Slide 9
Slide 9 text
автор tg: @PyotrAndreev
Future 2025+
CPython
Faster CPython: ideas 5x in 4 years: became ≈50% faster
PEP 659: Specializing Adaptive Interpreter
PEP 734: Multiple Interpreters in the Stdlib
PEP 744: JIT Compilation
Mark Shannon, Brandt Bucher,
Irit Katriel, Eric Snow,
Michael Droettboom, Guido van Rossum
Ken Jin, Savannah Bailey
Free-threading:
PEP 703: Making the GIL Optional
PEP 779: Criteria for supported status for ft
Phase III: default ft [to be continued …]
16 June: officially supported (3.14b3+):
🗹 all core features work: threading, asyncio, … 🗹 passed all CPython tests
🗹 support: NumPy, Cython, Pandas, PyTorch, … 🗹 support the same OS & CPU-archs
🗹 single-T overhead v/s GIL: -15% perf., -20% memory 🗹 ...
Sam Gross, Matt Page,
Dino Viehland, Thomas Wouters
PyData core team: UX, ecosystem:
supports: NumPy, Numba, Cython, scikit-learn, …
PEP 617: New parser PEG/AST LL(1)/CST: PyCon
PEP 701: Improve f-strings
gather ft-info; improve: ft, bytecode, …
does not support
Faster CPython will be maintained
by a Performance Council?
ft compatibility: Status Tracking / Wheels / Checker
Labs Team
9