Performance

• hayahash64: fast hashing without requiring SIMD or wide multiply 2026-08-02

  I built a 64-bit hash whose algorithm requires only ordinary 64-bit multiplies, shifts, rotates, and XORs, and it passes all 188 default tests of a pinned SMHasher3 revision. On native CPUs that usually costs it the crown. In scalar WebAssembly the tables turn, and on Zen 5 a compiler-vectorized build reaches 61 GB/s from portable C. Here's how it works, and a browser playground where you can race it yourself.

• Go disassembly: branches on arm64 vs amd64 2025-11-16

  A tiny branch in Go then disassembled on arm64 and amd64. We'll look at the key instructions and micro-benchmark the cost.

• RSA vs ECDSA in CoreDNS DNSSEC plugin: The CPU Cost of Signing NXDOMAINs 2025-10-29

  On-the-fly DNSSEC signing of negative responses is dramatically cheaper with ECDSA than RSA in CoreDNS. This benchmark showcases how RSA burns 30x the amount of CPU time compared to ECDSA for the same load, confirming CoreDNS's guidance to prefer ECDSA.