Mechanical ComparisonSharedArrayBuffer and Atomics Synchronization: Baseline · WidelyTypedArrays, ArrayBuffer, and Zero-Copy Transfers: Baseline · Widely

SharedArrayBuffer and Atomics Synchronization vs TypedArrays, ArrayBuffer, and Zero-Copy Transfers

Mechanical Verdict

SharedArrayBuffer shares memory concurrently across threads, whereas ArrayBuffer transferables move exclusive ownership.

Try SharedArrayBuffer and Atomics Synchronization LabTry TypedArrays, ArrayBuffer, and Zero-Copy Transfers LabRead SharedArrayBuffer and Atomics Synchronization DissectionRead TypedArrays, ArrayBuffer, and Zero-Copy Transfers Dissection

Structured Comparison Matrix

DimensionSharedArrayBuffer and Atomics SynchronizationTypedArrays, ArrayBuffer, and Zero-Copy Transfers
Mental ModelShared linear memory pages mapped across multiple JS threads, synchronized with hardware atomic operations and futex wait/notify queues.Contiguous binary buffers (ArrayBuffer) paired with unboxed numeric views (Uint8Array, Float32Array) that transfer across workers in O(1) via ownership handoff.
Maturity & DomainBaseline · Widely · Web APIs
Chrome 68+ · Firefox 79+ · Safari 15.2+
Baseline · Widely · JS Runtime
Chrome 7+ · Firefox 4+ · Safari 5.1+
Prerequisites
When Not to UseWhen not to use SharedArrayBuffer: Standard request/response worker tasks work cleanly with TypedArrays & Zero-Copy Transfers via postMessage and avoid COOP/COEP deployment constraints on third-party embeds.When not to use TypedArray: Avoid TypedArray for small, sparse, or mixed-type lists, creating a backing ArrayBuffer costs more than a small literal [1, 2, 3]; for hashing or encryption over binary buffers, pass the ArrayBuffer directly to Web Crypto (SubtleCrypto).

Engine Mechanics & Trade-Offs

SharedArrayBuffer and Atomics SynchronizationBaseline · Widely

Multiple JS isolates map the same virtual memory pages and coordinate reads/writes using CPU atomic instructions and OS futex queues.

Key Pitfalls: Calling Atomics.wait() on the main thread: Throws a TypeError synchronously; use Atomics.waitAsync() if the UI thread must observe a shared slot. Data races on non-atomic views: Mixing plain view[0] = 1 with Atomics.load(view, 0) causes torn reads on 64-bit values or stale CPU cache lines.

TypedArrays, ArrayBuffer, and Zero-Copy TransfersBaseline · Widely

ArrayBuffer allocates a raw contiguous byte block while TypedArray views index it directly without JS object boxing, enabling O(1) transfer between workers.

Key Pitfalls: Host Endianness Trap: Uint32Array reads bytes in host CPU order (little-endian on x86/ARM), silently scrambling big-endian network headers. Use-After-Transfer Crash: Reading a TypedArray after transferring its ArrayBuffer throws TypeError or returns empty slices.

Seen in Architecture Autopsies