Concurrency in Python

Mini PyCon MY 2014
March 2014, Kuala Lumpur

Transcript

1. Concurrency In Python Farzad Ghanei www.ghanei.net
2. != Parallel computing Concurrency == Handling tasks while others are still running
3. Blocked by IO Why? Processing takes long
4. Threads Solutions Processes Coroutines
5. Code Outline
6. Simple, sync IO API Threads Preemptive scheduling by OS Unpredictable execution ordering
7. IO bound threads
9. CPU bound threads
10. CPU bound threads
12. Threads release and acquire GIL Global Interpreter Lock The one thread that has GIL runs at a time
13. 1 CPU bound thread
14. 3 CPU bound threads
15. Not good for CPU bound Threads Good for user Interactivity, maybe IO Shared state might cause difficulties
16. Simple, sync IO API Processes Preemptive scheduling by OS Unpredictable execution ordering
17. IO bound processes
18. CPU bound processes
19. CPU bound processes
21. Processes Good for parallel processing Use more resources than threads Shared state might cause difficulties
22. Predictable execution ordering Coroutines Cooperative concurrency
23. PEP 380 (yield from) Coroutines Generators (yield) Greenlets (green threads)
24. CPU bound greenlets
25. CPU bound greenlets
27. Coroutines Have early results Controlling execution
28. Async IO No blocking calls Respond to IO Events
29. Event driven concurrency Callback based (Twisted) Coroutine based (Gevent) Both (asyncio)
30. Gevent Implicit coroutines (using greenlets) threading like API Monkey patching libev event loop
31. Gevent
33. Twisted Explicit event loop Batteries included Deferreds No blocking calls
34. Twisted
36. asyncio (tulip) PEP 3156 (asyncio) Callbacks and coroutines API Standard library Explicit event loop
37. Thanks