I went exploring parallelism in Python.

Below are some comparisons benchmarking performance of different parallel programming approaches for solving arbitrary I/O and compute tasks. The tasks vary between approaches (i.e. between async/await and multiprocessing) and are not directly comparable.

Results:

Async await
   Slower approach: Finished 'no_async' in 5.0055s
   Faster approach: Finished 'yes_async' in 1.0057s
   Speed improvement 4.98x

Multiprocessing
   Slower approach: Finished 'no_multiprocessing' in 2.1139s
   Faster approach: Finished 'yes_multiprocessing' in 0.1232s
   Speed improvement 17.16x

Multiprocessing with execution pools
   Slower approach: Finished 'no_multiprocessing' in 2.1328s
   Faster approach: Finished 'yes_multiprocessing' in 0.1227s
   Speed improvement 17.39x

Threading
   Slower approach: Finished 'no_threading' in 4.0146s
   Faster approach: Finished 'yes_threading' in 1.0011s
   Speed improvement 4.01x

Threading with execution pools
   Slower approach: Finished 'no_threading_execution_pool' in 4.0097s
   Faster approach: Finished 'yes_threading_execution_pool' in 1.0066s
   Speed improvement 3.98x

Note: The speed improvement multipliers will typically be even larger for larger amounts of work, as the "spin-up" time of each approach will make up a smaller fraction of the total work/waiting time. This is a blunt benchmark.

To reproduce:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python src/demos/main.py