Extracting Cycles_

Cache-friendly reads, user-hostile writes.

Low-Level Performance, One Wrong Turn at a Time

Fast code is the destination. This blog is about the road there.

SWAR tricks, bit-twiddling, SIMD, branch-killing, cache-friendly layouts… the final benchmark is rarely the interesting part. The wrong turns are. The dead ends, the small intuitions, and those “wait, why does THIS make it faster?!” moments along the way.

Pick a post below, and if you think my benchmark is wrong, tell me, I’d genuinely love that :P

Latest posts

2026-10-13 IPv4 Parsing: Having Your Cake and Eating It Too

A fast IPv4 parser that doesn't read out of bounds.

2026-06-20 The IPv4 Parser AI Couldn't Have Written

A branchless SWAR IPv4 parser in Zig using pext/pdep, and why AI can't reach for tricks like it.