Improving NumPy Performance on Free-Threaded Python
Summary
Christopher Trudeau joins the host to discuss Python 3.15 RC, Django 6.1, and new PyPI security measures. They cover PEPs 837 and 842, Adam Johnson’s Django-Crawl tool, and an article on pure Python SIMD optimizations. The episode also features a course on managing imports with `__all__` and details how NumPy is scaling on free-threaded Python by resolving hidden locking bottlenecks.
Topics discussed
Intro: NumPy scaling, PyCoders Weekly, and Modern OOP book
Python 3.15 RC and Django 6.1 release notes
PyPI security update: 14-day file update restriction
PEP 837: Extensible JSON Serialization and __dunder_export__
Django Crawl: Tool for auditing Content Security Policy headers
SIMD in Pure Python: Bitwise operations for vector math
Managing Wildcard Imports with __all__ and namespace control
Promo: Modern Object-Oriented Python book
NumPy performance on free-threaded Python: Lock-free paths
CNS Plots: Publication-ready scientific data visualization
ORM Lens: Visualizing Django models and schema changes
Outro and thanks to Christopher Trudeau
Listen ad-free on Castria