A cache-friendly IPv6 LPM with AVX-512 (linearized B+-tree, real BGP benchmarks)

Article

• PlanB: C++17 IPv6 longest-prefix-match using linearized B+-tree
• AVX-512 SIMD path with scalar fallback; wait-free lookups via rebuild-and-swap
• Benchmarked on real RIPE RIS BGP data (~254K prefixes)

Discussion

• Question raised: why runtime AVX-512 detection instead of compile-time #ifdef
• Curiosity about RISC-V vector instruction equivalent
• Brief lament that it’s C++ rather than C

Discuss on HN