Victor Stinner

Hi, > I've done a little experiment: I run timeit.repeat 1000 times on a really simple statement Are you talking about the timeit module of the Python standard library? You...

> WARNING: the benchmark result may be unstable > * the standard deviation (3.53 ns) is 15% of the mean (23.5 ns) > * the maximum (50.2 ns) is 114%...

The pyperf API gives you access to all values. You can use min(), max(), median(), etc. You're free to ignore outliers. But I don't want to change the default: mean...

"system tune" depends a lot on the platform. WSL is currently not supported. Patches are welcomed!

I'm not really excited by the idea of moving the venv feature from pyperformance into pyperf. Can't you use another tool to setup a venv to run benchmarks? I would...

Hi. Which API do you want to specify which clock should be used? A parameter to Runner()? A parameter to `bench*()` functions? Both?

Can't you avoid importing the code which uses click in your benchmark? Or maybe mock sys.argv when it's imported. Why does your code parse command line arguments even if it's...

pyperf spawns multiple processes and it uses the command line to pass arguments to worker processes: https://pyperf.readthedocs.io/en/latest/run_benchmark.html#pyperf-architecture You should avoid using click and pyperf in the same file.

Can't you move your benchmark code into a separated script, and import your functions in the benchmark script? Something like that: https://pyperf.readthedocs.io/en/latest/examples.html#bench-func-method Maybe pyperf should have an option to not...

Is there someone interested to implement this idea?