Unified Boot Time Measurement: A Precise, Scalable Framework for Multi-Processor Systems
summary
A vendor-agnostic framework that standardizes boot-stage records across cores, uses hardware timers to generate synchronized timestamps, and writes boot markers to a reserved DDR memory region that a userspace utility parses via devmem to produce merged timelines. Implemented on AM62x and BeaglePlay, it outputs console and HTML reports, shows SPL→U-Boot→kernel→MCU durations, and already reports up to first userspace. Current limits: static reserved memory, single-master support, and time‑synchronization issues for cascaded boots; plans include dynamic allocation and better clock handling.
AI-written summary · about
chapters (17)
- 0:00Introduction and speakers
- 0:30Problem statement and goals
- 6:27System boot architecture
- 7:45Boot measurement challenges
- 10:53Existing and proposed methods
- 16:43System-wide design approach
- 22:53Record propagation implementation
- 26:23Non-Linux component integration
- 28:06Measurement and report workflow
- 29:06Parsing and reporting tools
- 33:00Console and graphical demos
- 35:01Graphical interface overview
- 35:16Current status and community
- 38:37Future roadmap and references
- 39:30Credits and acknowledgement
- 40:32Q&A session
- 41:36Discussion on tracing methods
related talks
41mBoot-Time BOF
24mSoftware Optimization for Early Linux Boot for Time Critical...
21mAccelerating Boot Time: Lessons from a New Platform
39mTiming Boot Time Reduction Techniques
37mAccelerating Linux Boot Time: Techniques and Strategies for Optimal Performance
33mU-Boot's New Standard Boot and What's Next
42mImproving Linux Boot Time for Automotive Use Cases
49mEmbedded Linux BOF
