Drift, Jitter, and Expectations: A Story of Kernel Timers
summary
Kernel timers in Zephyr get concrete validation here: an automated timer drift test samples many firings, reports min max and standard deviation, and enforces statistical bounds so driver regressions are caught. On Nordic hardware a 1 ms request actually yields a ~1.1 ms minimum with measurable jitter and long‑term drift; the test uses sample collections, simple math and scope traces to demonstrate the discrepancy and successfully exposes broken timer drivers.
AI-written summary · about
chapters (5)
- 0:00Start
- 0:40Realistic Kernel Timers
- 4:13Clock Jitter: Loss of precision
- 6:28Clock Drift: Loss of accuracy
- 8:14Expectations of Kernel Timers
related talks
05mLightning Talk: Micro C++ Benchmarks & Murphy's Law
01h 38mKeynote: Benchmarking - It's About Time
42mWhen the OS gets in the way
30mHow to get realistic c-state latency and residency?
27mTracking app responsiveness at scale using JankStats
49mResponding in a Timely Manner
54mChallenging idle period predictions - Where are we ?
10mDo We Need a Faster Mouse? Empirical Evaluation of Asynchronicity-Induced Jitter