Meta: Enabling Programming Languages to Learn from the Crowd
summary
Meta adds a Python decorator and community DB to share functions, record runtime args, returns, types, timings, and exceptions, then uses that data for search, replacement, and synthesis. It runs dynamic equivalence tests and an optimizer that finds 44 optimizations, 1.45× median speedup and reports 7 incorrect suggestions. It offers auto-patcher swaps functions at runtime and records calls at 1/n sampling; load cost ≈47 ms, call overhead ≈0.31 ms.
AI-written summary · about
chapters (8)
- 0:00Start
- 3:22Example of a Typical Meta Function
- 6:16Load and Use an Existing Function from the Community
- 8:16Dictionary Merging
- 8:38Optimization
- 10:59Meta Functions
- 15:05Future Work
- 20:43Final Questions
related talks
25mHow two teams are working together to make Python better for all
09mAccelerating ML Prototyping The Pythonic Way
33mMetaprogramming in Python
01h 50mSpreadsheet-Driven Meta-Programming for Python
29mMetaprogramming with Decorators, Metaclasses, and Dynamic Code Generation
24mHow two undergrads from the other side of the planet are speeding up...
47mWriting Python Codemods for Fun and Profit
08mAutomatically enhancing error messages