Liquorix Kernel 7.1-15 Ships With Tightened Project-C Scheduler Optimizations
The latest enthusiast kernel push targets microsecond savings in prio_balance() and context-switch hot paths.
Liquorix 7.1-15 just landed, and if you're running an AMD64 system chasing lower latency, Steven Barrett delivered a fresh scheduler tune-up. The new build delivers eight targeted patches aimed squarely at Project-C's hottest scheduling paths. No base kernel merge. No upstream feature backports. Just code that runs faster when the system makes a scheduling decision.
Liquorix has never been a kernel for people who want a quiet background process. It's been an enthusiast-grade drop-in replacement for over a decade, trading power efficiency for predictable frame times and snappier desktop interactivity. Gamers, audio engineers, and anyone who has ever flinched at a dropped sample have relied on it. This release continues that tradition, though Barrett chose to refine existing paths instead of chasing new upstream merges.
What actually changed
The changelog is tight. Eight patches, all touching Project-C's internal sched/alt paths. The work focuses on prio_balance(), sched_update_rq_clock(), and a handful of hot-path decisions that run on virtually every context switch.
Barrett replaces expensive bit operations with plain cpumask_clear() calls on per-CPU masks. Condition checks now evaluate early-exit paths before touching topology queries. Dead code gets stripped from hrtick_schedule_exit() and the sibling idle checks that run when a whole core goes quiet. The most important tweak is probably the cheap early return in sched_update_rq_clock(). That function runs on every scheduler tick and every context switch. Make it cheaper and you trim latency across the board.
Priority updates now route through the policy seam instead of raw field writes, which keeps behavior consistent when switching between fair, real-time, and deadline modes. A dead policy hook and an unused macro parameter finally get the trash can. The patches sound granular because they are.
The Project-C story keeps writing itself
Project-C (completely unrelated to NVIDIA's AI codename sharing the same letters) has been Barrett's pet project for over two years. It's an alternative to CFS designed for low latency and predictable performance. The scheduler has evolved from early PDS roots and a 1000Hz tick into a hardened build that actually ships in mainstream packaging. The previous release, 7.0-18, fixed a use-after-free in the active balance task. 7.1-15 marks the first major optimization sprint since that stability patch landed, which says something about the project's current cadence.
It's a rather focused release for what is essentially a micro-optimization pass. You won't see it in a synthetic benchmark headline. But if you're running a system that wakes and sleeps hundreds of tasks per minute, those microsecond savings compound. Audio workstations with heavy thread counts, gaming rigs under load, and desktops running dozens of background services will all feel the difference in tail latency. Keep in mind that you're trading power efficiency for responsiveness. The kernel intentionally burns more watts to keep your UI snappy and your audio stream glitch-free.
AMD64 remains the only supported architecture. Packages ship as .deb files for Debian Stable, Testing, and Unstable, with an Ubuntu PPA available if that's your distro of choice. Arch users can grab it straight from the AUR. The standard one-liner still works:
curl -s 'https://liquorix.net/install-liquorix.sh' | sudo bash
If you're already on a Liquorix kernel, the upgrade is essentially a package manager away. The install script handles bootloader setup and dependency resolution automatically. The project stays strictly within GPL-2.0, and every commit is GPG-signed.
