Introduction
For gamers running Linux on older or heavily loaded systems, a promising kernel patch could soon deliver noticeable performance improvements. Developer Peter Zijlstra has released the second version of a patch series called "sched: Flatten the pick" aimed at correcting inefficiencies in how the Linux scheduler handles cgroup (control group) scheduling. When multiple tasks compete for CPU time, the current scheduler can inadvertently fragment a cgroup's total priority weight across all available cores, dramatically reducing the effective priority of critical applications like games. This patch seeks to fix that fragmentation, especially beneficial on machines with many cores or under heavy multitasking loads.
The Problem: Cgroup Scheduling Fragmentation
Zijlstra described the issue as "a pain in the ar*e" and traced it to the way a cgroup's total weight is distributed among every CPU in the system. On a 64-core machine, each cgroup effectively gets the priority of a task at nice +19 per CPU. On a 256-core server (not uncommon today), that margin shrinks even further. The problem has two main parts:
- Weight inflation: A typical workaround is to inflate the group weight by the number of CPUs, but if a group‘s entire load suddenly lands on a single core, that weight skyrockets past nice -20, breaking mathematical assumptions in the scheduler.
- Task selection complexity: The scheduler currently traverses multiple cgroup levels to pick the next task to run. This adds overhead and can delay decisions.
The Fix: Flattening the Pick
The “Flatten the pick” patch addresses both issues. For the weight inflation problem, it modifies the computation so that group weight doesn’t balloon artificially. For task selection, it collapses the multi‑level traversal into a single level, reducing overhead and making scheduling decisions faster. This streamlined approach is expected to improve responsiveness especially in scenarios where many background processes compete with a foreground game.
Experimental Results on an Old Rig
To validate the patch, Zijlstra conducted a stress test using an aging setup: an Intel Core i7-2600K (old hardware) and an AMD Radeon RX 580. He launched Shadows: Awakening from GOG via Lutris, using GE-Proton10-34 and Steam Runtime 3 (sniper). To simulate a typical multi‑tasking environment, he added 8 spinner processes (one per CPU thread) in the background—similar to having Discord, a browser with many tabs, and a system update running concurrently.
With the default scheduler slice, the game went from playable to “almost unplayable, as in proper terrible.” He then repeated the test with a shorter scheduler time slice (set to one‑tenth of the default via chrt) and recorded both sessions using MangoHud. The results are striking:
| Metric | Default slice | Shorter slice (patch active) |
|---|---|---|
| FPS min | 3.8 | 20.6 |
| FPS avg | 48.0 | 57.2 |
| FPS max | 87.4 | 80.3 |
| Frame time min (ms) | 9.4 | 8.4 |
| Frame time avg (ms) | 34.5 | 19.5 |
| Frame time max (ms) | 107.4 | 37.2 |
The shorter slice (with the flattened pick) roughly quadrupled the minimum FPS and halved the average frame time, turning a stuttering mess into a smooth experience. Zijlstra noted that he did not compare against a kernel without the flat patch; his goal was to ensure the patch worked under non‑trivial workloads.
Implications for Everyday Gaming
If you game on Linux, especially on older hardware, this fix could be transformative. The stress test mirrors real‑world usage: a game fighting for CPU time with Discord, a browser, system updates, and other background tasks. The scheduler inefficiency is not limited to old rigs; high core counts actually worsen the weight fragmentation issue. Modern systems with 16 or more cores could also benefit—the problem scales with core count, so fixing it helps everyone.
Moreover, the patch’s approach of flattening the pick reduces scheduling overhead, which can improve latency and responsiveness across the board, not just in gaming. Users running virtual machines, compilation jobs, or any workload where multiple cgroups compete should also see benefits.
Future Outlook
As of now, the patch series has not been merged into the mainline Linux kernel. It still requires review from the relevant kernel maintainers, and revisions are likely before it lands in a stable release. However, the positive results from Zijlstra’s testing suggest it has strong potential. Enthusiasts can watch for its inclusion in future kernel versions (e.g., 6.x or later) or test it via custom kernels if they wish to experiment early.
For now, the “sched: Flatten the pick” patch is a promising development that could breathe new life into older gaming machines and improve performance on high‑core systems alike. Stay tuned to Linux kernel mailing lists and Phoronix for updates.