folly::bskip_detail::AtomicSlot

Declared in <folly/ConcurrentBSkipList‐detail.h>

The atomic load instruction IS the bracket: a writer's store either fully precedes or fully follows the reader's load and never tears. Readers need no surrounding protocol.

Wraps folly::relaxed_atomic<T>; synthesized copy/move ops let std::array element‐assignment compile in splitKeys/moveChildren (raw std::atomic isn't copyable).

Alignment inherits from std::atomic<T> and is STRICTER than alignof(T) when the lock‐free instruction is wider than T's natural alignment (e.g. 16‐byte cmpxchg16b on an 8‐byte‐aligned 16‐byte struct). Do NOT replace this wrapper with std::atomic_ref<T> over raw T storage: atomic_ref surrenders that alignment guarantee back to the caller AND regressed children_ traversal on 16B‐key benchmarks (misaligned loads).

Created with MrDocs