Proactive Energy-Aware Programming with PEEK
summary
PEEK automates energy measurement, integrates with Git and Eclipse, and generates source patches that are validated by re-running analyses. Evaluation reports an 8.4-fold speed-up in energy analysis and an average 25.3% energy improvement for optimized snapshots. Hardware uses a current mirror plus flip‑flop to convert fast drain transients into a slow square wave, yielding 0.1 µJ resolution and 6 ns temporal resolution. Function-level attribution is done by toggling GPIOs at function entry and exit and correlating measurements to builds.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:27Energy-Aware Programming: Status Quo
- 3:25Proactive Energy-Aware Programming with PEEK
- 4:43Design and System Architecture
- 8:44Implementation and Operation: Energy Analysis
- 9:30Implementation and Operation: Optimizations
- 13:54Electronic Measuring Device: Motivation
- 15:15Measuring Energy: Sampling Constraints
- 16:16Measuring Energy: PEEK Energy Measurement Device
- 18:42Evaluation Scenarios
- 19:59Evaluation Platform
- 20:44Evaluation Results: Power Modes
- 22:44Evaluation Results: Compiler Impact
- 24:34Evaluation Results: Energy Optimization Hints
related talks
01h 12mPower, Energy and Analysis
36mPower Management on Zephyr
40mProgramming a 144-computer chip to minimize power
39mZephyr Power Management for Wearable Devices
13mPReGO: A Generative Methodology for Satisfying Real-Time Requirements on COTS-Based Systems: Definit
46mDevice Power Management: Journey to 5uA
40mOptimizing Zephyr for Peak Performance
16mAnnotating Executable DSLs with Energy Estimation Formulas