Skip main navigation

Speeding up complex expressions with Numexpr

In this article we show how to speed up complex expressions with Numexpr.
© CC-BY-NC-SA 4.0 by CSC - IT Center for Science Ltd.

Complex expressions with large NumPy arrays present a bit of a catch-22 situation performance-wise.

On the one hand, using a one-liner for the expression is not a good idea due to high memory usage (unnecessary temporary arrays that are need for the evaluation).

On the other hand, evaluation of the expression with one operation at a time can lead into suboptimal performance. Effectively, one carries out multiple for loops in the NumPy C-code.

Numexpr package provides tools for fast evaluation of array expressions.

x = numpy.random.random((1000000, 1))
y = numpy.random.random((1000000, 1))

import numexpr
poly = numexpr.evaluate("((.25*x + .75)*x - 1.5)*x - 2")

The expression is enclosed in quotes and will be evaluated using a single
C-loop. Speed-ups in comparison to NumPy are typically between 0.95 and 4.
Performance improves normaly most with arrays that do not fit in CPU cache.

Supported operators and functions include e.g.:

  • +, -, *, /, **
  • sin, cos, tan
  • exp, log, sqrt

Thread support

By default, numexpr tries to use multiple threads, which can also speed up
the execution. The number of threads can be queried and set with:


The number of threads can also be controlled by the environment variables

© CC-BY-NC-SA 4.0 by CSC - IT Center for Science Ltd.
This article is from the free online

Python in High Performance Computing

Created by
FutureLearn - Learning For Life

Reach your personal and professional goals

Unlock access to hundreds of expert online courses and degrees from top universities and educators to gain accredited qualifications and professional CV-building certificates.

Join over 18 million learners to launch, switch or build upon your career, all at your own pace, across a wide range of topic areas.

Start Learning now