avian_lwma

High-performance LWMA (Zawy v3) target calculator for dual-algo Avian forks (X16RT + MinotaurX).

• No Boost/OpenSSL: ships a tiny 256-bit integer core (GCC/Clang required for __uint128_t).
• pybind11: import from Python and call at header-sync/checkpoint time.
• Per-algo windows: pass the last N headers of a given algo.
• Optional WASM FFI: compile the same core to .wasm if you want a web/Node verifier (not included in this zip).

Quick start (build wheel)

pip install -U pip build
pip install pybind11 scikit-build-core
pip install .

Windows: build with MSYS2/MinGW or Clang, or use WSL. (MSVC lacks __uint128_t; a MSVC-safe Big256 can be added.)

CMake build

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j

Python usage

from avian_lwma import HeaderLite, next_bits_window, filter_last_n_for_algo

N = 45           # LWMA window
T = 30           # target spacing (seconds)
POW_LIMITS = {
    0: "00000fffffffffffffffffffffffffffffffffffffffffffffffffffffffffff",  # X16RT
    1: "000fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff",  # MinotaurX
}

def algo_from_version(ver: int) -> int:
    return 1 if (ver & (1<<28)) else 0  # example mapping

# chain: list[HeaderLite] oldest->newest (fill from your header DB)
window = filter_last_n_for_algo(chain, N, algo_from_version, algo_id=1)
expected_bits = next_bits_window(window, T, len(window), POW_LIMITS[1])

API

• HeaderLite(time: uint32, version: uint32, bits: uint32)
• next_bits_window(window, target_spacing_s, N, pow_limit_hex) -> int (compact bits)
• filter_last_n_for_algo(chain, N, algo_from_version, algo_id) -> list[HeaderLite]

Notes

• Targets are full 256-bit integers decoded from compact bits. The result is clamped to pow_limit before re-encoding.
• The Big256 here uses __uint128_t for intermediate math. For MSVC, swap in a different Big256 (PRs welcome).