The Current Status and Future Direction of the LAVD Scheduler
summary
LAVD is a sched_ext scheduler that targets gaming latency and energy by picking CPU sets that meet a measured computing demand with minimum power. It reads kernel energy models via debugfs energy model (netlink updates landed), builds a mapping in a userspace Rust reader, and runs an energy-model-aware (EMA) algorithm to choose big/medium/little combinations. Early tests on a Snapdragon board render two games with similar or higher FPS and noticeably lower energy compared to a big-core–biased scheduler. Remaining work: model accuracy, medium-core policy, and Windows-sync preemption fixes.
AI-written summary · about
chapters (12)
- 0:00Introduction to LAVD scheduler
- 1:02What is the RAVD scheduler
- 2:48Development status of LAVD
- 3:08Energy-aware scheduling in hybrid CPUs
- 6:58Energy model implementation
- 8:20Energy model updates and optimization
- 9:43Energy model results on Snapdragon
- 10:47Task assignment in LAVD
- 12:10Early results and performance
- 13:13Future directions and discussions
- 14:17Plans for next year
- 15:23Q&A session
