Skip to content
Toolkit to run Python benchmarks
Python
Branch: master
Clone or download

README.rst

pyperf

Latest release on the Python Cheeseshop (PyPI) Build status of pyperf on Travis CI

The Python pyperf module is a toolkit to write, run and analyze benchmarks.

Features

  • Simple API to run reliable benchmarks
  • Automatically calibrate a benchmark for a time budget.
  • Spawn multiple worker processes.
  • Compute the mean and standard deviation.
  • Detect if a benchmark result seems unstable.
  • JSON format to store benchmark results.
  • Support multiple units: seconds, bytes and integer.

Usage

To run a benchmark use the pyperf timeit command (result written into bench.json):

$ python3 -m pyperf timeit '[1,2]*1000' -o bench.json
.....................
Mean +- std dev: 4.22 us +- 0.08 us

Or write a benchmark script bench.py:

#!/usr/bin/env python3
import pyperf

runner = pyperf.Runner()
runner.timeit(name="sort a sorted list",
              stmt="sorted(s, key=f)",
              setup="f = lambda x: x; s = list(range(1000))")

See the API docs for full details on the timeit function and the Runner class. To run the script and dump the results into a file named bench.json:

$ python3 bench.py -o bench.json

To analyze benchmark results use the pyperf stats command:

$ python3 -m pyperf stats bench.json
Total duration: 29.2 sec
Start date: 2016-10-21 03:14:19
End date: 2016-10-21 03:14:53
Raw value minimum: 177 ms
Raw value maximum: 183 ms

Number of calibration run: 1
Number of run with values: 40
Total number of run: 41

Number of warmup per run: 1
Number of value per run: 3
Loop iterations per value: 8
Total number of values: 120

Minimum:         22.1 ms
Median +- MAD:   22.5 ms +- 0.1 ms
Mean +- std dev: 22.5 ms +- 0.2 ms
Maximum:         22.9 ms

  0th percentile: 22.1 ms (-2% of the mean) -- minimum
  5th percentile: 22.3 ms (-1% of the mean)
 25th percentile: 22.4 ms (-1% of the mean) -- Q1
 50th percentile: 22.5 ms (-0% of the mean) -- median
 75th percentile: 22.7 ms (+1% of the mean) -- Q3
 95th percentile: 22.9 ms (+2% of the mean)
100th percentile: 22.9 ms (+2% of the mean) -- maximum

Number of outlier (out of 22.0 ms..23.0 ms): 0

There's also:

  • pyperf compare_to command tests if a difference is significant. It supports comparison between multiple benchmark suites (made of multiple benchmarks)

    $ python3 -m pyperf compare_to py2.json py3.json --table
    +-----------+---------+------------------------------+
    | Benchmark | py2     | py3                          |
    +===========+=========+==============================+
    | timeit    | 4.70 us | 4.22 us: 1.11x faster (-10%) |
    +-----------+---------+------------------------------+
    
  • pyperf system tune command to tune your system to run stable benchmarks.

  • Automatically collect metadata on the computer and the benchmark: use the pyperf metadata command to display them, or the pyperf collect_metadata command to manually collect them.

  • --track-memory and --tracemalloc options to track the memory usage of a benchmark.

Quick Links

Command to install pyperf on Python 3:

python3 -m pip install pyperf

pyperf supports Python 2.7 and Python 3. It is distributed under the MIT license.

You can’t perform that action at this time.