Off road profiling - when the automated profilers just don't cut it.
summary
Automatic profilers slow a 9‑hour CI job enough to make iteration impossible, so a lightweight manual profiler is built instead: manprof inserts time.time snapshots at checkpoints, accumulates per-checkpoint totals and a nested-total counter to prevent double counting, and computes net time as accumulated minus nested. It supports multiple independent profiler instances, decorator-based checkpointing, and exports to YAML/JSON. It handles short hot functions called frequently and phase‑specific buckets; results print per‑checkpoint gross and net seconds.
AI-written summary · about
chapters (8)
- 0:00Introduction and speaker background
- 1:12Why use manual profiling
- 2:36Limitations of automatic profilers
- 5:03Manual timing with time.time
- 10:04Handling nested function calls
- 14:43Introducing manprof library
- 16:12Using manprof checkpoints
- 21:55Q&A and future plans
related talks
24mPractical Profiling
37mTuning Python applications can dramatically increase performance
30mHow to Tune your Python Analysis Pipeline: A Profiler Guide
01h 31mDebugging and profiling techniques
32mProfiling at the speed of light
02h 33mMeasure, don't Guess - How to find out if and where to optimize
01h 33mPower up your work with compiling and profiling
22mExperiences making CPU-bound tasks run much faster