Parallel Python

Part 1: Functional Programming

To understand how best to write efficient parallel programs in Python, we first need to gain an understanding of “Functional Programming”. This is a style of programming in which functions are treated and manipulated as objects, i.e. functions can be assigned to variables, they can be passed as arguments, and they can be stored in containers along with other data.

Using the idea of functional programming, we can write parallel code that works by running lots of functions in parallel on large amounts of data. Your job, as a parallel functional programmer, is to design these functions. You then connect them together, with their dependencies, to create a team that can be distributed across the cores of a processor, or nodes of a cluster.

Functional programming is not the only way of writing an efficient parallel program. Other examples are shared-memory parallelism (e.g. check out the OpenMP course) or message passing (e.g. check out the MPI course). However, functional programming is, in my opinion, the easiest and most efficient way of extracting good performance from a parallel Python program.

Functional programming is useful also in its own right, and adopting it as a programming style can improve the readability and efficiency of even single-threaded Python code. It can make your intention as a programmer clearer to other developers, it can limit the amount of code duplication or retyping, and can also reduce the number of lines of code needed to represent your algorithm.