CPython’s Memory Engine: Allocation, Deallocation, and Free-List Tricks
How does CPython really handle memory?
This talk follows an object’s full journey: allocation, refcounting, deallocation, freelists, and allocator swaps. You’ll see why memory seems to stick, how pymalloc and mimalloc differ, and what actually happens in hot loops. Leave with a sharp mental model for debugging performance, fragmentation, and apparent leaks.
このトピックに関するあなた自身の経験を教えてください
- I teach deep Python/performance and translate hardware limits (caches, SIMD, SMT, memory bandwidth) into concrete coding decisions.
- I have a strong physics background, hands-on experience with lithography, and I follow the semiconductor industry closely—incl. attending semiconductor conferences.
- I’m comfortable with benchmark-driven explanations and profiling workflows (allocator behavior, churn, fragmentation, “leak” illusions).
My students and I contribute to CPython, speak at many conf
この発表を通じて、参加者とどんな議論が可能ですか？
About de/allocation pipeline, an intuition of the inner engine and what freelists and alternative allocators are really doing.
Specializes in CPython internals, optimization, and high-performance computing.
Driven by GPU acceleration, CPU vectorization. Evolved from ML systems to CPython core research engineer.
8+ years leading teams in AI, maths, and physics. PyCon speaker.
Lecturer at Moscow Institute of Physics and Technology – top 1 Russian university.