class Concurrent::Collection::AtomicReferenceMapBackend
@!visibility private
checks.
contention by only committing count updates upon these size
for sizes <= 64. The bulk putAll operation further reduces
variance for small table sizes, so we check on any collision
resizing, many fewer do so). But this approximation has high
meaning that only about 1 in 8 puts check threshold (and after
probability of this occurring at threshold is around 13%,
adding in others). Under uniform hash distributions, the
nodes (checked before adding in the x_if_absent
methods, after
contended, or upon adding to a bin already holding two or more
often, resizing is attempted either when a bin lock is
if read too frequently during concurrent access. To avoid reading so
which avoids contention on updates but can encounter cache thrashing
The element count is maintained using a Concurrent::ThreadSafe::Util::Adder
,
Lazy table initialization minimizes footprint until first use.
currently implemented.
provides support for shutdown-style clearing, which is also not
operations give up if ever forwarded to a null table, which
to support partitioned aggregate operations. Also, read-only
ranges of bins (via an alternate Traverser constructor)
The traversal scheme also applies to partial traversals of
expensive mechanics trigger only when necessary.
forwarding node, which is possible under the above encoding.) These more
new ones are established. (This sometimes requires transiently locking a
initialized with a reverse-forwarding node back to the old table until the
during a transfer, all earlier table bins may have become visible, so are
to the new table without revisiting nodes. However, when any node is skipped
towards the first. Upon seeing a forwarding node, traversals arrange to move
is arranged simply by proceeding from the last bin (+table.size - 1+) up
usable by any traversal. When there are no lock acquisition failures, this
also ensure that all accessible bins in both the old and new table are
none are available (i.e., only very rarely). The transfer operation must
have been skipped because of failure to acquire a lock, and blocks only if
later. Method rebuild
maintains a buffer of TRANSFER_BUFFER_SIZE bins that
transfer can skip a bin if it fails to acquire its lock, and revisit it
Each bin transfer requires its bin lock. However, unlike other cases, a
access and update operations restart, using the new table.
contains the next table as its key. On encountering a forwarding node,
contains only a special forwarding node (with hash field MOVED
) that
the midst of concurrently traversing table. Upon transfer, the old table bin
soon as they are no longer referenced by any reader thread that may be in
when a table doubles. The nodes they replace will be garbage collectable as
fields won’t change. On average, only about one-sixth of them need cloning
creation by catching cases where old nodes can be reused because their next
index, or move with a power of two offset. We eliminate unnecessary node
power-of-two expansion, the elements from each bin must either stay at same
bins, one by one, from the table to the next table. Because we are using
remains usable for reads and updates. Resizing proceeds by transferring
(using field size_control
, to arrange exclusion), but the table otherwise
(nominally, 0.75, but see below). Only a single thread performs the resize
The table is resized when occupancy exceeds a percentage threshold
roughly 1 / (8 * #elements) under random hashes.
Lock contention probability for two threads accessing distinct elements is
- more: less than 1 in ten million
- 8: 0.00000006
- 7: 0.00000094
- 6: 0.00001316
- 5: 0.00015795
- 4: 0.00157952
- 3: 0.01263606
- 2: 0.07581633
- 1: 0.30326533
- 0: 0.60653066
factorial(k)). The first values are:
expected occurrences of list size k are (exp(-0.5) * pow(0.5, k) /
large variance because of resizing granularity. Ignoring variance, the
about 0.5 on average, given the resizing threshold of 0.75, although with a
(en.wikipedia.org/wiki/Poisson_distribution) with a parameter of
Ideally, the frequency of nodes in bins follows a Poisson distribution
statistically, under random hash codes, this is not a common problem.
when user eql?
or mapping functions take a long time. However,
other nodes in a bin list protected by the same lock can stall, for example
The main disadvantage of per-bin locks is that other update operations on
Herlihy & Shavit.
This is a converse of sorts to the lazy locking technique described by
that only conditionally update may inspect nodes until the point of update.
deleted or the bin becomes invalidated (upon resizing). However, operations
appended to lists, once a node is first in a bin, it remains first until
first node after locking it, and retry if not. Because new nodes are always
When a node is locked, any update must first validate that it is still the
Using the first node of a list as a lock does not by itself suffice though:cheap_wait
/cheap_broadcast
constructions. See +Node#try_await_lock+.
monitors only for blocking and signalling using
hash field for lock control (see above), and so normally use builtintry_lock
construction, so we overlay these by using bits of the Node
support for these locks relies +Concurrent::ThreadSafe::Util::CheapLockable. However, we also need a
so instead use the first node of a bin list itself as a lock. Blocking
waste the space required to associate a distinct lock object with each bin,
operations (insert, delete, and replace) require locks. We do not want to
for put operations under most key/hash distributions. Other update
performed by just CASing it to the bin. This is by far the most common case
Insertion (via []=
or its variants) of the first node in an empty bin is
always have hash field == MOVED
).
except for forwarding nodes, for which the lower bits are zero (and so
The lower 28 bits of each Node
‘s hash field contain a the key’s hash code,
- 10 - Node
is a forwarding node
- 11 - Locked and may have a thread waiting for lock
- 01 - Locked
- 00 - Normal
below, these top bits are used as follows:
are available anyway because of addressing constraints. As explained further
We use the top two bits of Node
hash fields for control purposes – they
hash code and non-nullness of value before checking key equality.
always accurately traversable under volatile reads, so long as lookups check
volatile/atomic reads, writes, and CASes. The lists of nodes within bins are
often, the list has only zero or one Node
). Table accesses require
insertion. Each bin in the table normally contains a list of +Node+s (most
The table is lazily initialized to a power-of-two size upon the first
precludes factoring into smaller methods.
explained below leads to a lot of code sprawl because retry-control
Each key-value mapping is held in a Node
. The validation-based approach
initial insertion rates on an empty table by many threads.
consumption about the same or better than plain Hash
, and to support high
while minimizing update contention. Secondary goals are to keep space
readability (typically method []
, but also iteration and related methods)
The primary design goal of this hash table is to maintain concurrent
## Design overview
table.
exactly the same hash
is a sure way to slow down performance of any hash
to allocate for the given number of elements. Note that using many keys with
specifying the table density to be used in calculating the amount of space
argument provides a further means of customizing initial table capacity by
initializer argument. An additional optional :load_factor constructor
a good idea to provide a size estimate as an optional :initial_capacity
kind of hash table may be a relatively slow operation. When possible, it is
time/space tradeoff for hash tables. However, resizing this or any other
added and removed, but overall, this maintains a commonly accepted
resizing). There may be much variance around this average as mappings are
bins per mapping (corresponding to a 0.75 load factor threshold for
table size), with the expected average effect of maintaining roughly two
keys that have distinct hash codes but fall into the same slot modulo the
The table is dynamically expanded when there are too many collisions (i.e.,
for monitoring or estimation purposes, but not for program control.
the results of these methods reflect transient states that may be adequate
when a map is not undergoing concurrent updates in other threads. Otherwise
status methods including +size()+ and empty?
} are typically useful only
the start of the each_pair
. Bear in mind that the results of aggregate
elements reflecting the state of the hash table at some point at or since
removal of only some entries. Similarly, the each_pair
iterator yields
operations such as +clear()+, concurrent retrievals may reflect insertion ornil
) retrieval for that key reporting the updated value.) For aggregate
operation for a given key bears a happens-before relation with any (non
update operations holding upon their onset. (More formally, an update
operations. Retrievals reflect the results of the most recently completed
Retrieval operations generally do not block, so may overlap with update
access.
not any support for locking the entire table in a way that prevents all
thread-safe, retrieval operations do not entail locking, and there is
concurrency for updates. However, even though all operations are
A hash table supporting full concurrency of retrievals and high expected
size exceeds a threshold).
The Ruby port skips out the TreeBin
(red-black trees for use in bins whose
Original source code available here:
available in public domain.
A Ruby port of the Doug Lea’s jsr166e.ConcurrentHashMapV8 class version 1.59
def [](key)
def [](key) get_or_default(key) end
def []=(key, value)
def []=(key, value) get_and_set(key, value) value end
def attempt_compute(key, hash, current_table, i, node, node_hash)
def attempt_compute(key, hash, current_table, i, node, node_hash) added = false current_table.try_lock_via_hash(i, node, node_hash) do predecessor_node = nil while true if node.matches?(key, hash) && NULL != (value = node.value) if NULL == (node.value = value = yield(value)) current_table.delete_node_at(i, node, predecessor_node) decrement_size value = nil end return true, value end predecessor_node = node unless node = node.next if NULL == (value = yield(NULL)) value = nil else predecessor_node.next = Node.new(hash, key, value) added = true increment_size end return true, value end end end ensure check_for_resize if added end
def attempt_get_and_set(key, value, hash, current_table, i, node, node_hash)
def attempt_get_and_set(key, value, hash, current_table, i, node, node_hash) node_nesting = nil current_table.try_lock_via_hash(i, node, node_hash) do node_nesting = 1 old_value = nil found_old_value = false while node if node.matches?(key, hash) && NULL != (old_value = node.value) found_old_value = true node.value = value break end last = node unless node = node.next last.next = Node.new(hash, key, value) break end node_nesting += 1 end return true, old_value if found_old_value increment_size true end ensure check_for_resize if node_nesting && (node_nesting > 1 || current_table.size <= 64) end
def attempt_internal_compute_if_absent(key, hash, current_table, i, node, node_hash)
def attempt_internal_compute_if_absent(key, hash, current_table, i, node, node_hash) added = false current_table.try_lock_via_hash(i, node, node_hash) do while true if node.matches?(key, hash) && NULL != (value = node.value) return true, value end last = node unless node = node.next last.next = Node.new(hash, key, value = yield) added = true increment_size return true, value end end end ensure check_for_resize if added end
def attempt_internal_replace(key, expected_old_value, hash, current_table, i, node, node_hash)
def attempt_internal_replace(key, expected_old_value, hash, current_table, i, node, node_hash) current_table.try_lock_via_hash(i, node, node_hash) do predecessor_node = nil old_value = NULL begin if node.matches?(key, hash) && NULL != (current_value = node.value) if NULL == expected_old_value || expected_old_value == current_value # NULL == expected_old_value means whatever value old_value = current_value if NULL == (node.value = yield(old_value)) current_table.delete_node_at(i, node, predecessor_node) decrement_size end end break end predecessor_node = node end while node = node.next return true, old_value end end
def check_for_resize
transfers bins. Rechecks occupancy after a transfer to see if another
If table is too small and not already resizing, creates next table and
def check_for_resize while (current_table = table) && MAX_CAPACITY > (table_size = current_table.size) && NOW_RESIZING != (size_ctrl = size_control) && size_ctrl < @counter.sum try_in_resize_lock(current_table, size_ctrl) do self.table = rebuild(current_table) (table_size << 1) - (table_size >> 1) # 75% load factor end end end
def clear
def clear return self unless current_table = table current_table_size = current_table.size deleted_count = i = 0 while i < current_table_size if !(node = current_table.volatile_get(i)) i += 1 elsif (node_hash = node.hash) == MOVED current_table = node.key current_table_size = current_table.size elsif Node.locked_hash?(node_hash) decrement_size(deleted_count) # opportunistically update count deleted_count = 0 node.try_await_lock(current_table, i) else current_table.try_lock_via_hash(i, node, node_hash) do begin deleted_count += 1 if NULL != node.value # recheck under lock node.value = nil end while node = node.next current_table.volatile_set(i, nil) i += 1 end end end decrement_size(deleted_count) self end
def compute(key)
def compute(key) internal_compute(key) do |old_value| if (new_value = yield(NULL == old_value ? nil : old_value)).nil? NULL else new_value end end end
def compute_if_absent(key)
def compute_if_absent(key) hash = key_hash(key) current_table = table || initialize_table while true if !(node = current_table.volatile_get(i = current_table.hash_to_index(hash))) succeeded, new_value = current_table.try_to_cas_in_computed(i, hash, key) { yield } if succeeded increment_size return new_value end elsif (node_hash = node.hash) == MOVED current_table = node.key elsif NULL != (current_value = find_value_in_node_list(node, key, hash, node_hash & HASH_BITS)) return current_value elsif Node.locked_hash?(node_hash) try_await_lock(current_table, i, node) else succeeded, value = attempt_internal_compute_if_absent(key, hash, current_table, i, node, node_hash) { yield } return value if succeeded end end end
def compute_if_present(key)
def compute_if_present(key) new_value = nil internal_replace(key) do |old_value| if (new_value = yield(NULL == old_value ? nil : old_value)).nil? NULL else new_value end end new_value end
def decrement_size(by = 1)
def decrement_size(by = 1) @counter.add(-by) end
def delete(key)
def delete(key) replace_if_exists(key, NULL) end
def delete_pair(key, value)
def delete_pair(key, value) result = internal_replace(key, value) { NULL } if result && NULL != result !!result else false end end
def each_pair
def each_pair return self unless current_table = table current_table_size = base_size = current_table.size i = base_index = 0 while base_index < base_size if node = current_table.volatile_get(i) if node.hash == MOVED current_table = node.key current_table_size = current_table.size else begin if NULL != (value = node.value) # skip deleted or special nodes yield node.key, value end end while node = node.next end end if (i_with_base = i + base_size) < current_table_size i = i_with_base # visit upper slots if present else i = base_index += 1 end end self end
def empty?
def empty? size == 0 end
def find_value_in_node_list(node, key, hash, pure_hash)
def find_value_in_node_list(node, key, hash, pure_hash) do_check_for_resize = false while true if pure_hash == hash && node.key?(key) && NULL != (value = node.value) return value elsif node = node.next do_check_for_resize = true # at least 2 nodes -> check for resize pure_hash = node.pure_hash else return NULL end end ensure check_for_resize if do_check_for_resize end
def get_and_set(key, value) # internalPut in the original CHMV8
def get_and_set(key, value) # internalPut in the original CHMV8 hash = key_hash(key) current_table = table || initialize_table while true if !(node = current_table.volatile_get(i = current_table.hash_to_index(hash))) if current_table.cas_new_node(i, hash, key, value) increment_size break end elsif (node_hash = node.hash) == MOVED current_table = node.key elsif Node.locked_hash?(node_hash) try_await_lock(current_table, i, node) else succeeded, old_value = attempt_get_and_set(key, value, hash, current_table, i, node, node_hash) break old_value if succeeded end end end
def get_or_default(key, else_value = nil)
def get_or_default(key, else_value = nil) hash = key_hash(key) current_table = table while current_table node = current_table.volatile_get_by_hash(hash) current_table = while node if (node_hash = node.hash) == MOVED break node.key elsif (node_hash & HASH_BITS) == hash && node.key?(key) && NULL != (value = node.value) return value end node = node.next end end else_value end
def increment_size
def increment_size @counter.increment end
def initialize(options = nil)
def initialize(options = nil) super() @counter = Concurrent::ThreadSafe::Util::Adder.new initial_capacity = options && options[:initial_capacity] || DEFAULT_CAPACITY self.size_control = (capacity = table_size_for(initial_capacity)) > MAX_CAPACITY ? MAX_CAPACITY : capacity end
def initialize_copy(other)
def initialize_copy(other) super @counter = Concurrent::ThreadSafe::Util::Adder.new self.table = nil self.size_control = (other_table = other.table) ? other_table.size : DEFAULT_CAPACITY self end
def initialize_table
def initialize_table until current_table ||= table if (size_ctrl = size_control) == NOW_RESIZING Thread.pass # lost initialization race; just spin else try_in_resize_lock(current_table, size_ctrl) do initial_size = size_ctrl > 0 ? size_ctrl : DEFAULT_CAPACITY current_table = self.table = Table.new(initial_size) initial_size - (initial_size >> 2) # 75% load factor end end end current_table end
def internal_compute(key, &block)
def internal_compute(key, &block) hash = key_hash(key) current_table = table || initialize_table while true if !(node = current_table.volatile_get(i = current_table.hash_to_index(hash))) succeeded, new_value = current_table.try_to_cas_in_computed(i, hash, key, &block) if succeeded if NULL == new_value break nil else increment_size break new_value end end elsif (node_hash = node.hash) == MOVED current_table = node.key elsif Node.locked_hash?(node_hash) try_await_lock(current_table, i, node) else succeeded, new_value = attempt_compute(key, hash, current_table, i, node, node_hash, &block) break new_value if succeeded end end end
def internal_replace(key, expected_old_value = NULL, &block)
Someday when details settle down a bit more, it might be worth
if present), which also makes pre-emptive resize checks worthwhile.
* compute_if_absent prescans for mapping without lock (and fails to add
* Plain +get_and_set+ checks for and performs resize after insertion.
The others interweave other checks and/or alternative actions:
4. Lock and validate; if valid, scan and add or update
3. If bin stale, use new table
2. If bin empty, try to CAS new node
1. If table uninitialized, create
the same basic structure:
little more complicated than the last. All have
Internal versions of the insertion methods, each a
def internal_replace(key, expected_old_value = NULL, &block) hash = key_hash(key) current_table = table while current_table if !(node = current_table.volatile_get(i = current_table.hash_to_index(hash))) break elsif (node_hash = node.hash) == MOVED current_table = node.key elsif (node_hash & HASH_BITS) != hash && !node.next # precheck break # rules out possible existence elsif Node.locked_hash?(node_hash) try_await_lock(current_table, i, node) else succeeded, old_value = attempt_internal_replace(key, expected_old_value, hash, current_table, i, node, node_hash, &block) return old_value if succeeded end end NULL end
def key?(key)
def key?(key) get_or_default(key, NULL) != NULL end
def key_hash(key)
def key_hash(key) key.hash & HASH_BITS end
def lock_and_clean_up_reverse_forwarders(old_table, old_table_size, new_table, i, forwarder)
def lock_and_clean_up_reverse_forwarders(old_table, old_table_size, new_table, i, forwarder) # transiently use a locked forwarding node locked_forwarder = Node.new(moved_locked_hash = MOVED | LOCKED, new_table, NULL) if old_table.cas(i, nil, locked_forwarder) new_table.volatile_set(i, nil) # kill the potential reverse forwarders new_table.volatile_set(i + old_table_size, nil) # kill the potential reverse forwarders old_table.volatile_set(i, forwarder) locked_forwarder.unlock_via_hash(moved_locked_hash, MOVED) true end end
def merge_pair(key, value)
def merge_pair(key, value) internal_compute(key) do |old_value| if NULL == old_value || !(value = yield(old_value)).nil? value else NULL end end end
def rebuild(table)
def rebuild(table) old_table_size = table.size new_table = table.next_in_size_table # puts "#{old_table_size} -> #{new_table.size}" forwarder = Node.new(MOVED, new_table, NULL) rev_forwarder = nil locked_indexes = nil # holds bins to revisit; nil until needed locked_arr_idx = 0 bin = old_table_size - 1 i = bin while true if !(node = table.volatile_get(i)) # no lock needed (or available) if bin >= 0, because we're not popping values from locked_indexes until we've run through the whole table redo unless (bin >= 0 ? table.cas(i, nil, forwarder) : lock_and_clean_up_reverse_forwarders(table, old_table_size, new_table, i, forwarder)) elsif Node.locked_hash?(node_hash = node.hash) locked_indexes ||= ::Array.new if bin < 0 && locked_arr_idx > 0 locked_arr_idx -= 1 i, locked_indexes[locked_arr_idx] = locked_indexes[locked_arr_idx], i # swap with another bin redo end if bin < 0 || locked_indexes.size >= TRANSFER_BUFFER_SIZE node.try_await_lock(table, i) # no other options -- block redo end rev_forwarder ||= Node.new(MOVED, table, NULL) redo unless table.volatile_get(i) == node && node.locked? # recheck before adding to list locked_indexes << i new_table.volatile_set(i, rev_forwarder) new_table.volatile_set(i + old_table_size, rev_forwarder) else redo unless split_old_bin(table, new_table, i, node, node_hash, forwarder) end if bin > 0 i = (bin -= 1) elsif locked_indexes && !locked_indexes.empty? bin = -1 i = locked_indexes.pop locked_arr_idx = locked_indexes.size - 1 else return new_table end end end
def replace_if_exists(key, new_value)
def replace_if_exists(key, new_value) if (result = internal_replace(key) { new_value }) && NULL != result result end end
def replace_pair(key, old_value, new_value)
def replace_pair(key, old_value, new_value) NULL != internal_replace(key, old_value) { new_value } end
def size
def size (sum = @counter.sum) < 0 ? 0 : sum # ignore transient negative values end
def split_bin(new_table, i, node, node_hash)
def split_bin(new_table, i, node, node_hash) bit = new_table.size >> 1 # bit to split on run_bit = node_hash & bit last_run = nil low = nil high = nil current_node = node # this optimises for the lowest amount of volatile writes and objects created while current_node = current_node.next unless (b = current_node.hash & bit) == run_bit run_bit = b last_run = current_node end end if run_bit == 0 low = last_run else high = last_run end current_node = node until current_node == last_run pure_hash = current_node.pure_hash if (pure_hash & bit) == 0 low = Node.new(pure_hash, current_node.key, current_node.value, low) else high = Node.new(pure_hash, current_node.key, current_node.value, high) end current_node = current_node.next end new_table.volatile_set(i, low) new_table.volatile_set(i + bit, high) end
def split_old_bin(table, new_table, i, node, node_hash, forwarder)
def split_old_bin(table, new_table, i, node, node_hash, forwarder) table.try_lock_via_hash(i, node, node_hash) do split_bin(new_table, i, node, node_hash) table.volatile_set(i, forwarder) end end
def table_size_for(entry_count)
def table_size_for(entry_count) size = 2 size <<= 1 while size < entry_count size end
def try_await_lock(current_table, i, node)
def try_await_lock(current_table, i, node) check_for_resize # try resizing if can't get lock node.try_await_lock(current_table, i) end
def try_in_resize_lock(current_table, size_ctrl)
def try_in_resize_lock(current_table, size_ctrl) if cas_size_control(size_ctrl, NOW_RESIZING) begin if current_table == table # recheck under lock size_ctrl = yield # get new size_control end ensure self.size_control = size_ctrl end end end