Fast Python: NumPy and Cython

Abstract

Python can combine simplicity with speed in a unique way, but much care has to be paid in avoiding bottlenecks, particularly avoiding unnecessary loops and heavily using its vectorized functions. In this chapter I make a detailed overview of the types of Python, how to speedup calculations with array, how to efficiently select subsets of large arrays. By illustrating strides, I explain how Numerical Python can achieve its performance and how to benchmark them. Finally, I illustrate how to add vector and linear algebra calculations to an applied modeling software and eventually use Cython, the C extension to Python, to add special functions that cannot be performed using the vectorized NumPy library.

“Give me 6 hours to chop a tree, I will spend the first 4 sharpening my axe.” —Abe Lincoln