Dynamic Energy Model to handle leakage power
summary
A runtime-swappable energy model is implemented for EAS using an RCU pointer so a vendor kernel module can replace per-frequency power and cost arrays without exposing a userspace API; updating the whole array takes ~10 µs at 1 GHz. Measurements show GPU heating raising CPU temperature ~20°C and increasing big-core leakage by +15–18.5% while mid cores change +5–8%, creating an overlap where task placement should prefer mid cores; RFC forthcoming.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 0:21Current energy model issues
- 2:11Leakage power and temperature impact
- 6:54Runtime energy model adjustment
- 10:48Implementation and future plans
- 11:11Q&A discussion
- 15:02Closing and next session
related talks
20mUpdating Energy Model from Thermal
48mAn Unbiased Look at the Energy Aware Scheduler (EAS)
01h 11mPerformance engineering - being friendly to your hardware
27mMeasuring Carbon Footprint of Personal Computing
37mPower Measurement and Attribution for Processes and Hardware Devices in the Linux...
42mPower Measurement and Attribution in Kernel for Embedded System Processes
32mCPU Power Saving Methods for Real-time Workloads
27mRefactor Power Library for Vendor Agnostic Uncore APIs