benchmarkme

Ergonomic Python interface to the built-in timeit utility for quick-look profiling.

Out-Of-The-Box Features

• Python API with enhanced reporting
• Yields configuration, best & worst times and processing rates
• Default autoranging
• Automatic looping & repeat
• Customizable environment

Usage

Minimal example

>>> from time import sleep
>>> from benchmarkme import BenchMark
>>> b = BenchMark(f=sleep, fargs=[0.2,])
>>> b() # call the instance directly to run the main utility
Benchmark Results
-----------------
* Ran 1 trials 5 times.
* Time of one independent trial: 200 msec
* Best time of subsequent trials: 200 msec
* Worst time of subsequent trials: 200 msec
* Best speed: 5 items / sec
* Worst speed: 4.99 items / sec

Provide the items processed to get rate in the desired units (from benchmarkme/examples/basic.py)

import math
from benchmarkme import BenchMark

def fmath(x: list) -> None:
    """Calculate somethig complex-looking to time."""        
    return [math.log(math.acos(item / 1e5)) ** 2 / 1733 for item in x]

items = 100000
bmrk_math = BenchMark(
    f=fmath, 
    fargs=[[*range(items)],], 
    item_units='numbers', 
    items_processed=items
)

if __name__ == "__main__":
    bmrk_math()

You'll get something like the following, i.e. the function computes around 10 million numbers per second on this machine.

Benchmark Results
-----------------
* Ran 50 trials 5 times.
* Time of one independent trial: 8.92 msec
* Best time of subsequent trials: 8.89 msec
* Worst time of subsequent trials: 12.1 msec
* Best speed: 11.2 Mnumbers / sec
* Worst speed: 8.28 Mnumbers / sec