Python performance workshop

This mini-workshop is intended for Python developers who are interested in writing code with better performance and potentially running it on supercomputers.

Prerequisites

• Python programming basics (functions, for loops, if-else statements) and data structures (list, set, dict, tuple)

• Familiarity with well-known numeric libraries (for example, Numpy)

5 min	Installation
5 min	Introduction and motivation
10 min	Performance fundamentals
15 min	Benchmark
15 min	Profile
5-10 min	Break
15 min	Optimize
5 min	Parallelize

Preparation

• Installation

The lesson

• Introduction and motivation
• Performance fundamentals
• Benchmark
• Profile
• Optimize
• Parallelize

Reference

• Quick Reference
• Instructor’s guide

Who is the course for?

Software developers, researchers, students who use Python often and process a lot of data.

Credits

The lesson is inspired and derived from the following:

Creative Commons CC-BY 4.0 licensed material

• https://github.com/ENCCS/hpda-python

• https://github.com/ENCCS/word-count-hpda

• https://github.com/coderefinery/word-count

• https://coderefinery.github.io/reproducible-research

• https://hpc-carpentry.github.io/hpc-python/

• Images and description by authors of lectures.scientific-python.org and by authors of deep-learning-intro

• PyCon Sweden 2019 talk on https://talks.fluid.quest/

Other open-source licenced material

• Images by authors of Project Jupyter is licensed under BSD 3-Clause “New” or “Revised” License

• Images from The Noun Project is licensed under CC-BY 3.0

• Code from https://github.com/paugier/nbabel is licensed under GPLv2

• Video from https://en.wikipedia.org/wiki/File:Galaxy_collision.ogv licensed under CC-BY 3.0