flameprof

Flamegraph generator for python's cProfile <https://docs.python.org/3/library/profile.html>_ stats.

Flamegraphs <http://www.brendangregg.com/flamegraphs.html>_ allow to visualize relations between functions in a very compact and understandable manner.

Flameprof solves main problems of built-in cProfile reporting and can replace gprof2dot <https://github.com/jrfonseca/gprof2dot>_ because later outputs very huge graphs with a lot of noise.

Flameprof works with profile stat files obtained by Profile.dump_stats() <https://docs.python.org/3/library/profile.html#profile.Profile.dump_stats>_ call or via direct script profiling::

python -m cProfile -o myscript.prof myscript.py

Install

Via pip::

pip install flameprof

Or you can invoke flameprof.py directly::

python flameprof.py input.prof > output.svg

Native svg (--format=svg)

Native svg features:

• compact function names with full names in a tooltip
• precise timings (cumulative and total)
• call counts (in a tooltip on hover)
• green bars show stack frames where flameprof starts to guess timing ratios
• inverted flamegraph to show total time of all calls

Graph width, row height, font size and threshold can be set via appropriate cli options.

::

flameprof requests.prof > requests.svg

.. image:: https://cdn.rawgit.com/baverman/flameprof/master/img/requests.svg :alt: Requests profile :width: 100% :align: center

Inverted flamegraph:

.. image:: https://cdn.rawgit.com/baverman/flameprof/master/img/httplib.svg :alt: Requests profile :width: 100% :align: center

(Images are clickable)

Svg generated with flamegraph.pl (--format=log)

Also flameprof can output trace log suitable as input for flamegraph.pl.

You can treat "samples" as microseconds by default (see --log-mult option).

::

flameprof --format=log requests.prof | flamegraph > requests-flamegraph.svg

.. image:: https://cdn.rawgit.com/baverman/flameprof/master/img/requests-flamegraph.svg :alt: Requests profile with flamegraph.pl :width: 100% :align: center

(Image is clickable)