perfbench 0.2.0

perfbench is a perfomance benchmarking module for Python code.

License Build Status PyPI version

perfbench

About

perfbench is a perfomance benchmarking module for Python code.

Feature

Compatibility

perfbench works with Python 3.3 or higher.

Dependencies

• ipython(6.0.0 or higher.)

• tqdm(4.6.1 or higher.)

• plotly(2.7.0 or lower)

Installation

pip install perfbench

Usage

plotting a single figure.

import numpy as np
from perfbench.process import *


setups = [
    {'func': lambda n: np.random.uniform(low=-1., high=1., size=n).astype(np.float64), 'title': 'float64'}
]

kernels = [
    {'func': lambda x: np.around(x), 'label': 'around'},
    {'func': lambda x: np.rint(x), 'label': 'rint'}
]

ntimes = [2 ** n for n in range(15)]

bm = Benchmark(
    setups=setups,
    kernels=kernels,
    ntimes=ntimes,
    xlabel='samples',
    title='test',
    logx=True
)
bm.run()
bm.plot()

plot1

plotting multiple figures.

import numpy as np
from perfbench.process import *


setups = [
    {'func': lambda n: np.random.uniform(low=-1., high=1., size=n).astype(np.float16), 'title': 'float16'},
    {'func': lambda n: np.random.uniform(low=-1., high=1., size=n).astype(np.float32), 'title': 'float32'},
    {'func': lambda n: np.random.uniform(low=-1., high=1., size=n).astype(np.float64), 'title': 'float64'}
]

kernels = [
    {'func': lambda x: np.around(x), 'label': 'around'},
    {'func': lambda x: np.rint(x), 'label': 'rint'}
]

ntimes = [2 ** n for n in range(5)]

bm = Benchmark(
    setups=setups,
    kernels=kernels,
    ntimes=ntimes,
    xlabel='samples',
    title='test',
    logx=True
)
bm.run()
bm.plot()

plot2

License

This software is released under the MIT License, see LICENSE.