https://github.com/jcrist/ptime

IPython magic for parallel profiling (like `%time`, but parallel)
https://github.com/jcrist/ptime

Last synced: 6 days ago
JSON representation

IPython magic for parallel profiling (like `%time`, but parallel)

• Host: GitHub
• URL: https://github.com/jcrist/ptime
• Owner: jcrist
• License: bsd-3-clause
• Created: 2017-07-17T18:54:38.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-07-17T20:56:42.000Z (over 7 years ago)
• Last Synced: 2024-10-07T22:19:59.554Z (about 1 month ago)
• Language: Python
• Homepage:
• Size: 5.86 KB
• Stars: 71
• Watchers: 4
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE

Awesome Lists containing this project

README

        
ptime

=====

A tool for measuring serial and parallel execution, and comparing the results.

Provides an `IPython magic

`_ ``%ptime``.

This can be useful for measuring the benefits of parallelizing code, including

measuring the effect of the `Global Interpreter Lock

`_ (GIL).

Example

-------

.. code::

    In [1]: %load_ext ptime

    In [2]: import numpy as np

    In [3]: x = np.ones((5000, 10000))

    In [4]: %ptime x + x

    Total serial time:   0.42 s

    Total parallel time: 0.25 s

    For a 1.67X speedup across 2 threads

    In [5]: %ptime -n4 x + x  # use 4 threads

    Total serial time:   0.82 s

    Total parallel time: 0.31 s

    For a 2.60X speedup across 4 threads

    In [6]: res = %ptime -o x + x  # Get the result

    Total serial time:   0.41 s

    Total parallel time: 0.25 s

    For a 1.66X speedup across 2 threads

    In [7]: res.speedup

    Out[7]: 1.6610825669011922

    In [8]: %%ptime  # Use as a cell magic

    ...: x = np.ones((5000, 10000))

    ...: y = x + x

    ...:

    Total serial time:   0.72 s

    Total parallel time: 0.47 s

    For a 1.54X speedup across 2 threads

Install

-------

This package is available via pip:

.. code::

    pip install ptime