xxHash Calculator

xxHash is one of the fastest hash functions ever published — fast enough to saturate RAM bandwidth on a single CPU thread — which is why it powers LZ4, Zstandard, and countless databases. This calculator computes both the XXH32 and XXH64 variants with an optional seed, entirely in your browser.

How xxHash works: the striped accumulation design

Both variants follow the same structural shape, designed for CPU pipeline efficiency:

Phase 1 — striped accumulation (inputs longer than 16/32 bytes)

xxHash maintains four independent accumulators (v1–v4), each seeded from the input seed combined with constants. For every lane (4 bytes in XXH32, 8 bytes in XXH64), one accumulator absorbs a chunk through a round step:

v = v + (data * prime2)
v = rotateLeft(v, rotationAmount)
v = v * prime1

Processing four accumulators in rotation is what lets xxHash keep four CPU execution units busy simultaneously, achieving near-memory-bandwidth throughput.

Phase 2 — accumulator merging

The four accumulators are folded into a single value using rotates and additions:

h = rotl(v1, 1) + rotl(v2, 7) + rotl(v3, 12) + rotl(v4, 18)

For XXH64 each accumulator is then merged through an additional mixing step before the fold.

Phase 3 — tail bytes and length

Remaining bytes (up to 7 for XXH64, up to 3 for XXH32) are consumed in 8-byte, 4-byte, and 1-byte sub-rounds, and the total input length is added to the accumulator.

Phase 4 — finalisation avalanche

An avalanche sequence of XORs, shifts, and prime multiplications ensures every output bit depends on every input bit:

h = h XOR (h >> shift1)
h = h * prime2
h = h XOR (h >> shift2)
h = h * prime3
h = h XOR (h >> shift3)

XXH32 uses 32-bit lanes and arithmetic; XXH64 uses 64-bit lanes (BigInt in this browser tool for correctness).

xxHash constants

The constants are chosen from the golden-ratio-derived sequence floor(2^32 * phi) and its relatives — a property that ensures good bit mixing across the word width.

Reference values

Use abc with seed 0 to verify the tool matches the reference implementation.

Where xxHash is actually used

• LZ4 and Zstandard — both compression formats use XXH32 or XXH64 to checksum compressed frames.
• ClickHouse and other columnar databases — xxHash is used for sharding and deduplication of column data.
• RocksDB / LevelDB — some configurations use XXH as an alternative to CRC32.
• File synchronisation tools — several rsync-derivative and backup tools prefer xxHash over MD5 for its speed without the baggage of a broken cryptographic algorithm.
• Content-defined chunking — rolling-window chunking algorithms (e.g., FastCDC) use xxHash to fingerprint chunks for deduplication.

Comparison with similar hashes

xxHash is non-cryptographic — never use it where an adversary can engineer collisions. For security-sensitive contexts use SipHash (for keyed MACs) or SHA-256/SHA-3 (for unkeyed integrity). All computation in this tool is local; nothing is uploaded.