Tuning Python applications can dramatically increase performance
summary
An experimental profiler claims line-level hotspots with typically under 15% overhead, no source changes, and the ability to attach to live processes. It contrasts with cProfile and Python Tools (function-level) and line_profiler (high overhead). Real examples show speedups of 2x or more — a page from ~90s to ~28s and a report from ~300s to ~2s — and it supports Windows and Linux, 32/64-bit, plus remote SSH collection.
AI-written summary · about
chapters (9)
- 0:00Introduction and agenda
- 2:38Why optimize Python code
- 6:04Overview of profiling types
- 8:08Existing Python tools
- 10:56Demonstration of a slow example
- 16:14Live demo of the new profiler
- 22:04Prototype capabilities and results
- 27:30Tool availability and Q&A
- 28:31Technical Q&A
related talks
01h 31mDebugging and profiling techniques
32mProfiling at the speed of light
30mHow to Tune your Python Analysis Pipeline: A Profiler Guide
22mExperiences making CPU-bound tasks run much faster
24mPractical Profiling
28m"Performant Python"
30mWrite faster Python! Common performance anti patterns
01h 03mA tour of CPython's runtime