Benchmarks

“Statistics are like bikinis. What they reveal is suggestive, but what they conceal is vital.”

—Aaron Levenstein, 1951

When used as a REPL (Read-Eval-Print Loop), this module starts slow since it builds on the standard library asyncio, which loads all its submodules on import, and the event loop is quite difficult to implement.

The figures below are obtained by running each of the following sequentially with no warmup elevenfold in a fresh console session and discarding the first run, which is considered a warmup run:

Environment

Benchmarked on an old HP laptop.

• python -VV gives Python 3.15.2 (main, Oct 26 2026, 14:23:21) [MSC v.1951 64 bit (AMD64)]

• python -m platform gives Windows-11-10.0.26300-SP0

• __pycache__ directories are persisted across runs

Baseline: asyncio

python -SIqX importtime -c "import asyncio"

Cumulative import time of asyncio: 122.60 ± 10.14 ms; max 138.83 ms, min 103.49 ms; n = 10

Note

Above, 122.60 ms is the mean and 10.14 ms is the standard deviation.

asyncutils

python -SIqX importtime -c "import asyncutils"

Cumulative import time of asyncutils: 147.34 ± 7.53 ms; max 156.40 ms, min 131.94 ms; n = 10

Note

The figures below are relative to the time when asyncutils/__init__.py is executed, and may not reflect the actual time taken, because Python is not free to boot up itself, having to perform various initialization tasks.

python -SIqm asyncutils -dl

Time taken to start the console, which includes importing asyncio: 99.89 ± 6.19 ms; max 110.25 ms, min 91.97 ms, n = 10

python -SIqm asyncutils -dpl

Time taken to import asyncio along with all 31 ordinary submodules: 196.96 ± 18.30 ms; max 225.08 ms, min 172.51 ms, n = 10

Note

Up to 10 required internal submodules are also fetched.

Note

asyncio is still loaded early such that attribute accesses later on would not randomly take more than 100 ms, and you logically wouldn’t use this module without an async entry point, such that asyncio and its event loop are crucial and unavoidable.