NAME
Data::Pool::Shared - Fixed-size shared-memory object pool for Linux
SYNOPSIS
use Data::Pool::Shared;
# Raw byte pool — 100 slots of 64 bytes each
my $pool = Data::Pool::Shared->new('/tmp/pool.shm', 100, 64);
my $idx = $pool->alloc; # allocate a slot
$pool->set($idx, "hello world"); # write data
my $data = $pool->get($idx); # read data
$pool->free($idx); # release slot
# Typed pools
my $ints = Data::Pool::Shared::I64->new('/tmp/ints.shm', 1000);
my $i = $ints->alloc;
$ints->set($i, 42);
$ints->add($i, 8); # atomic add, returns 50
$ints->cas($i, 50, 99); # atomic CAS
say $ints->get($i); # 99
my $floats = Data::Pool::Shared::F64->new('/tmp/f.shm', 100);
my $strs = Data::Pool::Shared::Str->new('/tmp/s.shm', 100, 256);
# Guard — auto-free on scope exit
{
my ($idx, $guard) = $pool->alloc_guard;
$pool->set($idx, $data);
# ... use slot ...
} # auto-freed
# Lock-free primitives
my $prev = $ints->cmpxchg($i, 99, 200); # CAS returning old value
$prev = $ints->xchg($i, 300); # atomic exchange
# Batch operations
my $slots = $pool->alloc_n(10); # allocate 10 slots
$pool->free_n($slots); # batch free
# Zero-copy and raw pointers
my $sv = $pool->slot_sv($idx); # read-only SV over slot memory
my $ptr = $pool->ptr($idx); # C pointer (UV) for FFI/OpenGL
my @all = @{ $pool->allocated_slots }; # list all allocated indices
# Convenience
my $j = $ints->alloc_set(42); # alloc + set
$j = $ints->try_alloc_set(42); # non-blocking
# Crash recovery
my $n = $pool->recover_stale; # free slots owned by dead PIDs
# Cross-process via fork
if (fork == 0) {
my $child = Data::Pool::Shared::I64->new('/tmp/ints.shm', 1000);
my $i = $child->alloc;
$child->set($i, $$);
exit;
}
# Anonymous (fork-inherited)
$pool = Data::Pool::Shared::I64->new(undef, 100);
# memfd (fd-passable)
$pool = Data::Pool::Shared::I64->new_memfd("my_pool", 100);
my $fd = $pool->memfd;DESCRIPTION
Data::Pool::Shared provides a fixed-size object pool in shared memory. Slots are allocated and freed explicitly, like a memory allocator but for cross-process shared objects.
Unlike Data::Buffer::Shared (index-based array access), Pool provides allocate/free semantics: you request a slot, use it, and return it. The pool tracks which slots are in use via a lock-free bitmap.
Linux-only. Requires 64-bit Perl.
Variants
Allocation
Allocation uses a CAS-based bitmap scan (lock-free). Each 64-slot group is managed by one atomic uint64_t word. On contention, CAS retries automatically. When the pool is full, alloc blocks on a futex until a slot is freed.
Crash Safety
Each slot records the PID of its allocator. recover_stale scans for slots owned by dead processes and frees them. Call periodically or on startup for crash recovery.
CONSTRUCTORS
# Raw pool
my $p = Data::Pool::Shared->new($path, $capacity, $elem_size);
my $p = Data::Pool::Shared->new(undef, $capacity, $elem_size); # anonymous
my $p = Data::Pool::Shared->new_memfd($name, $capacity, $elem_size);
my $p = Data::Pool::Shared->new_from_fd($fd);
# I64 / I32 / F64 pools (elem_size is implicit)
my $p = Data::Pool::Shared::I64->new($path, $capacity);
my $p = Data::Pool::Shared::I32->new($path, $capacity);
my $p = Data::Pool::Shared::F64->new($path, $capacity);
my $p = Data::Pool::Shared::I64->new_memfd($name, $capacity);
# All variants support new_from_fd
my $p = Data::Pool::Shared::I64->new_from_fd($fd);
# Str pool
my $p = Data::Pool::Shared::Str->new($path, $capacity, $max_len);
my $p = Data::Pool::Shared::Str->new_memfd($name, $capacity, $max_len);
my $p = Data::Pool::Shared::Str->new_from_fd($fd);METHODS
Allocation
my $idx = $pool->alloc; # block until available
my $idx = $pool->alloc($timeout); # with timeout (seconds)
my $idx = $pool->alloc(0); # non-blocking
my $idx = $pool->try_alloc; # non-blocking (alias)Returns slot index on success, undef on failure/timeout.
$pool->free($idx); # release slot (returns true/false)Batch Operations
my $slots = $pool->alloc_n($n); # allocate N slots (blocking)
my $slots = $pool->alloc_n($n, $timeout); # with timeout
my $slots = $pool->alloc_n($n, 0); # non-blocking
# returns arrayref of indices, or undef (all-or-nothing)
my $freed = $pool->free_n(\@indices); # batch free, returns count freed
# single used-decrement + single futex wake (faster than N individual frees)
my $slots = $pool->allocated_slots; # arrayref of all allocated indicesData Access
my $val = $pool->get($idx); # read slot
$pool->set($idx, $val); # write slotFor I64/I32 variants:
my $ok = $pool->cas($idx, $old, $new); # atomic CAS, returns bool
my $old = $pool->cmpxchg($idx, $old, $new); # atomic CAS, returns old value
my $old = $pool->xchg($idx, $val); # atomic exchange, returns old
my $val = $pool->add($idx, $delta); # atomic add, returns new value
my $val = $pool->incr($idx); # atomic increment
my $val = $pool->decr($idx); # atomic decrementFor Str variant:
my $max = $pool->max_len; # maximum string lengthRaw Pointers
my $ptr = $pool->ptr($idx); # raw C pointer to slot data (UV)
my $ptr = $pool->data_ptr; # pointer to start of data sectionptr returns the memory address of a slot's data as an unsigned integer. Use with FFI::Platypus, OpenGL _c functions, or XS code that needs a void*.
data_ptr returns the base of the contiguous data region. Slots are laid out as data_ptr + idx * elem_size.
Warning: The returned pointer becomes dangling if the pool object is destroyed. Do not use after the pool goes out of scope.
Zero-Copy Access
my $sv = $pool->slot_sv($idx); # SV backed by slot memoryReturns a read-only scalar whose PV points directly into the shared memory slot. Reading the scalar reads the slot with no memcpy. Useful for large slots where avoiding copy matters.
The scalar must not outlive the pool object. To modify the slot, use set().
Status
my $ok = $pool->is_allocated($idx);
my $cap = $pool->capacity;
my $esz = $pool->elem_size;
my $n = $pool->used; # allocated count
my $n = $pool->available; # free count
my $pid = $pool->owner($idx); # PID of allocatorRecovery
my $n = $pool->recover_stale; # free slots owned by dead PIDs
$pool->reset; # free all slots (exclusive access only)Guards
my ($idx, $guard) = $pool->alloc_guard; # auto-free on scope exit
my ($idx, $guard) = $pool->alloc_guard($timeout);
my ($idx, $guard) = $pool->try_alloc_guard; # non-blockingConvenience
my $idx = $pool->alloc_set($val); # alloc + set
my $idx = $pool->alloc_set($val, $timeout); # with timeout
my $idx = $pool->try_alloc_set($val); # non-blocking
$pool->each_allocated(sub { my $idx = shift; ... });Common Methods
my $p = $pool->path; # backing file (undef if anon)
my $fd = $pool->memfd; # memfd fd (-1 if not memfd)
$pool->sync; # msync to disk
$pool->unlink; # remove backing file
my $s = $pool->stats; # diagnostic hashrefeventfd Integration
my $fd = $pool->eventfd; # create eventfd
$pool->eventfd_set($fd); # use existing fd
my $fd = $pool->fileno; # current eventfd (-1 if none)
$pool->notify; # signal eventfd
my $n = $pool->eventfd_consume; # drain counterSTATS
stats() returns a hashref with diagnostic counters. All values are approximate under concurrency.
capacity— total slot count (immutable)elem_size— bytes per slot (immutable)used— currently allocated slot countavailable— currently free slot count (capacity - used)waiters— processes currently blocked onallocmmap_size— total mmap region size in bytesallocs— cumulative successful allocationsfrees— cumulative frees (including stale recovery)waits—alloccalls that entered the retry looptimeouts—alloccalls that timed outrecoveries— slots freed byrecover_stale
SECURITY
The shared memory region (mmap) is writable by all processes that open it. A malicious process with write access to the backing file or memfd can corrupt header fields (bitmap, counters, slot data) and cause other processes to crash, spin, or return incorrect data. Do not share backing files with untrusted processes. Use anonymous mode or memfd with restricted fd passing for isolation.
PERFORMANCE
Allocation scans a bitmap of
ceil(capacity/64)words. O(capacity/64) worst case, O(1) amortized with scan_hint.Each allocation is a single CAS on one bitmap word. Under contention, CAS retries on the same word are ~10ns each.
When pool is full,
allocblocks on a futex (zero CPU). Woken by a singleFUTEX_WAKEsyscall onfree.free_nbatches N frees into a singleuseddecrement and a singleFUTEX_WAKEsyscall — faster than N individual frees.slot_svprovides zero-copy access to slot data, avoidingmemcpyoverhead for large slots.Typed variants (I64, I32) use atomic load/store/CAS/add directly on the mmap'd memory — no locking overhead.
BENCHMARKS
Measured on a single-socket x86_64 Linux system, Perl 5.40.
Single process (1M ops):
I64 alloc + free 3.3M/s
I64 get/set ~10M/s
I64 add/incr ~10M/s
I64 cas 9.8M/s
Str set (48B) ~10M/s
Str get (48B) 7.5M/s
alloc_set + free 1.9M/s
Multi-process (8 workers, 200K ops each, cap=64):
I64 alloc/free 4.7M/s aggregate
I64 alloc/set/get/free 5.1M/s aggregate
I64 atomic add 22.9M/s aggregate
Str alloc/set/get/free 4.9M/s aggregate
Batch (single process, alloc_n + free_n):
batch=1 ~2.3M/s
batch=16 ~400K/s (vs ~200K individual)
batch=64 ~110K/s (vs ~50K individual, 2x gain)Bottleneck is Perl XS call overhead, not the CAS or futex.
SEE ALSO
Data::Buffer::Shared - typed shared array (index-based, no alloc/free)
Data::Stack::Shared - LIFO stack
Data::Deque::Shared - double-ended queue
Data::Queue::Shared - FIFO queue
Data::ReqRep::Shared - request-reply
Data::Log::Shared - append-only log (WAL)
Data::Sync::Shared - synchronization primitives
Data::HashMap::Shared - concurrent hash table
Data::PubSub::Shared - publish-subscribe ring
AUTHOR
vividsnow
LICENSE
This is free software; you can redistribute it and/or modify it under the same terms as Perl itself.