A clean‑room, MIT‑licensed C++17 library re‑implements PlanB’s linearized B‑+tree IPv6 longest‑prefix‑match, adding a portable header‑only core, dynamic FIB support, and Python bindings. The code runs on any CPU, falling back to scalar paths when AVX‑512 is unavailable.

Benchmarks on an Intel i5‑1035G7 show single‑core throughput of 344 MLPS, with an 1.5× boost at an eight‑entry batch size. Compared to a Patricia radix trie, the tree delivers roughly twenty‑fold speed improvements while using only half the memory for 100 k prefixes.

The library’s modular design makes it suitable for ns‑3, SDN controllers, traffic classifiers, and educational tools. By exposing a public API and a dynamic rebuild‑and‑swap model, it offers a drop‑in replacement for legacy lookup engines, enabling researchers to evaluate new LPM structures without rebuilding entire routing stacks.