Steps Towards a Gaming-Optimized Scheduler
summary
Proposal shows a working sched_ext prototype called LAVD that targets Steam Deck/Wine+Proton stacks and claims measurable reduction in stutter by prioritizing latency-critical chains rather than raw throughput. Analysis uses a custom benchmark VaporMark plus perfetto traces, finds games form 20–30 correlated tasks with many runtimes under 100 µs and a minority >1 ms, and shows average FPS rises are less important than improving low 1% (tail) FPS. Key gaps: flaky benchmarks, tracing rare spikes, futex/ownership visibility and CPU/core placement tradeoffs.
AI-written summary · about
chapters (7)
- 0:00Introduction and motivation
- 2:56Key challenges in development
- 9:18Workload analysis insights
- 17:13Input responsiveness in gaming
- 23:34Benchmarking difficulties
- 31:19Windows games synchronization
- 34:55Future directions and Q&A
related talks
46mOptimizing Scheduler for Linux Gaming
25mLinux needs a Scheduler QOS API -- and it isn't nice(2)
21mPreview: Operating Systems
48mAn Unbiased Look at the Energy Aware Scheduler (EAS)
50mSCHED_DEADLINE: It's Alive!
44mRestartable Sequences: Scheduler-Aware Scaling of Memory Use on Many-Core Systems
30mScheduling at Scale: eBPF Schedulers with Sched_ext
52mEffectively Measure and Reduce Kernel Latencies for Real-time Constraints