Lock Free Multi Core Performance with Behavior Oriented Concurrency
summary
Claims a lockless, deadlock-free model that gives real multi‑core speed on CPython 3.12+ by scheduling decorated functions across sub‑interpreters; implemented as the boc library and demonstrated with a Boids demo and a 14‑core benchmark that shows near‑linear scaling until OS contention (good scaling through eight cores). Shared state uses concurrent ownership cones serialized via XIData (serialization is the current bottleneck). Behaviors attach with the when decorator and the approach works end‑to‑end.
AI-written summary · about
chapters (10)
- 0:00Behavior Oriented Concurrency intro
- 3:36Concurrent programming challenges
- 8:13Introducing Behavior Oriented Concurrency concepts
- 12:06Dining philosophers with BAC
- 15:12Concurrent ownership internals
- 17:16Behavior scheduling patterns
- 18:59BAC limitations and future
- 22:23BAC module performance
- 23:29Boids simulation demo
- 24:51Q&A on BAC and free threading
related talks
35mWriting robust, readable, and maintainable concurrent programs in Python
31mPython Multithreading and Multiprocessing: Concurrency and Parallelism
39mUnlocking the Parallel Universe: Subinterpreters and Free-Threading in...
30mParallel computing in Python: Current state and recent advances
44mWorking Around the GIL with asyncio
30mBuilding a NoGIL Load Balancer in 30 minutes
01h 29mStandardised Coroutine Concurrency in Python 3
18mBuilding Lightning-Fast Apps With asyncio