High Performance Python

Princeton mini-course

By Henry Schreiner, with Jim Pivarski

Installation

Binder

In the minicourse, if you haven't prepared beforehand, please use this link to run online via Binder:

Codespaces

GitHub provides 120 core-hours (60 real-time hours if you use the smallest (2-core) setting) of CodeSpaces usage every month. You can run this in a codespace:

Note that you should currently start jupyter lab manually from the VSCode terminal once it's built (3-5 minutes after starting it for the first time).

Local install:

If you are reading this at least 10 minutes before the course starts or you have pixi, anaconda, or miniconda installed, you will probably be best off installing miniconda. This way you will keep local edits and will have an environment to play with.

Get the repository:

git clone https://github.com/henryiii/python-performance-minicourse.git
cd python-performance-minicourse

If you have pixi, just run:

pixi run lab

---

If you don't have pixi (it's just a single binary written in Rust), classic instructions follow:

Download and install miniconda. On macOS with homebrew, just run brew cask install miniconda (see my recommendations).

Run:

conda env create

from this directory. This will create an environment performance-minicourse. To use:

conda activate performance-minicourse
./check.py # Check to see if you've installed this correctly
jupyter lab

And, to disable:

conda deactivate

or restart your terminal.

---

To keep this in sync, I'm using this to export the pixi environment into something conda/mamba can use:

pixi workspace export conda-environment -n performance-minicourse > environment.yml

Binder is using conda.

Lessons

• 00 Intro: The introduction
• 01 Fractal accelerate: A look at a fractal computation, and ways to accelerate it with NumPy changes, numexpr, and numba.
  • 01b Fractal interactive: An interactive example using Numba.
• 02 Temperatures: A look at reading files and array manipulation in NumPy and Pandas.
• 03 MCMC: A Marco Chain Monte Carlo generator (and metropolis generator) in Python and Numba, with a focus on profiling.
• 04 Runge-Kutta: Implementing a popular integration algorithm in NumPy and Numba.
• 05 Distributed: An exploration of ways to break up code (fractal) into chunks for multithreading, multiproccessing, and Dask distribution.
• 06 Tensorflow: A look at implementing a Negative Log Likelihood function (used for unbinned fitting) in NumPy and Google's Tensorflow.
• 07 Callables: A look at Scipy's LowLevelCallable, and how to implement one with Numba.

Class participants: please complete the survey that will be posted.