Skip to content

A mixed-precision Paterson–Stockmeyer algorithm for evaluating matrix polynomials.

License

Notifications You must be signed in to change notification settings

xiaobo-liu/mp-ps

Repository files navigation

Mixed-Precision Paterson--Stockmeyer Method for Evaluating Polynomials of Matrices

Function include/PS_mp.m computes the Taylor approximants to the matrix exponential in arbitrary precision; function include/cosm_hsd.m is the counterpart that works in double precision and uses only double, single, and half (simulated by chop) precisions (no mp computation). Function include/PS_mp_gen.m is for general polynomials of matrices and takes the scalar coefficients (of decaying modulus) as one of the inputs.

Details on the underlying algorithms can be found in the technical report:

X. Liu. Mixed-precision Paterson–Stockmeyer method for evaluating polynomials of matrices, preprint, https://arxiv.org/abs/2312.17396.

All codes used for generating the data in the above report are included in this repository.

Dependencies

The code in this repository may require the Advanpix Multiprecision Computing Toolbox for MATLAB (www.advanpix.com).

License

See license.txt for licensing information.

About

A mixed-precision Paterson–Stockmeyer algorithm for evaluating matrix polynomials.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages