Downloaded 12 times
Uploaded bySpeedyCloud
How to Use Multi-thread & Multi-process in Python
This document discusses multi-threading and multi-processing in Python. It explains that the Global Interpreter Lock (GIL) in Python prevents true concurrency between threads by serializing threads. For I/O-bound tasks, multi-threading can improve performance by allowing other threads to run while one thread is blocked on I/O. However, for CPU-bound tasks, multi-threading may not help or even hurt performance due to the GIL. It recommends using the multiprocessing module or ctypes for CPU-bound parallelism instead of threads. It also covers related topics like joining processes, file descriptors, and improving multi-threaded performance.
How to Use Multi-thread & Multi-process in Python
- 1. 北京迅达云成科技有限公司 www.speedycloud.cn 北京迅达云成科技有限公司 www.speedycloud.cn 2015年9月 北京 WELCOME Howto Use Multi-thread & Multi-process in Python
- 2. 北京迅达云成科技有限公司 www.speedycloud.cn Why usingmulti-thread? Productivity
- 3.
- 4.
- 5. 北京迅达云成科技有限公司 www.speedycloud.cn What isGIL？ In C-python, the global interpreter lock, or GIL, is a mutex that prevents multiple native threads from executing Python byte-codes at once. This lock is necessary mainly because C-Python’s memory management is not thread-safe.(However, since the GIL exists, other features have grown to depend on the guarantees that it enforces.) GIL will serialize threads and stop the concurrence of threads.
- 6.
- 7. 北京迅达云成科技有限公司 www.speedycloud.cn Python multi-thread,no use at all? NO!
- 8. 北京迅达云成科技有限公司 www.speedycloud.cn The GoodNews… I/O Intensive：
- 9. 北京迅达云成科技有限公司 www.speedycloud.cn IO IntensiveRunning Sequence
- 10. 北京迅达云成科技有限公司 www.speedycloud.cn Unfortunately… CPU IntensiveTask: Multi-thread’s productivity even worse than single thread
- 11. 北京迅达云成科技有限公司 www.speedycloud.cn How todeal with it? First: Ctypes
- 12.
- 13. 北京迅达云成科技有限公司 www.speedycloud.cn Been there,done that… Join()
- 14. 北京迅达云成科技有限公司 www.speedycloud.cn file discripter InLinux, all operations on all devices and/or files are done via file descriptor. Been there, done that…
- 15.
- 16. 北京迅达云成科技有限公司 www.speedycloud.cn Thanks! Beijing SpeedyCloudTechnology Co., Ltd., www.speedycloud.cn