sympy2c: From symbolic expressions to fast C/C++ functions and ODE solvers in Python


Loading...

Date

2023-01

Publication Type

Journal Article

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Computer algebra systems play an important role in science as they facilitate the development of new theoretical models. The resulting symbolic equations are often implemented in a compiled programming language in order to provide fast and portable codes for practical applications. We describe sympy2c, a new Python package designed to bridge the gap between the symbolic development and the numerical implementation of a theoretical model. sympy2c translates symbolic equations implemented in the SymPy Python package to C/C++ code that is optimized using symbolic transformations. The resulting functions can be conveniently used as an extension module in Python. sympy2c is used within the PyCosmo Python package to solve the Einstein–Boltzmann equations, a large system of ODEs describing the evolution of linear perturbations in the Universe. After reviewing the functionalities and usage of sympy2c, we describe its implementation and optimization strategies. This includes, in particular, a novel approach to generate optimized ODE solvers making use of the sparsity of the symbolic Jacobian matrix. We demonstrate its performance using the Einstein–Boltzmann equations as a test case. sympy2c is general and potentially useful for various areas of computational physics. sympy2c is publicly available at https://cosmology.ethz.ch/research/software-lab/sympy2c.html.

Publication status

published

Editor

Book title

Volume

42

Pages / Article No.

100666

Publisher

Elsevier

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

code generation; ODE solver; Python; Computer algebra

Organisational unit

03928 - Refregier, Alexandre / Refregier, Alexandre check_circle

Notes

Funding

192243 - Cosmological Weak Lensing and Neutral Hydrogen (SNF)

Related publications and datasets