simple_benchmark

A simple benchmarking package including visualization facilities.

The goal of this package is to provide a simple way to compare the performance of different approaches for different inputs and to visualize the result.

Documentation

Downloads

Installation

Using pip:

python -m pip install simple_benchmark

Or installing the most recent version directly from git:

python -m pip install git+https://github.com/MSeifert04/simple_benchmark.git

To utilize the all features of the library (for example visualization) you need to install the optional dependencies:

Or install them automatically using:

python -m pip install simple_benchmark[optional]

Getting started

Suppose you want to compare how NumPys sum and Pythons sum perform on lists of different sizes:

>>> from simple_benchmark import benchmark
>>> import numpy as np
>>> funcs = [sum, np.sum]
>>> arguments = {i: [1]*i for i in [1, 10, 100, 1000, 10000, 100000]}
>>> argument_name = 'list size'
>>> aliases = {sum: 'Python sum', np.sum: 'NumPy sum'}
>>> b = benchmark(funcs, arguments, argument_name, function_aliases=aliases)

The result can be visualized with pandas (needs to be installed):

>>> b
          Python sum  NumPy sum
1       9.640884e-08   0.000004
10      1.726930e-07   0.000004
100     7.935484e-07   0.000008
1000    7.040000e-06   0.000042
10000   6.910000e-05   0.000378
100000  6.899000e-04   0.003941

Or with matplotlib (has to be installed too):

>>> b.plot()

To save the plotted benchmark as PNG file:

>>> import matplotlib.pyplot as plt
>>> plt.savefig('sum_example.png')

Command-Line interface

Warning

The command line interface is highly experimental. It's very likely to change its API.

It's an experiment to run it as command-line tool, especially useful if you want to run it on multiple files and don't want the boilerplate.

File sum.py:

import numpy as np

def bench_sum(l, func=sum):  # <-- function name needs to start with "bench_"
   return func(l)

def bench_numpy_sum(l, func=np.sum):  # <-- using func parameter with the actual function helps
   return np.sum(l)

def args_list_length():  # <-- function providing the argument starts with "args_"
   for i in [1, 10, 100, 1000, 10000, 100000]:
      yield i, [1] * i

Then run:

$ python -m simple_benchmark sum.py sum.png

With a similar result:

Similar packages

perfplot by Nico Schlömer.