Updating Energy Model from Thermal
summary
The Energy Model can be updated at runtime and the Thermal framework already provides SoC temperatures, so the proposal is to adjust static power (leakage) values in the EM to reflect heat-driven leakage changes; historically leakage reached ~50% (2015–16) and a rule of thumb is leakage doubles per 20°C, making leakage ~20% of total power in some workloads and enough to skew schedulers and governors. Plan: refactor DTPM and thermal governor estimation into a single power cooling device with write access to EM, add sensor hierarchy / DT info and an upstream tuning module; interface must be privileged to avoid hardware risk.
AI-written summary · about
chapters (10)
- 0:00Introduction to energy model
- 1:08Current energy model usage
- 3:15Proposal for temperature-based updates
- 7:22Safety and user-space risks
- 9:12Critique of energy model updates
- 12:28Downstream tunings and upstream challenges
- 13:34Two-track enablement approach
- 17:13Refactoring duplicated code
- 18:38PMQoS and thermal governor coordination
- 19:20Closing remarks
related talks
16mDynamic Energy Model to handle leakage power
22mA generic energy model description
16mCombining DTPM with the thermal control framework
24mEnergy model accuracy
34mImprovisation and Demonstration of Linux Thermal Framework for Multiple Temperature Sensors
42mPower Measurement and Attribution in Kernel for Embedded System Processes
19msystem pressure on CPUs capacity and feedback to scheduler
37mPower Measurement and Attribution for Processes and Hardware Devices in the Linux...