列中U.putOrderedInt(q, QTOP, s + 1);//挪动下次任务存放的槽的位置U.putIntVolatile(q, QLOCK, 0);//队列解锁if (n <= 1)//当前数组元素少时，进行唤醒当前线程；或者当没有活动线程或线程数较少时，添加新的线程signalWork(ws, q);return;}pareAndSwapInt(q, QLOCK, 1, 0);//队列解锁}externalSubmit(task);//升级版的externalPush}volatile int runState;// lockable status锁定状态// runState: SHUTDOWN为负数，其他的为2的次幂private static final int RSLOCK= 1;private static final int RSIGNAL = 1 << 1;//唤醒private static final int STARTED = 1 << 2;//启动private static final int STOP = 1 << 29;//停止private static final int TERMINATED = 1 << 30;//结束private static final int SHUTDOWN = 1 << 31;//关闭

externalSubmit：队列添加任务失败，进行升级版操作，即创建队列数组和创建队列后，将任务放入新创建的队列中；

private void externalSubmit(ForkJoinTask<?> task) {int r;// initialize caller's probeif ((r = ThreadLocalRandom.getProbe()) == 0) {ThreadLocalRandom.localInit();r = ThreadLocalRandom.getProbe();}for (;;) {//自旋WorkQueue[] ws; WorkQueue q; int rs, m, k;boolean move = false;/***ForkJoinPool执行器停止工作了，抛出异常*ForkJoinPool extends AbstractExecutorService*abstract class AbstractExecutorService implements ExecutorService*interface ExecutorService extends Executor*interface Executor执行提交的对象Runnable任务*/if ((rs = runState) < 0) {tryTerminate(false, false); // help terminatethrow new RejectedExecutionException();}//第一次遍历，队列数组未创建，进行创建else if ((rs & STARTED) == 0 ||// initialize初始化((ws = workQueues) == null || (m = ws.length - 1) < 0)) {int ns = 0;rs = lockRunState();try {if ((rs & STARTED) == 0) {pareAndSwapObject(this, STEALCOUNTER, null,new AtomicLong());// create workQueues array with size a power of twoint p = config & SMASK; // ensure at least 2 slots，config是CPU核数int n = (p > 1) ? p - 1 : 1;n |= n >>> 1; n |= n >>> 2; n |= n >>> 4;n |= n >>> 8; n |= n >>> 16; n = (n + 1) << 1;workQueues = new WorkQueue[n];//创建ns = STARTED;}} finally {unlockRunState(rs, (rs & ~RSLOCK) | ns);}}//第三次遍历，把任务放入队列中else if ((q = ws[k = r & m & SQMASK]) != null) {if (q.qlock == 0 && pareAndSwapInt(q, QLOCK, 0, 1)) {ForkJoinTask<?>[] a = q.array;int s = q.top;boolean submitted = false; // initial submission or resizingtry { // locked version of pushif ((a != null && a.length > s + 1 - q.base) ||(a = q.growArray()) != null) {int j = (((a.length - 1) & s) << ASHIFT) + ABASE;U.putOrderedObject(a, j, task);U.putOrderedInt(q, QTOP, s + 1);submitted = true;}} finally {pareAndSwapInt(q, QLOCK, 1, 0);}if (submitted) {signalWork(ws, q);return;}}move = true; // move on failure}//第二次遍历，队列数组为空，创建队列else if (((rs = runState) & RSLOCK) == 0) { // create new queueq = new WorkQueue(this, null);q.hint = r;q.config = k | SHARED_QUEUE;q.scanState = INACTIVE;rs = lockRunState(); // publish indexif (rs > 0 && (ws = workQueues) != null &&k < ws.length && ws[k] == null)ws[k] = q; // else terminatedunlockRunState(rs, rs & ~RSLOCK);}elsemove = true; // move if busyif (move)r = ThreadLocalRandom.advanceProbe(r);}}

fork任务切分的提交：ForkJoinTask.fork() -> ForkJoinWorkerThread.workQueue.push(task)/mon.externalPush(task) -> ForkJoinPool.push(task)/externalPush(task)

fork：

public final ForkJoinTask<V> fork() {Thread t;if ((t = Thread.currentThread()) instanceof ForkJoinWorkerThread)//当前线程是workerThread，任务直接放入workerThread当前的workQueue((ForkJoinWorkerThread)t).workQueue.push(this);mon.externalPush(this);//将任务添加到随机选取的队列中或新创建的队列中return this;}

push：

public class ForkJoinPool extends AbstractExecutorService {static final class WorkQueue {final void push(ForkJoinTask<?> task) {ForkJoinTask<?>[] a; ForkJoinPool p;int b = base, s = top, n;if ((a = array) != null) { // ignore if queue removed，队列被移除忽略int m = a.length - 1;// fenced write for task visibilityU.putOrderedObject(a, ((m & s) << ASHIFT) + ABASE, task);//任务加入队列中U.putOrderedInt(this, QTOP, s + 1);//挪动下次任务存放的槽的位置if ((n = s - b) <= 1) {//当前数组元素少时，进行唤醒当前线程；或者当没有活动线程或线程数较少时，添加新的线程if ((p = pool) != null)p.signalWork(p.workQueues, this);}else if (n >= m)//数组所有元素都满了进行2倍扩容growArray();}}final ForkJoinTask<?>[] growArray() {ForkJoinTask<?>[] oldA = array;int size = oldA != null ? oldA.length << 1 : INITIAL_QUEUE_CAPACITY;//2倍扩容或初始化if (size > MAXIMUM_QUEUE_CAPACITY)throw new RejectedExecutionException("Queue capacity exceeded");int oldMask, t, b;ForkJoinTask<?>[] a = array = new ForkJoinTask<?>[size];if (oldA != null && (oldMask = oldA.length - 1) >= 0 &&(t = top) - (b = base) > 0) {int mask = size - 1;do { // emulate poll from old array, push to new array遍历从旧数组中取出放到新数组中ForkJoinTask<?> x;int oldj = ((b & oldMask) << ASHIFT) + ABASE;int j = ((b & mask) << ASHIFT) + ABASE;x = (ForkJoinTask<?>)U.getObjectVolatile(oldA, oldj);//从旧数组中取出if (x != null &&pareAndSwapObject(oldA, oldj, x, null))//将旧数组取出的位置的对象置为nullU.putObjectVolatile(a, j, x);//放入新数组} while (++b != t);}return a;}}}

任务的消费

任务的消费的执行链路是ForkJoinTask.doExec() -> RecursiveTask.exec()/RecursiveAction.exec() -> 覆盖重写的compute()

doExec：任务的执行入口

final int doExec() {int s; boolean completed;if ((s = status) >= 0) {try {completed = exec();//消费任务} catch (Throwable rex) {return setExceptionalCompletion(rex);}if (completed)s = setCompletion(NORMAL);//任务执行完设置状态为NORMAL，并唤醒其他等待任务}return s;}protected abstract boolean exec();private int setCompletion(int completion) {for (int s;;) {if ((s = status) < 0)return s;if (pareAndSwapInt(this, STATUS, s, s | completion)) {//任务状态修改为NORMALif ((s >>> 16) != 0)//状态不是SMASKsynchronized (this) { notifyAll(); }//唤醒其他等待任务return completion;}}}/** The run status of this task 任务的运行状态*/volatile int status; // accessed directly by pool and workers由ForkJoinPool池或ForkJoinWorkerThread控制static final int DONE_MASK = 0xf0000000; // mask out non-completion bitsstatic final int NORMAL= 0xf0000000; // must be negativestatic final int CANCELLED = 0xc0000000; // must be < NORMALstatic final int EXCEPTIONAL = 0x80000000; // must be < CANCELLEDstatic final int SIGNAL= 0x00010000; // must be >= 1 << 16static final int SMASK = 0x0000ffff; // short bits for tags

任务真正执行处理逻辑

任务提交到ForkJoinPool，最终真正的是由继承Thread的ForkJoinWorkerThread的run方法来执行消费任务的，ForkJoinWorkerThread处理哪个任务是由join来出队的；

ForkJoinTask.join()

public final V join() {int s;if ((s = doJoin() & DONE_MASK) != NORMAL)reportException(s);return getRawResult();//得到返回结果}private int doJoin() {int s; Thread t; ForkJoinWorkerThread wt; ForkJoinPool.WorkQueue w;/*** (s = status) < 0 判断任务是否已经完成，完成直接返回s* 任务未完成：*1）线程是ForkJoinWorkerThread，tryUnpush任务出队然后消费任务doExec* 1.1）出队或消费失败，执行awaitJoin进行自旋，如果任务状态是完成就退出，否则继续尝试出队，直到任务完成或超时为止；*2）如果线程不是ForkJoinWorkerThread，执行externalAwaitDone进行出队消费*/return (s = status) < 0 ? s :((t = Thread.currentThread()) instanceof ForkJoinWorkerThread) ?(w = (wt = (ForkJoinWorkerThread)t).workQueue).tryUnpush(this) && (s = doExec()) < 0 ? s :wt.pool.awaitJoin(w, this, 0L) :externalAwaitDone();}private void reportException(int s) {if (s == CANCELLED)//取消throw new CancellationException();if (s == EXCEPTIONAL)//异常rethrow(getThrowableException());}

awaitJoin：

public class ForkJoinPool{final int awaitJoin(WorkQueue w, ForkJoinTask<?> task, long deadline) {int s = 0;if (task != null && w != null) {ForkJoinTask<?> prevJoin = w.currentJoin;U.putOrderedObject(w, QCURRENTJOIN, task);CountedCompleter<?> cc = (task instanceof CountedCompleter) ?(CountedCompleter<?>)task : null;for (;;) {if ((s = task.status) < 0)//任务完成退出break;if (cc != null)//当前任务即将完成，检查是否还有其他的等待任务，如果有//运行当前队列的其他任务，若当前的队列中没有任务了，则窃取其他队列的任务并运行helpComplete(w, cc, 0);//当前队列没有任务了，或遍历当前队列有没有任务，如果有且在top端取出来运行，或在队列中间使用EmptyTask替代原位置取出来运行，如果没有，执行helpStealerelse if (w.base == w.top || w.tryRemoveAndExec(task))helpStealer(w, task);//窃取其他队列的任务if ((s = task.status) < 0)break;long ms, ns;if (deadline == 0L)ms = 0L;else if ((ns = deadline - System.nanoTime()) <= 0L)//超时退出break;else if ((ms = TimeUnit.NANOSECONDS.toMillis(ns)) <= 0L)ms = 1L;if (tryCompensate(w)) {//当前队列阻塞了task.internalWait(ms);//进行等待U.getAndAddLong(this, CTL, AC_UNIT);}}U.putOrderedObject(w, QCURRENTJOIN, prevJoin);}return s;}}

externalAwaitDone：

private int externalAwaitDone() {/*** 当前任务是CountedCompleter* 1）是则执行mon.externalHelpComplete()* 2）否则执行mon.tryExternalUnpush(this)进行任务出队* 2.1）出队成功，进行doExec()消费，否则进行阻塞等待*/int s = ((this instanceof CountedCompleter) ? // try mon.externalHelpComplete((CountedCompleter<?>)this, 0) :mon.tryExternalUnpush(this) ? doExec() : 0);if (s >= 0 && (s = status) >= 0) {//任务未完成boolean interrupted = false;do {if (pareAndSwapInt(this, STATUS, s, s | SIGNAL)) {//任务状态标记为SIGNALsynchronized (this) {if (status >= 0) {try {wait(0L);//阻塞等待} catch (InterruptedException ie) {//有中断异常interrupted = true;//设置中断标识为true}}elsenotifyAll();//任务完成唤醒其他任务}}} while ((s = status) >= 0);if (interrupted)Thread.currentThread().interrupt();//当前线程进行中断}return s;}final int externalHelpComplete(CountedCompleter<?> task, int maxTasks) {WorkQueue[] ws; int n;int r = ThreadLocalRandom.getProbe();//没有任务直接结束，有任务则执行helpComplete//helpComplete：运行随机选取的队列的任务，若选取的队列中没有任务了，则窃取其他队列的任务并运行return ((ws = workQueues) == null || (n = ws.length) == 0) ? 0 :helpComplete(ws[(n - 1) & r & SQMASK], task, maxTasks);}

run和工作窃取

任务是由workThread来窃取的，workThread是一个线程。线程的所有逻辑都是由run()方法执行：

public class ForkJoinWorkerThread extends Thread {public void run() {if (workQueue.array == null) { // only run onceThrowable exception = null;try {onStart();//初始化状态pool.runWorker(workQueue);//处理任务队列} catch (Throwable ex) {exception = ex;//记录异常} finally {try {onTermination(exception);} catch (Throwable ex) {if (exception == null)exception = ex;} finally {pool.deregisterWorker(this, exception);//注销工作线程}}}}}public class ForkJoinPool{final void runWorker(WorkQueue w) {w.growArray(); // allocate queue，队列初始化int seed = w.hint;// initially holds randomization hintint r = (seed == 0) ? 1 : seed; // avoid 0 for xorShiftfor (ForkJoinTask<?> t;;) {//自旋if ((t = scan(w, r)) != null)//从队列中窃取任务成功，scan()进行任务窃取w.runTask(t);//执行任务，内部方法调用了doExec()进行任务的消费else if (!awaitWork(w, r))//队列没有任务了则结束break;r ^= r << 13; r ^= r >>> 17; r ^= r << 5; // xorshift}}}

scan：

private ForkJoinTask<?> scan(WorkQueue w, int r) {WorkQueue[] ws; int m;if ((ws = workQueues) != null && (m = ws.length - 1) > 0 && w != null) {int ss = w.scanState; // initially non-negativefor (int origin = r & m, k = origin, oldSum = 0, checkSum = 0;;) {WorkQueue q; ForkJoinTask<?>[] a; ForkJoinTask<?> t;int b, n; long c;if ((q = ws[k]) != null) { //随机选中了非空队列 qif ((n = (b = q.base) - q.top) < 0 &&(a = q.array) != null) {// non-emptylong i = (((a.length - 1) & b) << ASHIFT) + ABASE; //从尾部出队,b是尾部下标if ((t = ((ForkJoinTask<?>)U.getObjectVolatile(a, i))) != null &&q.base == b) {if (ss >= 0) {if (pareAndSwapObject(a, i, t, null)) { //利用cas出队q.base = b + 1;if (n < -1) // signal otherssignalWork(ws, q);return t; //出队成功，成功窃取一个任务！}}else if (oldSum == 0 && // try to activate 队列没有激活，尝试激活w.scanState < 0)tryRelease(c = ctl, ws[m & (int)c], AC_UNIT);}if (ss < 0) // refreshss = w.scanState;r ^= r << 1; r ^= r >>> 3; r ^= r << 10;origin = k = r & m; // move and rescanoldSum = checkSum = 0;continue;}checkSum += b;}//k = k + 1表示取下一个队列 如果（k + 1） & m == origin表示已经遍历完所有队列了if ((k = (k + 1) & m) == origin) { // continue until stable if ((ss >= 0 || (ss == (ss = w.scanState))) &&oldSum == (oldSum = checkSum)) {if (ss < 0 || w.qlock < 0) // already inactivebreak;int ns = ss | INACTIVE; // try to inactivatelong nc = ((SP_MASK & ns) |(UC_MASK & ((c = ctl) - AC_UNIT)));w.stackPred = (int)c; // hold prev stack topU.putInt(w, QSCANSTATE, ns);if (pareAndSwapLong(this, CTL, c, nc))ss = ns;elsew.scanState = ss; // back out}checkSum = 0;}}}return null;}

ForkJoinPool.runTask：

volatile int scanState; // versioned, <0: inactive; odd:scanning，版本标记，小于0暂停，奇数进行扫描其他任务static final int SCANNING= 1; // false when running tasks，有任务执行是false/*** Executes the given task and any remaining local tasks.* 执行给定的任务和任何剩余的本地任务*/final void runTask(ForkJoinTask<?> task) {if (task != null) {scanState &= ~SCANNING; // mark as busy，暂停扫描，当前有任务执行(currentSteal = task).doExec();//执行窃取的任务U.putOrderedObject(this, QCURRENTSTEAL, null); // release for GC，窃取的任务执行完置为nullexecLocalTasks();//执行本地的任务，即自己workQueue的任务，调用doExec执行到workQueue空为止ForkJoinWorkerThread thread = owner;if (++nsteals < 0)// collect on overflow，窃取计数溢出transferStealCount(pool);//重置窃取计数scanState |= SCANNING;//继续扫描队列if (thread != null)thread.afterTopLevelExec();}}static final class InnocuousForkJoinWorkerThread extends ForkJoinWorkerThread {@Override // to erase ThreadLocals，清除threadLocalsvoid afterTopLevelExec() {eraseThreadLocals();}/*** Erases ThreadLocals by nulling out Thread maps.*/final void eraseThreadLocals() {U.putObject(this, THREADLOCALS, null);//threadLocals置为nullU.putObject(this, INHERITABLETHREADLOCALS, null);//inheritablethreadlocals置为null}}

四.总结

对于fork/join来说，在使用时还是存在下面的一些问题的：

在使用JVM的时候我们要考虑OOM的问题，如果我们的任务处理时间非常耗时，并且处理的数据非常大的时候，会造成OOM；ForkJoin是通过多线程的方式进行处理任务，那么我们不得不考虑是否应该使用ForkJoin。因为当数据量不是特别大的时候，我们没有必要使用ForkJoin。因为多线程会涉及到上下文的切换，所以数据量不大的时候使用串行比使用多线程快； 项目中进行本地测试发现，业务层Service进行excel表数据(数据量几百)的复杂处理，进行单线程for循环统计消耗时间，然后与使用fork/join进行处理统计消耗时间，发现fork/join的消耗时间是单线程for的2倍；