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
    Data::Pool::Shared - raw byte slots (any elem_size)
    Data::Pool::Shared::I64 - int64_t (atomic get/set/cas/add)
    Data::Pool::Shared::F64 - double
    Data::Pool::Shared::I32 - int32_t (atomic get/set/cas/add)
    Data::Pool::Shared::Str - fixed-length strings

  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 indices

  Data Access
        my $val = $pool->get($idx);         # read slot
        $pool->set($idx, $val);             # write slot

    For 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 decrement

    For Str variant:

        my $max = $pool->max_len;           # maximum string length

  Raw Pointers
        my $ptr = $pool->ptr($idx);     # raw C pointer to slot data (UV)
        my $ptr = $pool->data_ptr;      # pointer to start of data section

    "ptr" 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 memory

    Returns 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 allocator

  Recovery
        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-blocking

  Convenience
        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 hashref

   eventfd 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 counter

STATS
    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 count
    "available" — currently free slot count ("capacity - used")
    "waiters" — processes currently blocked on "alloc"
    "mmap_size" — total mmap region size in bytes
    "allocs" — cumulative successful allocations
    "frees" — cumulative frees (including stale recovery)
    "waits" — "alloc" calls that entered the retry loop
    "timeouts" — "alloc" calls that timed out
    "recoveries" — slots freed by "recover_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, "alloc" blocks on a futex (zero CPU). Woken by a
        single "FUTEX_WAKE" syscall on "free".

    *   "free_n" batches N frees into a single "used" decrement and a single
        "FUTEX_WAKE" syscall — faster than N individual frees.

    *   "slot_sv" provides zero-copy access to slot data, avoiding "memcpy"
        overhead 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                  ~10M/s
          I64 set                  ~10M/s
          I64 add/incr             ~10M/s
          I64 cas                   9.4M/s
          Str set (48B)            ~10M/s
          Str get (48B)             7.6M/s
          alloc_set + free          1.8M/s

        Multi-process (8 workers, 200K ops each, cap=64):
          I64 alloc/free            5.4M/s aggregate
          I64 alloc/set/get/free    5.3M/s aggregate
          I64 atomic add           28.5M/s aggregate
          Str alloc/set/get/free    4.8M/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.

AUTHOR
    vividsnow

LICENSE
    This is free software; you can redistribute it and/or modify it under
    the same terms as Perl itself.