Measuring Carbon Footprint of Personal Computing
summary
Proposes a kernel module that estimates per‑process energy by combining multivariate regression—trained via device sweeps that turn components on one at a time—with runtime metrics from kernel process accounting (/proc) to output joules per process. Claims it runs lightweight (no neural nets), targets upstreaming to Linux, shows a mock UI and IDE integration, and notes validation requires spot checks with external meters; argues powertop’s discrete samples and vendor tricks are insufficient.
AI-written summary · about
chapters (17)
- 0:00Start
- 0:39Brief Introduction
- 0:53Outline
- 1:16Background
- 2:06Energy Consumption
- 4:48Hardware power attribution
- 6:36Windows Energy Estimation Engine (E3)
- 7:04Windows Task Manager Integration
- 7:19macOS
- 8:17Power TOP
- 8:55PowerTOP: The solution?
- 10:56Device-Specific Measurements
- 14:11Kernel Process Accounting Infrastructure
- 15:28End-users
- 16:37Programmers
- 17:15Conclusion
- 19:59System Design
related talks
42mPower Measurement and Attribution in Kernel for Embedded System Processes
37mPower Measurement and Attribution for Processes and Hardware Devices in the Linux...
50mResponsible computing: how about a survey of measurement tools?
34mCarbon Footprint–Aware Software Development
20mMonitoring Energy Consumption and Building Energy Efficient Systems the Cloud Native Way
31mTrack your code's CO2 emissions with Code Carbon
44mCloud computing and carbon emissions
24mEnergy model accuracy