November 6, 2025
Seven seconds, infinite comments
The seven second kernel compile
Old 7‑second Linux build resurfaces and the comments go wild: nostalgia, nitpicks, and nerd flexes
TLDR: A 2002 article about a seven‑second Linux build on a 32‑CPU IBM server is back, sparking debates over missing dates, modern compile times, and whether code growth beats hardware gains. Nostalgia stories and jokes stole the show, highlighting how tech feats and community culture collide.
An old IBM lab write‑up from 2002 just resurfaced bragging about a Linux kernel compiling in seven seconds on a giant 32‑processor server, and the comments section immediately set itself on fire. The piece explains how kernel devs use compile time as a quick‑and‑dirty test, how this mega‑machine hit seven seconds, and how the final “link” step couldn’t use all CPUs—translation: even monster hardware can get bottlenecked. But the community? They zeroed in on vibes and context.
The loudest voice is the “label your reposts” crowd, demanding dates for ancient tech triumphs. Another camp wants receipts: what does a kernel build take on today’s gear? Cue the spicy think piece energy—one commenter wonders if the codebase is growing so fast that faster chips can’t keep up, basically pitting Moore’s law against the ever‑ballooning Linux kernel. Meanwhile, nostalgia hits hard: confessions of skipping parties to compile and war stories of overnight builds on creaky 486 PCs had everyone laughing (and wincing). The meme‑y punchline: is this a meaningful benchmark or just a glorious nerd flex? The thread lands somewhere between “wow, engineering” and “ok but show me modern numbers,” with a heavy sprinkle of “back in my day” energy.
Key Points
- •The study targets a seven-second Linux kernel compile using the 2.5 kernel on a 32-way PowerPC64 IBM pSeries p690.
- •Martin Bligh reduced compile times on a 2.4.18 kernel from 47s to 23s with patches, exposing 2.4 series NUMA scalability issues.
- •The IBM pSeries p690 (POWER4) scales to 32-way SMP via multi-chip modules and supports logical partitioning with a hypervisor.
- •Initial 24-way p690 runs achieved a 10.3-second compile, with average utilization of 19.57 CPUs and significant system time.
- •A single-threaded final link stage limited CPU utilization, highlighting remaining serialization bottlenecks in the build process.