• Dispatch XXH3 on x86 targets at runtime: link xxh_x86dispatch.c
  into vendored x86 and x86_64 builds so one baseline wheel selects
  SSE2, AVX2, or AVX512 for xxh3_64 and xxh3_128 inputs
  longer than 240 bytes, including streaming update(). On AVX2
  those inputs are about 1.5x faster.
• Upgrade bundled xxHash from v0.8.3 to v0.8.4.

• Build wheels for Python 3.15

Full list of changes: v3.6.0...v3.6.1

• Build wheels for Python 3.14 and 3.15

Full list of changes: v3.5.0...v3.5.1

• Build wheels for Python 3.13, 3.14 and 3.15

Full list of changes: v3.4.1...v3.4.2

• Improve type annotation compatibility with hashlib

Full Changelog: v4.0.1...v4.0.2

• Build wheels for Python 3.15

Full list of changes: v3.8.1...v3.8.2

• Build wheels for Python 3.15

Full list of changes: v3.7.2...v3.7.3

• Clean up _parse_init_args keyword handling so the defensive check
  for a duplicate data keyword raises a proper error message
• Add _xxhash module docstring

Full Changelog: v4.0.0...v4.0.1

• Breaking change: Drop support for Python 3.8, require Python >= 3.9
• Breaking change: Remove deprecated xxhash.VERSION_TUPLE
• Breaking change: The input keyword argument is renamed to data
  across constructors, update(), and one-shot functions
• Breaking change: str input is no longer accepted and raises
  TypeError: Strings must be encoded before hashing; encode to
  bytes before hashing
• Upgrade xxHash from v0.8.2 to v0.8.3. Note: on GCC/Clang source builds
  that target AVX2 (e.g. -march=x86-64-v3), upstream v0.8.3
  autovectorizes XXH64_update() and makes the xxh64 streaming path
  about 2x slower. The shipped wheels are built for baseline x86-64 and
  are unaffected. Source builds can work around it by adding
  -fno-tree-vectorize to the compiler flags.
• Add per-object locking for thread safety, with sub-interpreter and
  free-threaded (no-GIL) Python support
• Speed up hash constructors by switching them to tp_vectorcall.
• Build pyodide wasm32 wheels
• Add s390x big-endian test job
• Add Python 3.15 classifier
• CI: shard the PyPI upload into parallel groups and create the GitHub
  Release automatically

• Register the "benchmark" pytest mark to avoid PytestUnknownMarkWarning
• Update C extension docstrings and remove stale comments

Full list of changes: v3.8.0...v3.8.1