High-Performance Multi-Precision Tool for Floating-Point Computations

Abstract

This paper presents a MATLAB toolbox for multiple-precision arithmetic, enabling high-precision numerical computations beyond the standard double-precision format. The development step and functionality of the toolbox are described, and how it integrates seamlessly with the MATLAB computational ecosystem, offering a user-friendly interface. The performance enhancement and usability of the toolbox are demonstrated through four use cases where floating-point arithmetic becomes an issue. These benchmark results illustrate the toolbox's computational accuracy and performance, highlighting its ability to mitigate numerical instabilities and roundoff errors in sensitive computations. The toolbox empowers researchers and engineers to implement more reliable models, test advanced algorithms, and validate system designs and a diverse set of operations for applications that require enhanced precision, such as numerical analysis, scientific computing, and applied mathematics.