Multithreading vs Multiprocessing in Python

Pre-requisites: Multithreading and Multiprocessing

Remember the golden rule: multithreading for I/O bound tasks and multiprocessing for CPU bound task

Python threads vs process

Following table shows python specific details of program_processes_threads

Global Interpreter lock (GIL)

1. Ensures thread safety

2. Improves single thread performance

3. Prevents simultaneous multi-threading

4. Bad for CPU limited tasks

5. I/O limited threads are hardly affected

Outcomes

Following benchmarks were run on a 8 core / 16 threads CPU
We have measuring number of function calls, higher the better
(NOTE: below multithreading only uses 1 process and either 4/8/16 threads. Multiprocessing uses 4/8/16 processes with 1 thread each)

Conclusion

1. Python is multithreaded

• It is concurrent but not parallel
• CPython will consider switching threads every 15 ms or when an I/0 operation is encountered

1. Multi-threaded does not strictly mean single-core

• the OS may switch the python process between physical and virtual CPU cores!

1. multiprocessing is the standard python way to increase processing power if needed

2. Most numerical libraries (numpy, scipy, tensor flow) are simultaneously multi-threaded behind the scenes

3. Since threading allows process data to be accessed, data may need to be protected with Lock()

4. In multiprocessing, Pipe() and Queue() are used for processes to share data between processes

5. Pipe() may be corrupted if accessed simultaneously

• It is NOT thread safe

1. Queue() can be accessed by multiple users

• It is thread and process safe

References

1. threading vs multiprocessing in python
2. https://towardsdatascience.com/multithreading-vs-multiprocessing-in-python-3afeb73e105f