/**
* Copyright 2014 Netflix, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package rx.internal.util;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicIntegerArray;
import java.util.concurrent.atomic.AtomicReference;
import java.util.concurrent.atomic.AtomicReferenceArray;
import rx.Subscription;
import rx.functions.Func1;
/**
* Add/Remove without object allocation (after initial construction).
* <p>
* This is meant for hundreds or single-digit thousands of elements that need
* to be rapidly added and randomly or sequentially removed while avoiding object allocation.
* <p>
* On Intel Core i7, 2.3Mhz, Mac Java 8:
* <p>
* - adds per second single-threaded => ~32,598,500 for 100
* - adds per second single-threaded => ~23,200,000 for 10,000
* - adds + removes per second single-threaded => 15,562,100 for 100
* - adds + removes per second single-threaded => 8,760,000 for 10,000
*
* <pre> {@code
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 263571.721 9856.994 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 1763.417 211.998 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 139850.115 17143.705 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 809.982 72.931 ops/s
* } </pre>
*
* @param <E>
*/
public final class IndexedRingBuffer<E> implements Subscription {
private static final ObjectPool<IndexedRingBuffer> POOL = new ObjectPool<IndexedRingBuffer>() {
@Override
protected IndexedRingBuffer createObject() {
return new IndexedRingBuffer();
}
};
public final static IndexedRingBuffer getInstance() {
return POOL.borrowObject();
}
private final ElementSection<E> elements = new ElementSection<E>();
private final IndexSection removed = new IndexSection();
/* package for unit testing */final AtomicInteger index = new AtomicInteger();
/* package for unit testing */final AtomicInteger removedIndex = new AtomicInteger();
// default size of ring buffer
/**
* Set at 256 ... Android defaults far smaller which likely will never hit the use cases that require the higher buffers.
* <p>
* The 10000 size test represents something that should be a rare use case (merging 10000 concurrent Observables for example)
*
* <pre> {@code
* ./gradlew benchmarks '-Pjmh=-f 1 -tu s -bm thrpt -wi 5 -i 5 -r 1 .*IndexedRingBufferPerf.*'
*
* 1024
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 269292.006 6013.347 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 2217.103 163.396 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 139349.608 9397.232 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 1045.323 30.991 ops/s
*
* 512
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 270919.870 5381.793 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 1724.436 42.287 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 141478.813 3696.030 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 719.447 75.629 ops/s
*
*
* 256
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 272042.605 7954.982 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 1101.329 23.566 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 140479.804 6389.060 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 397.306 24.222 ops/s
*
* 128
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 263065.312 11168.941 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 581.708 17.397 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 138051.488 4618.935 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 176.873 35.669 ops/s
*
* 32
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 250737.473 17260.148 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 144.725 26.284 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 118832.832 9082.658 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 32.133 8.048 ops/s
*
* 8
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 209192.847 25558.124 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 26.520 3.100 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 100200.463 1854.259 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 8.456 2.114 ops/s
*
* 2
*
* Benchmark (size) Mode Samples Score Score error Units
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 100 thrpt 5 96549.208 4427.239 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAdd 10000 thrpt 5 6.637 2.025 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 100 thrpt 5 34553.169 4904.197 ops/s
* r.i.IndexedRingBufferPerf.indexedRingBufferAddRemove 10000 thrpt 5 2.159 0.700 ops/s
* } </pre>
*
* Impact of IndexedRingBuffer size on merge
*
* <pre> {@code
* ./gradlew benchmarks '-Pjmh=-f 1 -tu s -bm thrpt -wi 5 -i 5 -r 1 .*OperatorMergePerf.*'
*
* 512
*
* Benchmark (size) Mode Samples Score Score error Units
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1 thrpt 5 5282500.038 530541.761 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000 thrpt 5 49327.272 6382.189 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000000 thrpt 5 53.025 4.724 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1 thrpt 5 97395.148 2489.303 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1000 thrpt 5 4.723 1.479 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1 thrpt 5 4534067.250 116321.725 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 100 thrpt 5 458561.098 27652.081 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1000 thrpt 5 43267.381 2648.107 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1 thrpt 5 5581051.672 144191.849 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1000 thrpt 5 50.643 4.354 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1 thrpt 5 76437.644 959.748 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1000 thrpt 5 2965.306 272.928 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1 thrpt 5 5026522.098 364196.255 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000 thrpt 5 34926.819 938.612 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000000 thrpt 5 33.342 1.701 ops/s
*
*
* 128
*
* Benchmark (size) Mode Samples Score Score error Units
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1 thrpt 5 5144891.776 271990.561 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000 thrpt 5 53580.161 2370.204 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000000 thrpt 5 53.265 2.236 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1 thrpt 5 96634.426 1417.430 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1000 thrpt 5 4.648 0.255 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1 thrpt 5 4601280.220 53157.938 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 100 thrpt 5 463394.568 58612.882 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1000 thrpt 5 50503.565 2394.168 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1 thrpt 5 5490315.842 228654.817 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1000 thrpt 5 50.661 3.385 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1 thrpt 5 74716.169 7413.642 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1000 thrpt 5 3009.476 277.075 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1 thrpt 5 4953313.642 307512.126 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000 thrpt 5 35335.579 2368.377 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000000 thrpt 5 37.450 0.655 ops/s
*
* 32
*
* Benchmark (size) Mode Samples Score Score error Units
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1 thrpt 5 4975957.497 365423.694 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000 thrpt 5 52141.226 5056.658 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000000 thrpt 5 53.663 2.671 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1 thrpt 5 96507.893 1833.371 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1000 thrpt 5 4.850 0.782 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1 thrpt 5 4557128.302 118516.934 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 100 thrpt 5 339005.037 10594.737 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1000 thrpt 5 50781.535 6071.787 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1 thrpt 5 5604920.068 209285.840 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1000 thrpt 5 50.413 7.496 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1 thrpt 5 76098.942 558.187 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1000 thrpt 5 2988.137 193.255 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1 thrpt 5 5177255.256 150253.086 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000 thrpt 5 34772.490 909.967 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000000 thrpt 5 34.847 0.606 ops/s
*
* 8
*
* Benchmark (size) Mode Samples Score Score error Units
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1 thrpt 5 5027331.903 337986.410 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000 thrpt 5 51746.540 3585.450 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000000 thrpt 5 52.682 4.026 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1 thrpt 5 96805.587 2868.112 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1000 thrpt 5 4.598 0.290 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1 thrpt 5 4390912.630 300687.310 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 100 thrpt 5 458615.731 56125.958 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1000 thrpt 5 49033.105 6132.936 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1 thrpt 5 5090614.100 649439.778 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1000 thrpt 5 48.548 3.586 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1 thrpt 5 72285.482 16820.952 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1000 thrpt 5 2981.576 316.140 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1 thrpt 5 4993609.293 267975.397 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000 thrpt 5 33228.972 1554.924 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000000 thrpt 5 32.994 3.615 ops/s
*
*
* 2
*
* Benchmark (size) Mode Samples Score Score error Units
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1 thrpt 5 5103812.234 939461.192 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000 thrpt 5 51491.116 3790.056 ops/s
* r.o.OperatorMergePerf.merge1SyncStreamOfN 1000000 thrpt 5 54.043 2.340 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1 thrpt 5 96575.834 13416.541 ops/s
* r.o.OperatorMergePerf.mergeNAsyncStreamsOfN 1000 thrpt 5 4.740 0.047 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1 thrpt 5 4435909.832 899133.671 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 100 thrpt 5 392382.445 59814.783 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOf1 1000 thrpt 5 50429.258 7489.849 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1 thrpt 5 5637321.803 161838.195 ops/s
* r.o.OperatorMergePerf.mergeNSyncStreamsOfN 1000 thrpt 5 51.065 2.138 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1 thrpt 5 76366.764 2631.710 ops/s
* r.o.OperatorMergePerf.mergeTwoAsyncStreamsOfN 1000 thrpt 5 2978.302 296.418 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1 thrpt 5 5280829.290 1602542.493 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000 thrpt 5 35070.518 3565.672 ops/s
* r.o.OperatorMergePerf.oneStreamOfNthatMergesIn1 1000000 thrpt 5 34.501 0.991 ops/s
*
* } </pre>
*/
static int _size = 256;
static {
// lower default for Android (https://github.com/ReactiveX/RxJava/issues/1820)
if (PlatformDependent.isAndroid()) {
_size = 8;
}
// possible system property for overriding
String sizeFromProperty = System.getProperty("rx.indexed-ring-buffer.size"); // also see RxRingBuffer
if (sizeFromProperty != null) {
try {
_size = Integer.parseInt(sizeFromProperty);
} catch (Exception e) {
System.err.println("Failed to set 'rx.indexed-ring-buffer.size' with value " + sizeFromProperty + " => " + e.getMessage());
}
}
}
/* package for unit testing */static final int SIZE = _size;
/**
* This resets the arrays, nulls out references and returns it to the pool.
* This extra CPU cost is far smaller than the object allocation cost of not pooling.
*/
public void releaseToPool() {
// need to clear all elements so we don't leak memory
int maxIndex = index.get();
int realIndex = 0;
ElementSection<E> section = elements;
outer: while (section != null) {
for (int i = 0; i < SIZE; i++, realIndex++) {
if (realIndex >= maxIndex) {
section = null;
break outer;
}
// we can use lazySet here because we are nulling things out and not accessing them again
// (relative on Mac Intel i7) lazySet gets us ~30m vs ~26m ops/second in the JMH test (100 adds per release)
section.array.set(i, null);
}
section = section.next.get();
}
index.set(0);
removedIndex.set(0);
POOL.returnObject(this);
}
@Override
public void unsubscribe() {
releaseToPool();
}
private IndexedRingBuffer() {
}
/**
* Add an element and return the index where it was added to allow removal.
*
* @param e
* @return
*/
public int add(E e) {
int i = getIndexForAdd();
if (i < SIZE) {
// fast-path when we are in the first section
elements.array.set(i, e);
return i;
} else {
int sectionIndex = i % SIZE;
getElementSection(i).array.set(sectionIndex, e);
return i;
}
}
public E remove(int index) {
E e;
if (index < SIZE) {
// fast-path when we are in the first section
e = elements.array.getAndSet(index, null);
} else {
int sectionIndex = index % SIZE;
e = getElementSection(index).array.getAndSet(sectionIndex, null);
}
pushRemovedIndex(index);
return e;
}
private IndexSection getIndexSection(int index) {
// short-cut the normal case
if (index < SIZE) {
return removed;
}
// if we have passed the first array we get more complicated and do recursive chaining
int numSections = index / SIZE;
IndexSection a = removed;
for (int i = 0; i < numSections; i++) {
a = a.getNext();
}
return a;
}
private ElementSection<E> getElementSection(int index) {
// short-cut the normal case
if (index < SIZE) {
return elements;
}
// if we have passed the first array we get more complicated and do recursive chaining
int numSections = index / SIZE;
ElementSection<E> a = elements;
for (int i = 0; i < numSections; i++) {
a = a.getNext();
}
return a;
}
private synchronized int getIndexForAdd() {
/*
* Synchronized as I haven't yet figured out a way to do this in an atomic way that doesn't involve object allocation
*/
int i;
int ri = getIndexFromPreviouslyRemoved();
if (ri >= 0) {
if (ri < SIZE) {
// fast-path when we are in the first section
i = removed.getAndSet(ri, -1);
} else {
int sectionIndex = ri % SIZE;
i = getIndexSection(ri).getAndSet(sectionIndex, -1);
}
if (i == index.get()) {
// if it was the last index removed, when we pick it up again we want to increment
index.getAndIncrement();
}
} else {
i = index.getAndIncrement();
}
return i;
}
/**
* Returns -1 if nothing, 0 or greater if the index should be used
*
* @return
*/
private synchronized int getIndexFromPreviouslyRemoved() {
/*
* Synchronized as I haven't yet figured out a way to do this in an atomic way that doesn't involve object allocation
*/
// loop because of CAS
while (true) {
int currentRi = removedIndex.get();
if (currentRi > 0) {
// claim it
if (removedIndex.compareAndSet(currentRi, currentRi - 1)) {
return currentRi - 1;
}
} else {
// do nothing
return -1;
}
}
}
private synchronized void pushRemovedIndex(int elementIndex) {
/*
* Synchronized as I haven't yet figured out a way to do this in an atomic way that doesn't involve object allocation
*/
int i = removedIndex.getAndIncrement();
if (i < SIZE) {
// fast-path when we are in the first section
removed.set(i, elementIndex);
} else {
int sectionIndex = i % SIZE;
getIndexSection(i).set(sectionIndex, elementIndex);
}
}
@Override
public boolean isUnsubscribed() {
return false;
}
public int forEach(Func1<? super E, Boolean> action) {
return forEach(action, 0);
}
/**
*
* @param action
* that processes each item and returns true if it wants to continue to the next
* @return int of next index to process, or last index seen if it exited early
*/
public int forEach(Func1<? super E, Boolean> action, int startIndex) {
int endedAt = forEach(action, startIndex, index.get());
if (startIndex > 0 && endedAt == index.get()) {
// start at the beginning again and go up to startIndex
endedAt = forEach(action, 0, startIndex);
} else if (endedAt == index.get()) {
// start back at the beginning
endedAt = 0;
}
return endedAt;
}
private int forEach(Func1<? super E, Boolean> action, int startIndex, int endIndex) {
int lastIndex = startIndex;
int maxIndex = index.get();
int realIndex = startIndex;
ElementSection<E> section = elements;
if (startIndex >= SIZE) {
// move into the correct section
section = getElementSection(startIndex);
startIndex = startIndex % SIZE;
}
outer: while (section != null) {
for (int i = startIndex; i < SIZE; i++, realIndex++) {
if (realIndex >= maxIndex || realIndex >= endIndex) {
section = null;
break outer;
}
E element = section.array.get(i);
if (element == null) {
continue;
}
lastIndex = realIndex;
boolean continueLoop = action.call(element);
if (!continueLoop) {
return lastIndex;
}
}
section = section.next.get();
startIndex = 0; // reset to start for next section
}
// return the OutOfBounds index position if we processed all of them ... the one we should be less-than
return realIndex;
}
private static class ElementSection<E> {
private final AtomicReferenceArray<E> array = new AtomicReferenceArray<E>(SIZE);
private final AtomicReference<ElementSection<E>> next = new AtomicReference<ElementSection<E>>();
ElementSection<E> getNext() {
if (next.get() != null) {
return next.get();
} else {
ElementSection<E> newSection = new ElementSection<E>();
if (next.compareAndSet(null, newSection)) {
// we won
return newSection;
} else {
// we lost so get the value that won
return next.get();
}
}
}
}
private static class IndexSection {
private final AtomicIntegerArray unsafeArray = new AtomicIntegerArray(SIZE);
public int getAndSet(int expected, int newValue) {
return unsafeArray.getAndSet(expected, newValue);
}
public void set(int i, int elementIndex) {
unsafeArray.set(i, elementIndex);
}
private final AtomicReference<IndexSection> _next = new AtomicReference<IndexSection>();
IndexSection getNext() {
if (_next.get() != null) {
return _next.get();
} else {
IndexSection newSection = new IndexSection();
if (_next.compareAndSet(null, newSection)) {
// we won
return newSection;
} else {
// we lost so get the value that won
return _next.get();
}
}
}
}
}