How two undergrads from the other side of the planet are speeding up...
summary
CPython 3.13 ships an experimental optimizer that turns specialized bytecode traces into bigger-region optimizations, with globals-to-constants and partial guard elimination already implemented and heavier passes planned for 3.14. The pipeline produces traces up to ~800 bytecode instructions, uses an abstract interpreter for type/constant propagation, and feeds a JIT; early results show ~30% reduction in function-call overhead and single-digit wins from top-of-stack caching, but the JIT is off by default and not yet production-ready.
AI-written summary · about
chapters (4)
related talks
01h 03mA tour of CPython's runtime
29mWhat they don't tell you about building a JIT compiler for CPython
55mMaking CPython Fast Using Trace-based Optimisations
32mWriting performant code for modern Python interpreters
25mHow two teams are working together to make Python better for all
38mDynamic Code Instrumentation with Hacked Interpreters
23mInside CPython 3.11's new specializing, adaptive interpreter
41mHow to write a JIT compiler in 30 minutes