An open API service indexing awesome lists of open source software.

https://github.com/nuitka/nuitka-speedcenter

Tool to create a website with Nuitka performance statistics
https://github.com/nuitka/nuitka-speedcenter

nuitka

Last synced: 7 months ago
JSON representation

Tool to create a website with Nuitka performance statistics

Awesome Lists containing this project

README

          

Nuitka Speedcenter
------------------

Welcome to Nuitka speedcenter, the underlying tool behind the site dedicated
to performance and Nuitka: https://speedcenter.nuitka.net

Currently this is very newly released and relatively ugly, hope is for you
people to join and improve this in entirely new ways.

Installation
------------

This needs Python 3.9 and Linux currently. Also ``Valgrind`` will have to be
installed.

Then do these commands (may take a while, it is compile parts of the
dependencies via C it seems, and not all of it is available as wheels):

.. code-block::

python3.9 -m pip install pipenv
python3.9 -m pipenv install

Now the tools are runnable, the main frontend currently for building the
Nikola site is this:

.. code-block::

python3.9 -m pipenv run python ./update.py --update-all

Note it will attempt to deploy and fail for you. The UI is currently not really good,
the tool is mostly meant to run on CI servers, feel free to improve it for interactive
use.

To look at it, consider using this:

.. code-block::

python3.9 -m pipenv run nikola serve

Improvements Needed
-------------------

Construct based tests are a nice way to take performance. But maybe test cases
should be generated via Jinja2 and not via hand made pre-processor. Not sure
really, because one of my initial goals in doing it was to have proper syntax
for Python. But Jinja2 will be better at generating and therefore also at
maintaining more test cases easier.

For the rendering, Nikola and a graphing plugin are used. Nikola has a better
thing for diff display. We either expand on this, or we turn to something
entirely different.

And lastly, on Windows there is no Valgrind. Is there an alternative way to
measure performance as accurate as Valgrind can for Linux?