ViewVC Help

View File | Revision Log | Show Annotations | Download File | Root Listing

root/jsr166/jsr166/src/jsr166e/ConcurrentHashMapV8.java

Comparing jsr166/src/jsr166e/ConcurrentHashMapV8.java (file contents):
Revision 1.23 by jsr166, Sun Sep 11 04:25:00 2011 UTC vs.
Revision 1.24 by dl, Thu Sep 15 14:25:46 2011 UTC

#	Line 6 | Line 6
6
7		package jsr166e;
8		import jsr166e.LongAdder;
9	+	import java.util.Arrays;
10		import java.util.Map;
11		import java.util.Set;
12		import java.util.Collection;
#	Line 19 | Line 20 | import java.util.Enumeration;
20		import java.util.ConcurrentModificationException;
21		import java.util.NoSuchElementException;
22		import java.util.concurrent.ConcurrentMap;
23	+	import java.util.concurrent.locks.LockSupport;
24		import java.io.Serializable;
25
26		/**
#	Line 54 | Line 56 | import java.io.Serializable;
56		* <p> The table is dynamically expanded when there are too many
57		* collisions (i.e., keys that have distinct hash codes but fall into
58		* the same slot modulo the table size), with the expected average
59	<	* effect of maintaining roughly two bins per mapping. There may be
60	<	* much variance around this average as mappings are added and
61	<	* removed, but overall, this maintains a commonly accepted time/space
62	<	* tradeoff for hash tables.  However, resizing this or any other kind
63	<	* of hash table may be a relatively slow operation. When possible, it
64	<	* is a good idea to provide a size estimate as an optional {@code
59	>	* effect of maintaining roughly two bins per mapping (corresponding
60	>	* to a 0.75 load factor threshold for resizing). There may be much
61	>	* variance around this average as mappings are added and removed, but
62	>	* overall, this maintains a commonly accepted time/space tradeoff for
63	>	* hash tables.  However, resizing this or any other kind of hash
64	>	* table may be a relatively slow operation. When possible, it is a
65	>	* good idea to provide a size estimate as an optional {@code
66		* initialCapacity} constructor argument. An additional optional
67		* {@code loadFactor} constructor argument provides a further means of
68		* customizing initial table capacity by specifying the table density
#	Line 121 | Line 124 | public class ConcurrentHashMapV8<K, V>
124		* The primary design goal of this hash table is to maintain
125		* concurrent readability (typically method get(), but also
126		* iterators and related methods) while minimizing update
127	<	* contention.
127	>	* contention. Secondary goals are to keep space consumption about
128	>	* the same or better than java.util.HashMap, and to support high
129	>	* initial insertion rates on an empty table by many threads.
130		*
131		* Each key-value mapping is held in a Node.  Because Node fields
132		* can contain special values, they are defined using plain Object
#	Line 139 | Line 144 | public class ConcurrentHashMapV8<K, V>
144		* we use intrinsics (sun.misc.Unsafe) operations.  The lists of
145		* nodes within bins are always accurately traversable under
146		* volatile reads, so long as lookups check hash code and
147	<	* non-nullness of value before checking key equality. (All valid
148	<	* hash codes are nonnegative. Negative values are reserved for
149	<	* special forwarding nodes; see below.)
147	>	* non-nullness of value before checking key equality.
148	>	*
149	>	* We use the top two bits of Node hash fields for control
150	>	* purposes -- they are available anyway because of addressing
151	>	* constraints.  As explained further below, these top bits are
152	>	* usd as follows:
153	>	*  00 - Normal
154	>	*  01 - Locked
155	>	*  11 - Locked and may have a thread waiting for lock
156	>	*  10 - Node is a forwarding node
157	>	*
158	>	* The lower 30 bits of each Node's hash field contain a
159	>	* transformation (for better randomization -- method "spread") of
160	>	* the key's hash code, except for forwarding nodes, for which the
161	>	* lower bits are zero (and so always have hash field == "MOVED").
162		*
163		* Insertion (via put or putIfAbsent) of the first node in an
164		* empty bin is performed by just CASing it to the bin.  This is
165	<	* on average by far the most common case for put operations.
166	<	* Other update operations (insert, delete, and replace) require
167	<	* locks.  We do not want to waste the space required to associate
168	<	* a distinct lock object with each bin, so instead use the first
169	<	* node of a bin list itself as a lock, using plain "synchronized"
170	<	* locks. These save space and we can live with block-structured
171	<	* lock/unlock operations. Using the first node of a list as a
172	<	* lock does not by itself suffice though. When a node is locked,
173	<	* any update must first validate that it is still the first node,
174	<	* and retry if not. Because new nodes are always appended to
175	<	* lists, once a node is first in a bin, it remains first until
176	<	* deleted or the bin becomes invalidated.  However, operations
177	<	* that only conditionally update can and sometimes do inspect
178	<	* nodes until the point of update. This is a converse of sorts to
179	<	* the lazy locking technique described by Herlihy & Shavit.
165	>	* by far the most common case for put operations.  Other update
166	>	* operations (insert, delete, and replace) require locks.  We do
167	>	* not want to waste the space required to associate a distinct
168	>	* lock object with each bin, so instead use the first node of a
169	>	* bin list itself as a lock. Blocking support for these locks
170	>	* relies on the builtin "synchronized" monitors.  However, we
171	>	* also need a tryLock construction, so we overlay these by using
172	>	* bits of the Node hash field for lock control (see above), and
173	>	* so normally use builtin monitors only for blocking and
174	>	* signalling using wait/notifyAll constructions. See
175	>	* Node.tryAwaitLock.
176	>	*
177	>	* Using the first node of a list as a lock does not by itself
178	>	* suffice though: When a node is locked, any update must first
179	>	* validate that it is still the first node after locking it, and
180	>	* retry if not. Because new nodes are always appended to lists,
181	>	* once a node is first in a bin, it remains first until deleted
182	>	* or the bin becomes invalidated.  However, operations that only
183	>	* conditionally update may inspect nodes until the point of
184	>	* update. This is a converse of sorts to the lazy locking
185	>	* technique described by Herlihy & Shavit.
186		*
187	<	* The main disadvantage of this approach is that most update
187	>	* The main disadvantage of per-bin locks is that other update
188		* operations on other nodes in a bin list protected by the same
189		* lock can stall, for example when user equals() or mapping
190		* functions take a long time.  However, statistically, this is
#	Line 187 | Line 210 | public class ConcurrentHashMapV8<K, V>
210		* that these assumptions hold unless users define exactly the
211		* same value for too many hashCodes.
212		*
213	<	* The table is resized when occupancy exceeds a threshold.  Only
214	<	* a single thread performs the resize (using field "resizing", to
215	<	* arrange exclusion), but the table otherwise remains usable for
216	<	* reads and updates. Resizing proceeds by transferring bins, one
217	<	* by one, from the table to the next table.  Upon transfer, the
218	<	* old table bin contains only a special forwarding node (with
219	<	* negative hash field) that contains the next table as its
220	<	* key. On encountering a forwarding node, access and update
221	<	* operations restart, using the new table. To ensure concurrent
222	<	* readability of traversals, transfers must proceed from the last
223	<	* bin (table.length - 1) up towards the first.  Upon seeing a
224	<	* forwarding node, traversals (see class InternalIterator)
225	<	* arrange to move to the new table for the rest of the traversal
226	<	* without revisiting nodes.  This constrains bin transfers to a
227	<	* particular order, and so can block indefinitely waiting for the
228	<	* next lock, and other threads cannot help with the transfer.
229	<	* However, expected stalls are infrequent enough to not warrant
230	<	* the additional overhead of access and iteration schemes that
231	<	* could admit out-of-order or concurrent bin transfers.
213	>	* The table is resized when occupancy exceeds an occupancy
214	>	* threshold (nominally, 0.75, but see below).  Only a single
215	>	* thread performs the resize (using field "sizeCtl", to arrange
216	>	* exclusion), but the table otherwise remains usable for reads
217	>	* and updates. Resizing proceeds by transferring bins, one by
218	>	* one, from the table to the next table.  Because we are using
219	>	* power-of-two expansion, the elements from each bin must either
220	>	* stay at same index, or move with a power of two offset. We
221	>	* eliminate unnecessary node creation by catching cases where old
222	>	* nodes can be reused because their next fields won't change.  On
223	>	* average, only about one-sixth of them need cloning when a table
224	>	* doubles. The nodes they replace will be garbage collectable as
225	>	* soon as they are no longer referenced by any reader thread that
226	>	* may be in the midst of concurrently traversing table.  Upon
227	>	* transfer, the old table bin contains only a special forwarding
228	>	* node (with hash field "MOVED") that contains the next table as
229	>	* its key. On encountering a forwarding node, access and update
230	>	* operations restart, using the new table.
231	>	*
232	>	* Each bin transfer requires its bin lock. However, unlike other
233	>	* cases, a transfer can skip a bin if it fails to acquire its
234	>	* lock, and revisit it later. Method rebuild maintains a buffer
235	>	* of TRANSFER_BUFFER_SIZE bins that have been skipped because of
236	>	* failure to acquire a lock, and blocks only if none are
237	>	* available (i.e., only very rarely).  The transfer operation
238	>	* must also ensure that all accessible bins in both the old and
239	>	* new table are usable by any traversal.  When there are no lock
240	>	* acquisition failures, this is arranged simply by proceeding
241	>	* from the last bin (table.length - 1) up towards the first.
242	>	* Upon seeing a forwarding node, traversals (see class
243	>	* InternalIterator) arrange to move to the new table without
244	>	* revisiting nodes.  However, when any node is skipped during a
245	>	* transfer, all earlier table bins may have become visible, so
246	>	* are initialized with a reverse-forwarding node back to the old
247	>	* table until the new ones are established. (This sometimes
248	>	* requires transiently locking a forwarding node, which is
249	>	* possible under the above encoding.) These more expensive
250	>	* mechanics trigger only when necessary.
251		*
252	<	* This traversal scheme also applies to partial traversals of
252	>	* The traversal scheme also applies to partial traversals of
253		* ranges of bins (via an alternate InternalIterator constructor)
254		* to support partitioned aggregate operations (that are not
255		* otherwise implemented yet).  Also, read-only operations give up
#	Line 218 | Line 260 | public class ConcurrentHashMapV8<K, V>
260		* Lazy table initialization minimizes footprint until first use,
261		* and also avoids resizings when the first operation is from a
262		* putAll, constructor with map argument, or deserialization.
263	<	* These cases attempt to override the targetCapacity used in
264	<	* growTable. These harmlessly fail to take effect in cases of
223	<	* races with other ongoing resizings. Uses of the threshold and
224	<	* targetCapacity during attempted initializations or resizings
225	<	* are racy but fall back on checks to preserve correctness.
263	>	* These cases attempt to override the initial capacity settings,
264	>	* but harmlessly fail to take effect in cases of races.
265		*
266		* The element count is maintained using a LongAdder, which avoids
267		* contention on updates but can encounter cache thrashing if read
#	Line 233 | Line 272 | public class ConcurrentHashMapV8<K, V>
272		* is around 13%, meaning that only about 1 in 8 puts check
273		* threshold (and after resizing, many fewer do so). But this
274		* approximation has high variance for small table sizes, so we
275	<	* check on any collision for sizes <= 64.  Further, to increase
237	<	* the probability that a resize occurs soon enough, we offset the
238	<	* threshold (see THRESHOLD_OFFSET) by the expected number of puts
239	<	* between checks.
275	>	* check on any collision for sizes <= 64.
276		*
277		* Maintaining API and serialization compatibility with previous
278		* versions of this class introduces several oddities. Mainly: We
279		* leave untouched but unused constructor arguments refering to
280	<	* concurrencyLevel. We also declare an unused "Segment" class
281	<	* that is instantiated in minimal form only when serializing.
280	>	* concurrencyLevel. We accept a loadFactor constructor argument,
281	>	* but apply it only to initial table capacity (which is the only
282	>	* time that we can guarantee to honor it.) We also declare an
283	>	* unused "Segment" class that is instantiated in minimal form
284	>	* only when serializing.
285		*/
286
287		/* ---------------- Constants -------------- */
#	Line 250 | Line 289 | public class ConcurrentHashMapV8<K, V>
289		/**
290		* The largest possible table capacity.  This value must be
291		* exactly 1<<30 to stay within Java array allocation and indexing
292	<	* bounds for power of two table sizes.
292	>	* bounds for power of two table sizes, and is further required
293	>	* because the top two bits of 32bit hash fields are used for
294	>	* control purposes.
295		*/
296		private static final int MAXIMUM_CAPACITY = 1 << 30;
297
#	Line 261 | Line 302 | public class ConcurrentHashMapV8<K, V>
302		private static final int DEFAULT_CAPACITY = 16;
303
304		/**
305	+	* The largest possible (non-power of two) array size.
306	+	* Needed by toArray and related methods.
307	+	*/
308	+	static final int MAX_ARRAY_SIZE = Integer.MAX_VALUE - 8;
309	+
310	+	/**
311	+	* The default concurrency level for this table. Unused but
312	+	* defined for compatibility with previous versions of this class.
313	+	*/
314	+	private static final int DEFAULT_CONCURRENCY_LEVEL = 16;
315	+
316	+	/**
317		* The load factor for this table. Overrides of this value in
318		* constructors affect only the initial table capacity.  The
319	<	* actual floating point value isn't normally used, because it is
320	<	* simpler to rely on the expression {@code n - (n >>> 2)} for the
321	<	* associated resizing threshold.
319	>	* actual floating point value isn't normally used -- it is
320	>	* simpler to use expressions such as {@code n - (n >>> 2)} for
321	>	* the associated resizing threshold.
322		*/
323		private static final float LOAD_FACTOR = 0.75f;
324
325		/**
326	<	* The count value to offset thresholds to compensate for checking
327	<	* for the need to resize only when inserting into bins with two
328	<	* or more elements. See above for explanation.
326	>	* The buffer size for skipped bins during transfers. The
327	>	* value is arbitrary but should be large enough to avoid
328	>	* most locking stalls during resizes.
329	>	*/
330	>	private static final int TRANSFER_BUFFER_SIZE = 32;
331	>
332	>	/*
333	>	* Encodings for special uses of Node hash fields. See above for
334	>	* explanation.
335		*/
336	<	private static final int THRESHOLD_OFFSET = 8;
336	>	static final int MOVED     = 0x80000000; // hash field for fowarding nodes
337	>	static final int LOCKED    = 0x40000000; // set/tested only as a bit
338	>	static final int WAITING   = 0xc0000000; // both bits set/tested together
339	>	static final int HASH_BITS = 0x3fffffff; // usable bits of normal node hash
340	>
341	>	/* ---------------- Fields -------------- */
342
343		/**
344	<	* The default concurrency level for this table. Unused except as
345	<	* a sizing hint, but defined for compatibility with previous
282	<	* versions of this class.
344	>	* The array of bins. Lazily initialized upon first insertion.
345	>	* Size is always a power of two. Accessed directly by iterators.
346		*/
347	<	private static final int DEFAULT_CONCURRENCY_LEVEL = 16;
347	>	transient volatile Node[] table;
348	>
349	>	/**
350	>	* The counter maintaining number of elements.
351	>	*/
352	>	private transient final LongAdder counter;
353	>
354	>	/**
355	>	* Table initialization and resizing control.  When negative, the
356	>	* table is being initialized or resized. Otherwise, when table is
357	>	* null, holds the initial table size to use upon creation, or 0
358	>	* for default. After initialization, holds the next element count
359	>	* value upon which to resize the table.
360	>	*/
361	>	private transient volatile int sizeCtl;
362	>
363	>	// views
364	>	private transient KeySet<K,V> keySet;
365	>	private transient Values<K,V> values;
366	>	private transient EntrySet<K,V> entrySet;
367	>
368	>	/** For serialization compatibility. Null unless serialized; see below */
369	>	private Segment<K,V>[] segments;
370
371		/* ---------------- Nodes -------------- */
372
373		/**
374		* Key-value entry. Note that this is never exported out as a
375	<	* user-visible Map.Entry. Nodes with a negative hash field are
376	<	* special, and do not contain user keys or values.  Otherwise,
377	<	* keys are never null, and null val fields indicate that a node
378	<	* is in the process of being deleted or created. For purposes of
379	<	* read-only access, a key may be read before a val, but can only
380	<	* be used after checking val.  (For an update operation, when a
381	<	* lock is held on a node, order doesn't matter.)
375	>	* user-visible Map.Entry (see WriteThroughEntry and SnapshotEntry
376	>	* below). Nodes with a negative hash field are special, and do
377	>	* not contain user keys or values.  Otherwise, keys are never
378	>	* null, and null val fields indicate that a node is in the
379	>	* process of being deleted or created. For purposes of read-only
380	>	* access, a key may be read before a val, but can only be used
381	>	* after checking val to be non-null.
382		*/
383		static final class Node {
384	<	final int hash;
384	>	volatile int hash;
385		final Object key;
386		volatile Object val;
387		volatile Node next;
#	Line 307 | Line 392 | public class ConcurrentHashMapV8<K, V>
392		this.val = val;
393		this.next = next;
394		}
310	–	}
311	–
312	–	/**
313	–	* Sign bit of node hash value indicating to use table in node.key.
314	–	*/
315	–	private static final int SIGN_BIT = 0x80000000;
395
396	<	/* ---------------- Fields -------------- */
396	>	/** CompareAndSet the hash field */
397	>	final boolean casHash(int cmp, int val) {
398	>	return UNSAFE.compareAndSwapInt(this, hashOffset, cmp, val);
399	>	}
400
401	<	/**
402	<	* The array of bins. Lazily initialized upon first insertion.
403	<	* Size is always a power of two. Accessed directly by iterators.
322	<	*/
323	<	transient volatile Node[] table;
401	>	/** The number of spins before blocking for a lock */
402	>	static final int MAX_SPINS =
403	>	Runtime.getRuntime().availableProcessors() > 1 ? 64 : 1;
404
405	<	/** The counter maintaining number of elements. */
406	<	private transient final LongAdder counter;
407	<	/** Nonzero when table is being initialized or resized. Updated via CAS. */
408	<	private transient volatile int resizing;
409	<	/** The next element count value upon which to resize the table. */
410	<	private transient int threshold;
411	<	/** The target capacity; volatile to cover initialization races. */
412	<	private transient volatile int targetCapacity;
405	>	/**
406	>	* Spins a while if LOCKED bit set and this node is the first
407	>	* of its bin, and then sets WAITING bits on hash field and
408	>	* blocks (once) if they are still set.  It is OK for this
409	>	* method to return even if lock is not available upon exit,
410	>	* which enables these simple single-wait mechanics.
411	>	*
412	>	* The corresponding signalling operation is performed within
413	>	* callers: Upon detecting that WAITING has been set when
414	>	* unlocking lock (via a failed CAS from non-waiting LOCKED
415	>	* state), unlockers acquire the sync lock and perform a
416	>	* notifyAll.
417	>	*/
418	>	final void tryAwaitLock(Node[] tab, int i) {
419	>	if (tab != null && i >= 0 && i < tab.length) { // bounds check
420	>	int spins = MAX_SPINS, h;
421	>	while (tabAt(tab, i) == this && ((h = hash) & LOCKED) != 0) {
422	>	if (spins >= 0) {
423	>	if (--spins == MAX_SPINS >>> 1)
424	>	Thread.yield();  // heuristically yield mid-way
425	>	}
426	>	else if (casHash(h, h | WAITING)) {
427	>	synchronized(this) {
428	>	if (tabAt(tab, i) == this &&
429	>	(hash & WAITING) == WAITING) {
430	>	try {
431	>	wait();
432	>	} catch (InterruptedException ie) {
433	>	Thread.currentThread().interrupt();
434	>	}
435	>	}
436	>	else
437	>	notifyAll(); // possibly won race vs signaller
438	>	}
439	>	break;
440	>	}
441	>	}
442	>	}
443	>	}
444
445	<	// views
446	<	private transient KeySet<K,V> keySet;
447	<	private transient Values<K,V> values;
337	<	private transient EntrySet<K,V> entrySet;
445	>	// Unsafe mechanics for casHash
446	>	private static final sun.misc.Unsafe UNSAFE;
447	>	private static final long hashOffset;
448
449	<	/** For serialization compatibility. Null unless serialized; see below */
450	<	private Segment<K,V>[] segments;
449	>	static {
450	>	try {
451	>	UNSAFE = getUnsafe();
452	>	Class<?> k = Node.class;
453	>	hashOffset = UNSAFE.objectFieldOffset
454	>	(k.getDeclaredField("hash"));
455	>	} catch (Exception e) {
456	>	throw new Error(e);
457	>	}
458	>	}
459	>	}
460
461		/* ---------------- Table element access -------------- */
462
#	Line 365 | Line 484 | public class ConcurrentHashMapV8<K, V>
484		UNSAFE.putObjectVolatile(tab, ((long)i<<ASHIFT)+ABASE, v);
485		}
486
368	–	/* ----------------Table Initialization and Resizing -------------- */
369	–
370	–	/**
371	–	* Returns a power of two table size for the given desired capacity.
372	–	* See Hackers Delight, sec 3.2
373	–	*/
374	–	private static final int tableSizeFor(int c) {
375	–	int n = c - 1;
376	–	n |= n >>> 1;
377	–	n |= n >>> 2;
378	–	n |= n >>> 4;
379	–	n |= n >>> 8;
380	–	n |= n >>> 16;
381	–	return (n < 0) ? 1 : (n >= MAXIMUM_CAPACITY) ? MAXIMUM_CAPACITY : n + 1;
382	–	}
383	–
384	–	/**
385	–	* If not already resizing, initializes or creates next table and
386	–	* transfers bins. Initial table size uses the capacity recorded
387	–	* in targetCapacity.  Rechecks occupancy after a transfer to see
388	–	* if another resize is already needed because resizings are
389	–	* lagging additions.
390	–	*
391	–	* @return current table
392	–	*/
393	–	private final Node[] growTable() {
394	–	if (resizing == 0 &&
395	–	UNSAFE.compareAndSwapInt(this, resizingOffset, 0, 1)) {
396	–	try {
397	–	for (;;) {
398	–	Node[] tab = table;
399	–	int n, c, m;
400	–	if (tab == null)
401	–	n = (c = targetCapacity) > 0 ? c : DEFAULT_CAPACITY;
402	–	else if ((m = tab.length) < MAXIMUM_CAPACITY &&
403	–	counter.sum() >= (long)threshold)
404	–	n = m << 1;
405	–	else
406	–	break;
407	–	threshold = n - (n >>> 2) - THRESHOLD_OFFSET;
408	–	Node[] nextTab = new Node[n];
409	–	if (tab != null)
410	–	transfer(tab, nextTab,
411	–	new Node(SIGN_BIT, nextTab, null, null));
412	–	table = nextTab;
413	–	if (tab == null)
414	–	break;
415	–	}
416	–	} finally {
417	–	resizing = 0;
418	–	}
419	–	}
420	–	else if (table == null)
421	–	Thread.yield(); // lost initialization race; just spin
422	–	return table;
423	–	}
424	–
425	–	/**
426	–	* Reclassifies nodes in each bin to new table.  Because we are
427	–	* using power-of-two expansion, the elements from each bin must
428	–	* either stay at same index, or move with a power of two
429	–	* offset. We eliminate unnecessary node creation by catching
430	–	* cases where old nodes can be reused because their next fields
431	–	* won't change.  Statistically, only about one-sixth of them need
432	–	* cloning when a table doubles. The nodes they replace will be
433	–	* garbage collectable as soon as they are no longer referenced by
434	–	* any reader thread that may be in the midst of concurrently
435	–	* traversing table.
436	–	*
437	–	* Transfers are done from the bottom up to preserve iterator
438	–	* traversability. On each step, the old bin is locked,
439	–	* moved/copied, and then replaced with a forwarding node.
440	–	*/
441	–	private static final void transfer(Node[] tab, Node[] nextTab, Node fwd) {
442	–	int n = tab.length;
443	–	Node ignore = nextTab[n + n - 1]; // force bounds check
444	–	for (int i = n - 1; i >= 0; --i) {
445	–	for (Node e;;) {
446	–	if ((e = tabAt(tab, i)) != null) {
447	–	boolean validated = false;
448	–	synchronized (e) {
449	–	if (tabAt(tab, i) == e) {
450	–	validated = true;
451	–	Node lo = null, hi = null, lastRun = e;
452	–	int runBit = e.hash & n;
453	–	for (Node p = e.next; p != null; p = p.next) {
454	–	int b = p.hash & n;
455	–	if (b != runBit) {
456	–	runBit = b;
457	–	lastRun = p;
458	–	}
459	–	}
460	–	if (runBit == 0)
461	–	lo = lastRun;
462	–	else
463	–	hi = lastRun;
464	–	for (Node p = e; p != lastRun; p = p.next) {
465	–	int ph = p.hash;
466	–	Object pk = p.key, pv = p.val;
467	–	if ((ph & n) == 0)
468	–	lo = new Node(ph, pk, pv, lo);
469	–	else
470	–	hi = new Node(ph, pk, pv, hi);
471	–	}
472	–	setTabAt(nextTab, i, lo);
473	–	setTabAt(nextTab, i + n, hi);
474	–	setTabAt(tab, i, fwd);
475	–	}
476	–	}
477	–	if (validated)
478	–	break;
479	–	}
480	–	else if (casTabAt(tab, i, e, fwd))
481	–	break;
482	–	}
483	–	}
484	–	}
485	–
487		/* ---------------- Internal access and update methods -------------- */
488
489		/**
490		* Applies a supplemental hash function to a given hashCode, which
491		* defends against poor quality hash functions.  The result must
492	<	* be non-negative, and for reasonable performance must have good
493	<	* avalanche properties; i.e., that each bit of the argument
494	<	* affects each bit (except sign bit) of the result.
492	>	* be have the top 2 bits clear. For reasonable performance, this
493	>	* function must have good avalanche properties; i.e., that each
494	>	* bit of the argument affects each bit of the result. (Although
495	>	* we don't care about the unused top 2 bits.)
496		*/
497		private static final int spread(int h) {
498		// Apply base step of MurmurHash; see http://code.google.com/p/smhasher/
#	Line 498 | Line 500 | public class ConcurrentHashMapV8<K, V>
500		h *= 0x85ebca6b;
501		h ^= h >>> 13;
502		h *= 0xc2b2ae35;
503	<	return (h >>> 16) ^ (h & 0x7fffffff); // mask out sign bit
503	>	return ((h >>> 16) ^ h) & HASH_BITS; // mask out top bits
504		}
505
506		/** Implementation for get and containsKey */
507		private final Object internalGet(Object k) {
508		int h = spread(k.hashCode());
509		retry: for (Node[] tab = table; tab != null;) {
510	<	Node e; Object ek, ev; int eh;  // locals to read fields once
510	>	Node e; Object ek, ev; int eh;    // locals to read fields once
511		for (e = tabAt(tab, (tab.length - 1) & h); e != null; e = e.next) {
512	<	if ((eh = e.hash) == h) {
513	<	if ((ev = e.val) != null &&
512	<	((ek = e.key) == k || k.equals(ek)))
513	<	return ev;
514	<	}
515	<	else if (eh < 0) {          // sign bit set
516	<	tab = (Node[])e.key;    // bin was moved during resize
512	>	if ((eh = e.hash) == MOVED) {
513	>	tab = (Node[])e.key;      // restart with new table
514		continue retry;
515		}
516	+	if ((eh & HASH_BITS) == h && (ev = e.val) != null &&
517	+	((ek = e.key) == k || k.equals(ek)))
518	+	return ev;
519		}
520		break;
521		}
#	Line 527 | Line 527 | public class ConcurrentHashMapV8<K, V>
527		int h = spread(k.hashCode());
528		Object oldVal = null;               // previous value or null if none
529		for (Node[] tab = table;;) {
530	<	Node e; int i; Object ek, ev;
530	>	int i; Node f; int fh; Object fk, fv;
531		if (tab == null)
532	<	tab = growTable();
533	<	else if ((e = tabAt(tab, i = (tab.length - 1) & h)) == null) {
532	>	tab = initTable();
533	>	else if ((f = tabAt(tab, i = (tab.length - 1) & h)) == null) {
534		if (casTabAt(tab, i, null, new Node(h, k, v, null)))
535		break;                   // no lock when adding to empty bin
536		}
537	<	else if (e.hash < 0)             // resized -- restart with new table
538	<	tab = (Node[])e.key;
539	<	else if (!replace && e.hash == h && (ev = e.val) != null &&
540	<	((ek = e.key) == k || k.equals(ek))) {
541	<	if (tabAt(tab, i) == e) {    // inspect and validate 1st node
542	<	oldVal = ev;             // without lock for putIfAbsent
543	<	break;
544	<	}
537	>	else if ((fh = f.hash) == MOVED)
538	>	tab = (Node[])f.key;
539	>	else if (!replace && (fh & HASH_BITS) == h && (fv = f.val) != null &&
540	>	((fk = f.key) == k || k.equals(fk))) {
541	>	oldVal = fv;                // precheck 1st node for putIfAbsent
542	>	break;
543		}
544	<	else {
544	>	else if ((fh & LOCKED) != 0)
545	>	f.tryAwaitLock(tab, i);
546	>	else if (f.casHash(fh, fh | LOCKED)) {
547		boolean validated = false;
548		boolean checkSize = false;
549	<	synchronized (e) {           // lock the 1st node of bin list
550	<	if (tabAt(tab, i) == e) {
549	>	try {
550	>	if (tabAt(tab, i) == f) {
551		validated = true;    // retry if 1st already deleted
552	<	for (Node first = e;;) {
553	<	if (e.hash == h &&
554	<	((ek = e.key) == k || k.equals(ek)) &&
555	<	(ev = e.val) != null) {
552	>	for (Node e = f;;) {
553	>	Object ek, ev;
554	>	if ((e.hash & HASH_BITS) == h &&
555	>	(ev = e.val) != null &&
556	>	((ek = e.key) == k || k.equals(ek))) {
557		oldVal = ev;
558		if (replace)
559		e.val = v;
#	Line 561 | Line 562 | public class ConcurrentHashMapV8<K, V>
562		Node last = e;
563		if ((e = e.next) == null) {
564		last.next = new Node(h, k, v, null);
565	<	if (last != first || tab.length <= 64)
565	>	if (last != f || tab.length <= 64)
566		checkSize = true;
567		break;
568		}
569		}
570		}
571	+	} finally {                  // unlock and signal if needed
572	+	if (!f.casHash(fh | LOCKED, fh)) {
573	+	f.hash = fh;
574	+	synchronized(f) { f.notifyAll(); };
575	+	}
576		}
577		if (validated) {
578	+	int sc;
579		if (checkSize && tab.length < MAXIMUM_CAPACITY &&
580	<	resizing == 0 && counter.sum() >= (long)threshold)
580	>	(sc = sizeCtl) >= 0 && counter.sum() >= (long)sc)
581		growTable();
582		break;
583		}
#	Line 588 | Line 595 | public class ConcurrentHashMapV8<K, V>
595		*/
596		private final Object internalReplace(Object k, Object v, Object cv) {
597		int h = spread(k.hashCode());
598	+	Object oldVal = null;
599		for (Node[] tab = table;;) {
600	<	Node e; int i;
600	>	Node f; int i, fh;
601		if (tab == null ||
602	<	(e = tabAt(tab, i = (tab.length - 1) & h)) == null)
603	<	return null;
604	<	else if (e.hash < 0)
605	<	tab = (Node[])e.key;
606	<	else {
607	<	Object oldVal = null;
602	>	(f = tabAt(tab, i = (tab.length - 1) & h)) == null)
603	>	break;
604	>	else if ((fh = f.hash) == MOVED)
605	>	tab = (Node[])f.key;
606	>	else if ((fh & HASH_BITS) != h && f.next == null) // precheck
607	>	break;                          // rules out possible existence
608	>	else if ((fh & LOCKED) != 0)
609	>	f.tryAwaitLock(tab, i);
610	>	else if (f.casHash(fh, fh | LOCKED)) {
611		boolean validated = false;
612		boolean deleted = false;
613	<	synchronized (e) {
614	<	if (tabAt(tab, i) == e) {
613	>	try {
614	>	if (tabAt(tab, i) == f) {
615		validated = true;
616	<	Node pred = null;
606	<	do {
616	>	for (Node e = f, pred = null;;) {
617		Object ek, ev;
618	<	if (e.hash == h &&
619	<	((ek = e.key) == k || k.equals(ek)) &&
620	<	((ev = e.val) != null)) {
618	>	if ((e.hash & HASH_BITS) == h &&
619	>	((ev = e.val) != null) &&
620	>	((ek = e.key) == k || k.equals(ek))) {
621		if (cv == null || cv == ev || cv.equals(ev)) {
622		oldVal = ev;
623		if ((e.val = v) == null) {
#	Line 621 | Line 631 | public class ConcurrentHashMapV8<K, V>
631		}
632		break;
633		}
634	<	} while ((e = (pred = e).next) != null);
634	>	pred = e;
635	>	if ((e = e.next) == null)
636	>	break;
637	>	}
638	>	}
639	>	} finally {
640	>	if (!f.casHash(fh | LOCKED, fh)) {
641	>	f.hash = fh;
642	>	synchronized(f) { f.notifyAll(); };
643		}
644		}
645		if (validated) {
646		if (deleted)
647		counter.decrement();
648	<	return oldVal;
648	>	break;
649		}
650		}
651		}
652	+	return oldVal;
653		}
654
655		/** Implementation for computeIfAbsent and compute. Like put, but messier. */
656	+	// Todo: Somehow reinstate non-termination check
657		@SuppressWarnings("unchecked")
658		private final V internalCompute(K k,
659	<	MappingFunction<? super K, ? extends V> f,
659	>	MappingFunction<? super K, ? extends V> fn,
660		boolean replace) {
661		int h = spread(k.hashCode());
662		V val = null;
663		boolean added = false;
664		Node[] tab = table;
665		outer:for (;;) {
666	<	Node e; int i; Object ek, ev;
666	>	Node f; int i, fh; Object fk, fv;
667		if (tab == null)
668	<	tab = growTable();
669	<	else if ((e = tabAt(tab, i = (tab.length - 1) & h)) == null) {
670	<	Node node = new Node(h, k, null, null);
668	>	tab = initTable();
669	>	else if ((f = tabAt(tab, i = (tab.length - 1) & h)) == null) {
670	>	Node node = new Node(fh = h | LOCKED, k, null, null);
671		boolean validated = false;
672	<	synchronized (node) {  // must lock while computing value
673	<	if (casTabAt(tab, i, null, node)) {
674	<	validated = true;
675	<	try {
676	<	val = f.map(k);
677	<	if (val != null) {
678	<	node.val = val;
679	<	added = true;
680	<	}
681	<	} finally {
682	<	if (!added)
683	<	setTabAt(tab, i, null);
672	>	if (casTabAt(tab, i, null, node)) {
673	>	validated = true;
674	>	try {
675	>	val = fn.map(k);
676	>	if (val != null) {
677	>	node.val = val;
678	>	added = true;
679	>	}
680	>	} finally {
681	>	if (!added)
682	>	setTabAt(tab, i, null);
683	>	if (!node.casHash(fh, h)) {
684	>	node.hash = fh;
685	>	synchronized(node) { node.notifyAll(); };
686		}
687		}
688		}
689		if (validated)
690		break;
691		}
692	<	else if (e.hash < 0)
693	<	tab = (Node[])e.key;
694	<	else if (!replace && e.hash == h && (ev = e.val) != null &&
695	<	((ek = e.key) == k || k.equals(ek))) {
696	<	if (tabAt(tab, i) == e) {
697	<	val = (V)ev;
692	>	else if ((fh = f.hash) == MOVED)
693	>	tab = (Node[])f.key;
694	>	else if (!replace && (fh & HASH_BITS) == h && (fv = f.val) != null &&
695	>	((fk = f.key) == k || k.equals(fk))) {
696	>	if (tabAt(tab, i) == f) {
697	>	val = (V)fv;
698		break;
699		}
700		}
701	<	else if (Thread.holdsLock(e))
702	<	throw new IllegalStateException("Recursive map computation");
703	<	else {
701	>	else if ((fh & LOCKED) != 0)
702	>	f.tryAwaitLock(tab, i);
703	>	else if (f.casHash(fh, fh | LOCKED)) {
704		boolean validated = false;
705		boolean checkSize = false;
706	<	synchronized (e) {
707	<	if (tabAt(tab, i) == e) {
706	>	try {
707	>	if (tabAt(tab, i) == f) {
708		validated = true;
709	<	for (Node first = e;;) {
710	<	if (e.hash == h &&
711	<	((ek = e.key) == k || k.equals(ek)) &&
712	<	((ev = e.val) != null)) {
713	<	Object fv;
714	<	if (replace && (fv = f.map(k)) != null)
715	<	ev = e.val = fv;
709	>	for (Node e = f;;) {
710	>	Object ek, ev, v;
711	>	if ((e.hash & HASH_BITS) == h &&
712	>	(ev = e.val) != null &&
713	>	((ek = e.key) == k || k.equals(ek))) {
714	>	if (replace && (v = fn.map(k)) != null)
715	>	ev = e.val = v;
716		val = (V)ev;
717		break;
718		}
719		Node last = e;
720		if ((e = e.next) == null) {
721	<	if ((val = f.map(k)) != null) {
721	>	if ((val = fn.map(k)) != null) {
722		last.next = new Node(h, k, val, null);
723		added = true;
724	<	if (last != first || tab.length <= 64)
724	>	if (last != f || tab.length <= 64)
725		checkSize = true;
726		}
727		break;
728		}
729		}
730		}
731	+	} finally {
732	+	if (!f.casHash(fh | LOCKED, fh)) {
733	+	f.hash = fh;
734	+	synchronized(f) { f.notifyAll(); };
735	+	}
736		}
737		if (validated) {
738	+	int sc;
739		if (checkSize && tab.length < MAXIMUM_CAPACITY &&
740	<	resizing == 0 && counter.sum() >= (long)threshold)
740	>	(sc = sizeCtl) >= 0 && counter.sum() >= (long)sc)
741		growTable();
742		break;
743		}
#	Line 728 | Line 756 | public class ConcurrentHashMapV8<K, V>
756		int i = 0;
757		Node[] tab = table;
758		while (tab != null && i < tab.length) {
759	<	Node e = tabAt(tab, i);
760	<	if (e == null)
759	>	int fh;
760	>	Node f = tabAt(tab, i);
761	>	if (f == null)
762		++i;
763	<	else if (e.hash < 0)
764	<	tab = (Node[])e.key;
765	<	else {
763	>	else if ((fh = f.hash) == MOVED)
764	>	tab = (Node[])f.key;
765	>	else if ((fh & LOCKED) != 0)
766	>	f.tryAwaitLock(tab, i);
767	>	else if (f.casHash(fh, fh | LOCKED)) {
768		boolean validated = false;
769	<	synchronized (e) {
770	<	if (tabAt(tab, i) == e) {
769	>	try {
770	>	if (tabAt(tab, i) == f) {
771		validated = true;
772	<	Node en;
742	<	do {
743	<	en = e.next;
772	>	for (Node e = f; e != null; e = e.next) {
773		if (e.val != null) { // currently always true
774		e.val = null;
775		--delta;
776		}
777	<	} while ((e = en) != null);
777	>	}
778		setTabAt(tab, i, null);
779		}
780	+	} finally {
781	+	if (!f.casHash(fh | LOCKED, fh)) {
782	+	f.hash = fh;
783	+	synchronized(f) { f.notifyAll(); };
784	+	}
785		}
786		if (validated)
787		++i;
#	Line 756 | Line 790 | public class ConcurrentHashMapV8<K, V>
790		counter.add(delta);
791		}
792
793	+	/* ----------------Table Initialization and Resizing -------------- */
794	+
795	+	/**
796	+	* Returns a power of two table size for the given desired capacity.
797	+	* See Hackers Delight, sec 3.2
798	+	*/
799	+	private static final int tableSizeFor(int c) {
800	+	int n = c - 1;
801	+	n |= n >>> 1;
802	+	n |= n >>> 2;
803	+	n |= n >>> 4;
804	+	n |= n >>> 8;
805	+	n |= n >>> 16;
806	+	return (n < 0) ? 1 : (n >= MAXIMUM_CAPACITY) ? MAXIMUM_CAPACITY : n + 1;
807	+	}
808	+
809	+	/**
810	+	* Initializes table, using the size recorded in sizeCtl.
811	+	*/
812	+	private final Node[] initTable() {
813	+	Node[] tab; int sc;
814	+	while ((tab = table) == null) {
815	+	if ((sc = sizeCtl) < 0)
816	+	Thread.yield(); // lost initialization race; just spin
817	+	else if (UNSAFE.compareAndSwapInt(this, sizeCtlOffset, sc, -1)) {
818	+	try {
819	+	if ((tab = table) == null) {
820	+	int n = (sc > 0) ? sc : DEFAULT_CAPACITY;
821	+	tab = table = new Node[n];
822	+	sc = n - (n >>> 2) - 1;
823	+	}
824	+	} finally {
825	+	sizeCtl = sc;
826	+	}
827	+	break;
828	+	}
829	+	}
830	+	return tab;
831	+	}
832	+
833	+	/**
834	+	* If not already resizing, creates next table and transfers bins.
835	+	* Rechecks occupancy after a transfer to see if another resize is
836	+	* already needed because resizings are lagging additions.
837	+	*/
838	+	private final void growTable() {
839	+	int sc = sizeCtl;
840	+	if (sc >= 0 && UNSAFE.compareAndSwapInt(this, sizeCtlOffset, sc, -1)) {
841	+	try {
842	+	Node[] tab; int n;
843	+	while ((tab = table) != null &&
844	+	(n = tab.length) > 0 && n < MAXIMUM_CAPACITY &&
845	+	counter.sum() >= (long)sc) {
846	+	table = rebuild(tab);
847	+	sc = (n << 1) - (n >>> 1) - 1;
848	+	}
849	+	} finally {
850	+	sizeCtl = sc;
851	+	}
852	+	}
853	+	}
854	+
855	+	/*
856	+	* Moves and/or copies the nodes in each bin to new table. See
857	+	* above for explanation.
858	+	*
859	+	* @return the new table
860	+	*/
861	+	private static final Node[] rebuild(Node[] tab) {
862	+	int n = tab.length;
863	+	Node[] nextTab = new Node[n << 1];
864	+	Node fwd = new Node(MOVED, nextTab, null, null);
865	+	int[] buffer = null;       // holds bins to revisit; null until needed
866	+	Node rev = null;           // reverse forwarder; null until needed
867	+	int nbuffered = 0;         // the number of bins in buffer list
868	+	int bufferIndex = 0;       // buffer index of current buffered bin
869	+	int bin = n - 1;           // current non-buffered bin or -1 if none
870	+
871	+	for (int i = bin;;) {      // start upwards sweep
872	+	int fh; Node f;
873	+	if ((f = tabAt(tab, i)) == null) {
874	+	if (bin >= 0) {    // no lock needed (or available)
875	+	if (!casTabAt(tab, i, f, fwd))
876	+	continue;
877	+	}
878	+	else {             // transiently use a locked forwarding node
879	+	Node g =  new Node(MOVED|LOCKED, nextTab, null, null);
880	+	if (!casTabAt(tab, i, f, g))
881	+	continue;
882	+	setTabAt(nextTab, i, null);
883	+	setTabAt(nextTab, i + n, null);
884	+	setTabAt(tab, i, fwd);
885	+	if (!g.casHash(MOVED|LOCKED, MOVED)) {
886	+	g.hash = MOVED;
887	+	synchronized(g) { g.notifyAll(); }
888	+	}
889	+	}
890	+	}
891	+	else if (((fh = f.hash) & LOCKED) == 0 && f.casHash(fh, fh|LOCKED)) {
892	+	boolean validated = false;
893	+	try {              // split to lo and hi lists; copying as needed
894	+	if (tabAt(tab, i) == f) {
895	+	validated = true;
896	+	Node e = f, lastRun = f;
897	+	Node lo = null, hi = null;
898	+	int runBit = e.hash & n;
899	+	for (Node p = e.next; p != null; p = p.next) {
900	+	int b = p.hash & n;
901	+	if (b != runBit) {
902	+	runBit = b;
903	+	lastRun = p;
904	+	}
905	+	}
906	+	if (runBit == 0)
907	+	lo = lastRun;
908	+	else
909	+	hi = lastRun;
910	+	for (Node p = e; p != lastRun; p = p.next) {
911	+	int ph = p.hash & HASH_BITS;
912	+	Object pk = p.key, pv = p.val;
913	+	if ((ph & n) == 0)
914	+	lo = new Node(ph, pk, pv, lo);
915	+	else
916	+	hi = new Node(ph, pk, pv, hi);
917	+	}
918	+	setTabAt(nextTab, i, lo);
919	+	setTabAt(nextTab, i + n, hi);
920	+	setTabAt(tab, i, fwd);
921	+	}
922	+	} finally {
923	+	if (!f.casHash(fh | LOCKED, fh)) {
924	+	f.hash = fh;
925	+	synchronized(f) { f.notifyAll(); };
926	+	}
927	+	}
928	+	if (!validated)
929	+	continue;
930	+	}
931	+	else {
932	+	if (buffer == null) // initialize buffer for revisits
933	+	buffer = new int[TRANSFER_BUFFER_SIZE];
934	+	if (bin < 0 && bufferIndex > 0) {
935	+	int j = buffer[--bufferIndex];
936	+	buffer[bufferIndex] = i;
937	+	i = j;         // swap with another bin
938	+	continue;
939	+	}
940	+	if (bin < 0 || nbuffered >= TRANSFER_BUFFER_SIZE) {
941	+	f.tryAwaitLock(tab, i);
942	+	continue;      // no other options -- block
943	+	}
944	+	if (rev == null)   // initialize reverse-forwarder
945	+	rev = new Node(MOVED, tab, null, null);
946	+	if (tabAt(tab, i) != f || (f.hash & LOCKED) == 0)
947	+	continue;      // recheck before adding to list
948	+	buffer[nbuffered++] = i;
949	+	setTabAt(nextTab, i, rev);     // install place-holders
950	+	setTabAt(nextTab, i + n, rev);
951	+	}
952	+
953	+	if (bin > 0)
954	+	i = --bin;
955	+	else if (buffer != null && nbuffered > 0) {
956	+	bin = -1;
957	+	i = buffer[bufferIndex = --nbuffered];
958	+	}
959	+	else
960	+	return nextTab;
961	+	}
962	+	}
963	+
964		/* ----------------Table Traversal -------------- */
965
966		/**
#	Line 830 | Line 1035 | public class ConcurrentHashMapV8<K, V>
1035		final void advance() {
1036		Node e = last = next;
1037		outer: do {
1038	<	if (e != null)                   // pass used or skipped node
1038	>	if (e != null)                  // advance past used/skipped node
1039		e = e.next;
1040	<	while (e == null) {              // get to next non-null bin
1041	<	Node[] t; int b, i, n;       // checks must use locals
1040	>	while (e == null) {             // get to next non-null bin
1041	>	Node[] t; int b, i, n;      // checks must use locals
1042		if ((b = baseIndex) >= baseLimit || (i = index) < 0 ||
1043		(t = tab) == null || i >= (n = t.length))
1044		break outer;
1045	<	else if ((e = tabAt(t, i)) != null && e.hash < 0)
1046	<	tab = (Node[])e.key;     // restarts due to null val
1047	<	else                         // visit upper slots if present
1045	>	else if ((e = tabAt(t, i)) != null && e.hash == MOVED)
1046	>	tab = (Node[])e.key;    // restarts due to null val
1047	>	else                        // visit upper slots if present
1048		index = (i += baseSize) < n ? i : (baseIndex = b + 1);
1049		}
1050		nextKey = e.key;
1051	<	} while ((nextVal = e.val) == null); // skip deleted or special nodes
1051	>	} while ((nextVal = e.val) == null);// skip deleted or special nodes
1052		next = e;
1053		}
1054		}
#	Line 855 | Line 1060 | public class ConcurrentHashMapV8<K, V>
1060		*/
1061		public ConcurrentHashMapV8() {
1062		this.counter = new LongAdder();
858	–	this.targetCapacity = DEFAULT_CAPACITY;
1063		}
1064
1065		/**
#	Line 875 | Line 1079 | public class ConcurrentHashMapV8<K, V>
1079		MAXIMUM_CAPACITY :
1080		tableSizeFor(initialCapacity + (initialCapacity >>> 1) + 1));
1081		this.counter = new LongAdder();
1082	<	this.targetCapacity = cap;
1082	>	this.sizeCtl = cap;
1083		}
1084
1085		/**
#	Line 885 | Line 1089 | public class ConcurrentHashMapV8<K, V>
1089		*/
1090		public ConcurrentHashMapV8(Map<? extends K, ? extends V> m) {
1091		this.counter = new LongAdder();
1092	<	this.targetCapacity = DEFAULT_CAPACITY;
1092	>	this.sizeCtl = DEFAULT_CAPACITY;
1093		putAll(m);
1094		}
1095
#	Line 936 | Line 1140 | public class ConcurrentHashMapV8<K, V>
1140		int cap =  ((size >= (long)MAXIMUM_CAPACITY) ?
1141		MAXIMUM_CAPACITY: tableSizeFor((int)size));
1142		this.counter = new LongAdder();
1143	<	this.targetCapacity = cap;
1143	>	this.sizeCtl = cap;
1144		}
1145
1146		/**
#	Line 956 | Line 1160 | public class ConcurrentHashMapV8<K, V>
1160		(int)n);
1161		}
1162
1163	+	final long longSize() { // accurate version of size needed for views
1164	+	long n = counter.sum();
1165	+	return (n < 0L) ? 0L : n;
1166	+	}
1167	+
1168		/**
1169		* Returns the value to which the specified key is mapped,
1170		* or {@code null} if this map contains no mapping for the key.
#	Line 1076 | Line 1285 | public class ConcurrentHashMapV8<K, V>
1285		if (m == null)
1286		throw new NullPointerException();
1287		/*
1288	<	* If uninitialized, try to adjust targetCapacity to
1080	<	* accommodate the given number of elements.
1288	>	* If uninitialized, try to preallocate big enough table
1289		*/
1290		if (table == null) {
1291		int size = m.size();
1292	<	int cap = (size >= (MAXIMUM_CAPACITY >>> 1)) ? MAXIMUM_CAPACITY :
1292	>	int n = (size >= (MAXIMUM_CAPACITY >>> 1)) ? MAXIMUM_CAPACITY :
1293		tableSizeFor(size + (size >>> 1) + 1);
1294	<	if (cap > targetCapacity)
1295	<	targetCapacity = cap;
1294	>	int sc = sizeCtl;
1295	>	if (n < sc)
1296	>	n = sc;
1297	>	if (sc >= 0 &&
1298	>	UNSAFE.compareAndSwapInt(this, sizeCtlOffset, sc, -1)) {
1299	>	try {
1300	>	if (table == null) {
1301	>	table = new Node[n];
1302	>	sc = n - (n >>> 2) - 1;
1303	>	}
1304	>	} finally {
1305	>	sizeCtl = sc;
1306	>	}
1307	>	}
1308	>	}
1309	>	for (Map.Entry<? extends K, ? extends V> e : m.entrySet()) {
1310	>	Object ek = e.getKey(), ev = e.getValue();
1311	>	if (ek == null || ev == null)
1312	>	throw new NullPointerException();
1313	>	internalPut(ek, ev, true);
1314		}
1089	–	for (Map.Entry<? extends K, ? extends V> e : m.entrySet())
1090	–	put(e.getKey(), e.getValue());
1315		}
1316
1317		/**
#	Line 1462 | Line 1686 | public class ConcurrentHashMapV8<K, V>
1686		Object k = nextKey;
1687		Object v = nextVal;
1688		advance();
1689	<	return new WriteThroughEntry<K,V>(map, (K)k, (V)v);
1689	>	return new WriteThroughEntry<K,V>((K)k, (V)v, map);
1690	>	}
1691	>	}
1692	>
1693	>	static final class SnapshotEntryIterator<K,V> extends ViewIterator<K,V>
1694	>	implements Iterator<Map.Entry<K,V>> {
1695	>	SnapshotEntryIterator(ConcurrentHashMapV8<K, V> map) { super(map); }
1696	>
1697	>	@SuppressWarnings("unchecked")
1698	>	public final Map.Entry<K,V> next() {
1699	>	if (next == null)
1700	>	throw new NoSuchElementException();
1701	>	Object k = nextKey;
1702	>	Object v = nextVal;
1703	>	advance();
1704	>	return new SnapshotEntry<K,V>((K)k, (V)v);
1705		}
1706		}
1707
1708		/**
1709	<	* Custom Entry class used by EntryIterator.next(), that relays
1471	<	* setValue changes to the underlying map.
1709	>	* Base of writeThrough and Snapshot entry classes
1710		*/
1711	<	static final class WriteThroughEntry<K,V> implements Map.Entry<K, V> {
1474	<	final ConcurrentHashMapV8<K, V> map;
1711	>	static abstract class MapEntry<K,V> implements Map.Entry<K, V> {
1712		final K key; // non-null
1713		V val;       // non-null
1714	<	WriteThroughEntry(ConcurrentHashMapV8<K, V> map, K key, V val) {
1478	<	this.map = map; this.key = key; this.val = val;
1479	<	}
1480	<
1714	>	MapEntry(K key, V val)        { this.key = key; this.val = val; }
1715		public final K getKey()       { return key; }
1716		public final V getValue()     { return val; }
1717		public final int hashCode()   { return key.hashCode() ^ val.hashCode(); }
#	Line 1492 | Line 1726 | public class ConcurrentHashMapV8<K, V>
1726		(v == val || v.equals(val)));
1727		}
1728
1729	+	public abstract V setValue(V value);
1730	+	}
1731	+
1732	+	/**
1733	+	* Entry used by EntryIterator.next(), that relays setValue
1734	+	* changes to the underlying map.
1735	+	*/
1736	+	static final class WriteThroughEntry<K,V> extends MapEntry<K,V>
1737	+	implements Map.Entry<K, V> {
1738	+	final ConcurrentHashMapV8<K, V> map;
1739	+	WriteThroughEntry(K key, V val, ConcurrentHashMapV8<K, V> map) {
1740	+	super(key, val);
1741	+	this.map = map;
1742	+	}
1743	+
1744		/**
1745		* Sets our entry's value and writes through to the map. The
1746		* value to return is somewhat arbitrary here. Since a
#	Line 1510 | Line 1759 | public class ConcurrentHashMapV8<K, V>
1759		}
1760		}
1761
1762	+	/**
1763	+	* Internal version of entry, that doesn't write though changes
1764	+	*/
1765	+	static final class SnapshotEntry<K,V> extends MapEntry<K,V>
1766	+	implements Map.Entry<K, V> {
1767	+	SnapshotEntry(K key, V val) { super(key, val); }
1768	+	public final V setValue(V value) { // only locally update
1769	+	if (value == null) throw new NullPointerException();
1770	+	V v = val;
1771	+	val = value;
1772	+	return v;
1773	+	}
1774	+	}
1775	+
1776		/* ----------------Views -------------- */
1777
1778	<	/*
1779	<	* These currently just extend java.util.AbstractX classes, but
1780	<	* may need a new custom base to support partitioned traversal.
1778	>	/**
1779	>	* Base class for views. This is done mainly to allow adding
1780	>	* customized parallel traversals (not yet implemented.)
1781		*/
1782	<
1520	<	static final class KeySet<K,V> extends AbstractSet<K> {
1782	>	static abstract class MapView<K, V> {
1783		final ConcurrentHashMapV8<K, V> map;
1784	<	KeySet(ConcurrentHashMapV8<K, V> map)   { this.map = map; }
1523	<
1784	>	MapView(ConcurrentHashMapV8<K, V> map)  { this.map = map; }
1785		public final int size()                 { return map.size(); }
1786		public final boolean isEmpty()          { return map.isEmpty(); }
1787		public final void clear()               { map.clear(); }
1788	+
1789	+	// implementations below rely on concrete classes supplying these
1790	+	abstract Iterator<?> iter();
1791	+	abstract public boolean contains(Object o);
1792	+	abstract public boolean remove(Object o);
1793	+
1794	+	private static final String oomeMsg = "Required array size too large";
1795	+
1796	+	public final Object[] toArray() {
1797	+	long sz = map.longSize();
1798	+	if (sz > (long)(MAX_ARRAY_SIZE))
1799	+	throw new OutOfMemoryError(oomeMsg);
1800	+	int n = (int)sz;
1801	+	Object[] r = new Object[n];
1802	+	int i = 0;
1803	+	Iterator<?> it = iter();
1804	+	while (it.hasNext()) {
1805	+	if (i == n) {
1806	+	if (n >= MAX_ARRAY_SIZE)
1807	+	throw new OutOfMemoryError(oomeMsg);
1808	+	if (n >= MAX_ARRAY_SIZE - (MAX_ARRAY_SIZE >>> 1) - 1)
1809	+	n = MAX_ARRAY_SIZE;
1810	+	else
1811	+	n += (n >>> 1) + 1;
1812	+	r = Arrays.copyOf(r, n);
1813	+	}
1814	+	r[i++] = it.next();
1815	+	}
1816	+	return (i == n) ? r : Arrays.copyOf(r, i);
1817	+	}
1818	+
1819	+	@SuppressWarnings("unchecked")
1820	+	public final <T> T[] toArray(T[] a) {
1821	+	long sz = map.longSize();
1822	+	if (sz > (long)(MAX_ARRAY_SIZE))
1823	+	throw new OutOfMemoryError(oomeMsg);
1824	+	int m = (int)sz;
1825	+	T[] r = (a.length >= m) ? a :
1826	+	(T[])java.lang.reflect.Array
1827	+	.newInstance(a.getClass().getComponentType(), m);
1828	+	int n = r.length;
1829	+	int i = 0;
1830	+	Iterator<?> it = iter();
1831	+	while (it.hasNext()) {
1832	+	if (i == n) {
1833	+	if (n >= MAX_ARRAY_SIZE)
1834	+	throw new OutOfMemoryError(oomeMsg);
1835	+	if (n >= MAX_ARRAY_SIZE - (MAX_ARRAY_SIZE >>> 1) - 1)
1836	+	n = MAX_ARRAY_SIZE;
1837	+	else
1838	+	n += (n >>> 1) + 1;
1839	+	r = Arrays.copyOf(r, n);
1840	+	}
1841	+	r[i++] = (T)it.next();
1842	+	}
1843	+	if (a == r && i < n) {
1844	+	r[i] = null; // null-terminate
1845	+	return r;
1846	+	}
1847	+	return (i == n) ? r : Arrays.copyOf(r, i);
1848	+	}
1849	+
1850	+	public final int hashCode() {
1851	+	int h = 0;
1852	+	for (Iterator<?> it = iter(); it.hasNext();)
1853	+	h += it.next().hashCode();
1854	+	return h;
1855	+	}
1856	+
1857	+	public final String toString() {
1858	+	StringBuilder sb = new StringBuilder();
1859	+	sb.append('[');
1860	+	Iterator<?> it = iter();
1861	+	if (it.hasNext()) {
1862	+	for (;;) {
1863	+	Object e = it.next();
1864	+	sb.append(e == this ? "(this Collection)" : e);
1865	+	if (!it.hasNext())
1866	+	break;
1867	+	sb.append(',').append(' ');
1868	+	}
1869	+	}
1870	+	return sb.append(']').toString();
1871	+	}
1872	+
1873	+	public final boolean containsAll(Collection<?> c) {
1874	+	if (c != this) {
1875	+	for (Iterator<?> it = c.iterator(); it.hasNext();) {
1876	+	Object e = it.next();
1877	+	if (e == null || !contains(e))
1878	+	return false;
1879	+	}
1880	+	}
1881	+	return true;
1882	+	}
1883	+
1884	+	public final boolean removeAll(Collection c) {
1885	+	boolean modified = false;
1886	+	for (Iterator<?> it = iter(); it.hasNext();) {
1887	+	if (c.contains(it.next())) {
1888	+	it.remove();
1889	+	modified = true;
1890	+	}
1891	+	}
1892	+	return modified;
1893	+	}
1894	+
1895	+	public final boolean retainAll(Collection<?> c) {
1896	+	boolean modified = false;
1897	+	for (Iterator<?> it = iter(); it.hasNext();) {
1898	+	if (!c.contains(it.next())) {
1899	+	it.remove();
1900	+	modified = true;
1901	+	}
1902	+	}
1903	+	return modified;
1904	+	}
1905	+
1906	+	}
1907	+
1908	+	static final class KeySet<K,V> extends MapView<K,V> implements Set<K> {
1909	+	KeySet(ConcurrentHashMapV8<K, V> map)   { super(map); }
1910		public final boolean contains(Object o) { return map.containsKey(o); }
1911		public final boolean remove(Object o)   { return map.remove(o) != null; }
1912	+
1913		public final Iterator<K> iterator() {
1914		return new KeyIterator<K,V>(map);
1915		}
1916	+	final Iterator<?> iter() {
1917	+	return new KeyIterator<K,V>(map);
1918	+	}
1919	+	public final boolean add(K e) {
1920	+	throw new UnsupportedOperationException();
1921	+	}
1922	+	public final boolean addAll(Collection<? extends K> c) {
1923	+	throw new UnsupportedOperationException();
1924	+	}
1925	+	public boolean equals(Object o) {
1926	+	Set<?> c;
1927	+	return ((o instanceof Set) &&
1928	+	((c = (Set<?>)o) == this ||
1929	+	(containsAll(c) && c.containsAll(this))));
1930	+	}
1931		}
1932
1933	<	static final class Values<K,V> extends AbstractCollection<V> {
1934	<	final ConcurrentHashMapV8<K, V> map;
1935	<	Values(ConcurrentHashMapV8<K, V> map)   { this.map = map; }
1537	<
1538	<	public final int size()                 { return map.size(); }
1539	<	public final boolean isEmpty()          { return map.isEmpty(); }
1540	<	public final void clear()               { map.clear(); }
1933	>	static final class Values<K,V> extends MapView<K,V>
1934	>	implements Collection<V>  {
1935	>	Values(ConcurrentHashMapV8<K, V> map)   { super(map); }
1936		public final boolean contains(Object o) { return map.containsValue(o); }
1937	+
1938	+	public final boolean remove(Object o) {
1939	+	if (o != null) {
1940	+	Iterator<V> it = new ValueIterator<K,V>(map);
1941	+	while (it.hasNext()) {
1942	+	if (o.equals(it.next())) {
1943	+	it.remove();
1944	+	return true;
1945	+	}
1946	+	}
1947	+	}
1948	+	return false;
1949	+	}
1950		public final Iterator<V> iterator() {
1951		return new ValueIterator<K,V>(map);
1952		}
1953	+	final Iterator<?> iter() {
1954	+	return new ValueIterator<K,V>(map);
1955	+	}
1956	+	public final boolean add(V e) {
1957	+	throw new UnsupportedOperationException();
1958	+	}
1959	+	public final boolean addAll(Collection<? extends V> c) {
1960	+	throw new UnsupportedOperationException();
1961	+	}
1962		}
1963
1964	<	static final class EntrySet<K,V> extends AbstractSet<Map.Entry<K,V>> {
1965	<	final ConcurrentHashMapV8<K, V> map;
1966	<	EntrySet(ConcurrentHashMapV8<K, V> map) { this.map = map; }
1550	<
1551	<	public final int size()                 { return map.size(); }
1552	<	public final boolean isEmpty()          { return map.isEmpty(); }
1553	<	public final void clear()               { map.clear(); }
1554	<	public final Iterator<Map.Entry<K,V>> iterator() {
1555	<	return new EntryIterator<K,V>(map);
1556	<	}
1964	>	static final class EntrySet<K,V>  extends MapView<K,V>
1965	>	implements Set<Map.Entry<K,V>> {
1966	>	EntrySet(ConcurrentHashMapV8<K, V> map) { super(map); }
1967
1968		public final boolean contains(Object o) {
1969		Object k, v, r; Map.Entry<?,?> e;
#	Line 1571 | Line 1981 | public class ConcurrentHashMapV8<K, V>
1981		(v = e.getValue()) != null &&
1982		map.remove(k, v));
1983		}
1984	+
1985	+	public final Iterator<Map.Entry<K,V>> iterator() {
1986	+	return new EntryIterator<K,V>(map);
1987	+	}
1988	+	final Iterator<?> iter() {
1989	+	return new SnapshotEntryIterator<K,V>(map);
1990	+	}
1991	+	public final boolean add(Entry<K,V> e) {
1992	+	throw new UnsupportedOperationException();
1993	+	}
1994	+	public final boolean addAll(Collection<? extends Entry<K,V>> c) {
1995	+	throw new UnsupportedOperationException();
1996	+	}
1997	+	public boolean equals(Object o) {
1998	+	Set<?> c;
1999	+	return ((o instanceof Set) &&
2000	+	((c = (Set<?>)o) == this ||
2001	+	(containsAll(c) && c.containsAll(this))));
2002	+	}
2003		}
2004
2005		/* ---------------- Serialization Support -------------- */
#	Line 1626 | Line 2055 | public class ConcurrentHashMapV8<K, V>
2055		this.segments = null; // unneeded
2056		// initialize transient final field
2057		UNSAFE.putObjectVolatile(this, counterOffset, new LongAdder());
1629	–	this.targetCapacity = DEFAULT_CAPACITY;
2058
2059		// Create all nodes, then place in table once size is known
2060		long size = 0L;
#	Line 1643 | Line 2071 | public class ConcurrentHashMapV8<K, V>
2071		}
2072		if (p != null) {
2073		boolean init = false;
2074	<	if (resizing == 0 &&
2075	<	UNSAFE.compareAndSwapInt(this, resizingOffset, 0, 1)) {
2074	>	int n;
2075	>	if (size >= (long)(MAXIMUM_CAPACITY >>> 1))
2076	>	n = MAXIMUM_CAPACITY;
2077	>	else {
2078	>	int sz = (int)size;
2079	>	n = tableSizeFor(sz + (sz >>> 1) + 1);
2080	>	}
2081	>	int sc = sizeCtl;
2082	>	if (n > sc &&
2083	>	UNSAFE.compareAndSwapInt(this, sizeCtlOffset, sc, -1)) {
2084		try {
2085		if (table == null) {
2086		init = true;
1651	–	int n;
1652	–	if (size >= (long)(MAXIMUM_CAPACITY >>> 1))
1653	–	n = MAXIMUM_CAPACITY;
1654	–	else {
1655	–	int sz = (int)size;
1656	–	n = tableSizeFor(sz + (sz >>> 1) + 1);
1657	–	}
1658	–	threshold = n - (n >>> 2) - THRESHOLD_OFFSET;
2087		Node[] tab = new Node[n];
2088		int mask = n - 1;
2089		while (p != null) {
#	Line 1667 | Line 2095 | public class ConcurrentHashMapV8<K, V>
2095		}
2096		table = tab;
2097		counter.add(size);
2098	+	sc = n - (n >>> 2) - 1;
2099		}
2100		} finally {
2101	<	resizing = 0;
2101	>	sizeCtl = sc;
2102		}
2103		}
2104		if (!init) { // Can only happen if unsafely published.
#	Line 1684 | Line 2113 | public class ConcurrentHashMapV8<K, V>
2113		// Unsafe mechanics
2114		private static final sun.misc.Unsafe UNSAFE;
2115		private static final long counterOffset;
2116	<	private static final long resizingOffset;
2116	>	private static final long sizeCtlOffset;
2117		private static final long ABASE;
2118		private static final int ASHIFT;
2119
#	Line 1695 | Line 2124 | public class ConcurrentHashMapV8<K, V>
2124		Class<?> k = ConcurrentHashMapV8.class;
2125		counterOffset = UNSAFE.objectFieldOffset
2126		(k.getDeclaredField("counter"));
2127	<	resizingOffset = UNSAFE.objectFieldOffset
2128	<	(k.getDeclaredField("resizing"));
2127	>	sizeCtlOffset = UNSAFE.objectFieldOffset
2128	>	(k.getDeclaredField("sizeCtl"));
2129		Class<?> sc = Node[].class;
2130		ABASE = UNSAFE.arrayBaseOffset(sc);
2131		ss = UNSAFE.arrayIndexScale(sc);

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines