A small comparison between different methods of accelerating numerical python code

To get started, install the required packages. I used conda:

$ conda create -n ndarray_comparison python=3.9 numpy numba=0.54.1 cython jupyterlab seaborn pythran watermark setuptools-rust --override-channels -c conda-forge

conda activate ndarray_comparison

Then to compile the extensions:

$ python setup.py build_ext --inplace

The benchmark can be found in the following ipython notebook: https://github.com/synapticarbors/ndarray_comparison/blob/main/comparison.ipynb