1.在哪里调用了ForkJoinPool?
1.1 parallelStream示例
int result = Stream.of(1,2,3,4,5).parallel().map(x -> x + 1).reduce((a, b) -> a + b).get();
@SafeVarargs
@SuppressWarnings("varargs") // Creating a stream from an array is safe
public static<T> Stream<T> of(T... values) {
return Arrays.stream(values);
}
@Override
@SuppressWarnings("unchecked")
public final S parallel() {
sourceStage.parallel = true;
return (S) this;
}
@Override
@SuppressWarnings("unchecked")
public final <R> Stream<R> map(Function<? super P_OUT, ? extends R> mapper) {
Objects.requireNonNull(mapper);
return new StatelessOp<P_OUT, R>(this, StreamShape.REFERENCE,
StreamOpFlag.NOT_SORTED | StreamOpFlag.NOT_DISTINCT) {
@Override
Sink<P_OUT> opWrapSink(int flags, Sink<R> sink) {
return new Sink.ChainedReference<P_OUT, R>(sink) {
@Override
public void accept(P_OUT u) {
downstream.accept(mapper.apply(u));
}
};
}
};
}
@Override
public final Optional<P_OUT> reduce(BinaryOperator<P_OUT> accumulator) {
return evaluate(ReduceOps.makeRef(accumulator));
}
/**
* Evaluate the pipeline with a terminal operation to produce a result.
*
* @param <R> the type of result
* @param terminalOp the terminal operation to be applied to the pipeline.
* @return the result
*/
final <R> R evaluate(TerminalOp<E_OUT, R> terminalOp) {
assert getOutputShape() == terminalOp.inputShape();
if (linkedOrConsumed)
throw new IllegalStateException(MSG_STREAM_LINKED);
linkedOrConsumed = true;
return isParallel()
? terminalOp.evaluateParallel(this, sourceSpliterator(terminalOp.getOpFlags()))
: terminalOp.evaluateSequential(this, sourceSpliterator(terminalOp.getOpFlags()));
}
@Override
public <P_IN> R evaluateParallel(PipelineHelper<T> helper,
Spliterator<P_IN> spliterator) {
return new ReduceTask<>(this, helper, spliterator).invoke().get();
}
1.2 ReduceTask会调用invoke
@Override
public <P_IN> R evaluateParallel(PipelineHelper<T> helper,
Spliterator<P_IN> spliterator) {
return new ReduceTask<>(this, helper, spliterator).invoke().get();
}
调用的是ForkJoinTask的invoke:
public final V invoke() {
int s;
if ((s = doInvoke() & DONE_MASK) != NORMAL)
reportException(s);
return getRawResult();
}
private int doInvoke() {
int s; Thread t; ForkJoinWorkerThread wt;
return (s = doExec()) < 0 ? s :
((t = Thread.currentThread()) instanceof ForkJoinWorkerThread) ?
(wt = (ForkJoinWorkerThread)t).pool.
awaitJoin(wt.workQueue, this, 0L) :
externalAwaitDone();
}
final int doExec() {
int s; boolean completed;
if ((s = status) >= 0) {
try {
completed = exec();
} catch (Throwable rex) {
return setExceptionalCompletion(rex);
}
if (completed)
s = setCompletion(NORMAL);
}
return s;
}
调用CountedCompleter的exec():
/**
* Implements execution conventions for CountedCompleters.
*/
protected final boolean exec() {
compute();
return false;
}
调用stream包下面的AbstractTask.compute:
@Override
public void compute() {
Spliterator<P_IN> rs = spliterator, ls; // right, left spliterators
long sizeEstimate = rs.estimateSize();
long sizeThreshold = getTargetSize(sizeEstimate);
boolean forkRight = false;
@SuppressWarnings("unchecked") K task = (K) this;
while (sizeEstimate > sizeThreshold && (ls = rs.trySplit()) != null) {
K leftChild, rightChild, taskToFork;
task.leftChild = leftChild = task.makeChild(ls);
task.rightChild = rightChild = task.makeChild(rs);
task.setPendingCount(1);
if (forkRight) {
forkRight = false;
rs = ls;
task = leftChild;
taskToFork = rightChild;
}
else {
forkRight = true;
task = rightChild;
taskToFork = leftChild;
}
taskToFork.fork();
sizeEstimate = rs.estimateSize();
}
task.setLocalResult(task.doLeaf());
task.tryComplete();
}
调用ForkJoinTask的fork:
public final ForkJoinTask<V> fork() {
Thread t;
if ((t = Thread.currentThread()) instanceof ForkJoinWorkerThread)
((ForkJoinWorkerThread)t).workQueue.push(this);
else
ForkJoinPool.common.externalPush(this);
return this;
}
2.并行的核心就是任务拆分
2.1 调用stream包下面的AbstractTask.compute
@Override
public void compute() {
Spliterator<P_IN> rs = spliterator, ls; // right, left spliterators
long sizeEstimate = rs.estimateSize();
long sizeThreshold = getTargetSize(sizeEstimate);
boolean forkRight = false;
@SuppressWarnings("unchecked") K task = (K) this;
while (sizeEstimate > sizeThreshold && (ls = rs.trySplit()) != null) {
K leftChild, rightChild, taskToFork;
task.leftChild = leftChild = task.makeChild(ls);
task.rightChild = rightChild = task.makeChild(rs);
task.setPendingCount(1);
if (forkRight) {
forkRight = false;
rs = ls;
task = leftChild;
taskToFork = rightChild;
}
else {
forkRight = true;
task = rightChild;
taskToFork = leftChild;
}
taskToFork.fork();
sizeEstimate = rs.estimateSize();
}
task.setLocalResult(task.doLeaf());
task.tryComplete();
}
2.2 ReduceTask.doLeaf()
private static final class ReduceTask<P_IN, P_OUT, R,
S extends AccumulatingSink<P_OUT, R, S>>
extends AbstractTask<P_IN, P_OUT, S, ReduceTask<P_IN, P_OUT, R, S>> {
private final ReduceOp<P_OUT, R, S> op;
ReduceTask(ReduceOp<P_OUT, R, S> op,
PipelineHelper<P_OUT> helper,
Spliterator<P_IN> spliterator) {
super(helper, spliterator);
this.op = op;
}
ReduceTask(ReduceTask<P_IN, P_OUT, R, S> parent,
Spliterator<P_IN> spliterator) {
super(parent, spliterator);
this.op = parent.op;
}
@Override
protected ReduceTask<P_IN, P_OUT, R, S> makeChild(Spliterator<P_IN> spliterator) {
return new ReduceTask<>(this, spliterator);
}
@Override
protected S doLeaf() {
return helper.wrapAndCopyInto(op.makeSink(), spliterator);
}
@Override
public void onCompletion(CountedCompleter<?> caller) {
if (!isLeaf()) {
S leftResult = leftChild.getLocalResult();
leftResult.combine(rightChild.getLocalResult());
setLocalResult(leftResult);
}
// GC spliterator, left and right child
super.onCompletion(caller);
}
}
根据Stream流水线原理中可知ReduceTask.doLeaf()也即求值操作,针对这部分流进行分部分求值:
@Override
final <P_IN, S extends Sink<E_OUT>> S wrapAndCopyInto(S sink, Spliterator<P_IN> spliterator) {
copyInto(wrapSink(Objects.requireNonNull(sink)), spliterator);
return sink;
}
2.3 ReduceTask.onCompletion()
AbstractTask.compute最后调用tryComplete()。
CountedCompleter.tryComplete():
public final void tryComplete() {
CountedCompleter<?> a = this, s = a;
for (int c;;) {
if ((c = a.pending) == 0) {
a.onCompletion(s);
if ((a = (s = a).completer) == null) {
s.quietlyComplete();
return;
}
}
else if (U.compareAndSwapInt(a, PENDING, c, c - 1))
return;
}
}
ReduceTask的onCompletion:
@Override
public void onCompletion(CountedCompleter<?> caller) {
if (!isLeaf()) {
S leftResult = leftChild.getLocalResult();
leftResult.combine(rightChild.getLocalResult());
setLocalResult(leftResult);
}
// GC spliterator, left and right child
super.onCompletion(caller);
}
根据Stream流水线原理中可知S代表终止操作的Sink:
public static TerminalOp<Integer, OptionalInt>
makeInt(IntBinaryOperator operator) {
Objects.requireNonNull(operator);
class ReducingSink
implements AccumulatingSink<Integer, OptionalInt, ReducingSink>, Sink.OfInt {
private boolean empty;
private int state;
public void begin(long size) {
empty = true;
state = 0;
}
@Override
public void accept(int t) {
if (empty) {
empty = false;
state = t;
}
else {
state = operator.applyAsInt(state, t);
}
}
@Override
public OptionalInt get() {
return empty ? OptionalInt.empty() : OptionalInt.of(state);
}
@Override
public void combine(ReducingSink other) {
if (!other.empty)
accept(other.state);
}
}
return new ReduceOp<Integer, OptionalInt, ReducingSink>(StreamShape.INT_VALUE) {
@Override
public ReducingSink makeSink() {
return new ReducingSink();
}
};
}
3.使用陷阱
public static String query(String question) {
List<String> engines = new ArrayList<String>() {{
add("http://www.google.com/?q=");
add("http://duckduckgo.com/?q=");
add("http://www.bing.com/search?q=");
}};
// get element as soon as it is available
Optional<String> result = engines.stream().parallel().map((base) -> {
String url = base + question;
// open connection and fetch the result
return WS.url(url).get();
}).findAny();
return result.get();
}
查询搜索引擎是一个阻塞操作。所以在某时刻所有线程都会调用get()方法并且在那里等待结果返回。某个时间所有ForkJoinPool.common()的线程都会被用光。也就是说,下一次你调用这个查询方法,就可能会在一个时间与其他的parallel stream同时运行,而导致第二个任务的性能大大受损。
这意味着任何依赖parallel streams的程序在什么别的东西占用着common ForkJoinPool时将会变得不可预知并且暗藏危机。