Software

Libraries

• librla: A Randomized Linear Algebra Library. A unified, multi-language library implementing randomized algorithms for low-rank matrix factorizations. Provides efficient randomized-sketching-based methods for large-scale matrix factorizations with consistent APIs across Python, MATLAB/Octave, and Julia. Performance is similar to PyTorch but with more functionality.
• librla_applications: The application of the librla library to commonly visited examples from the randomized linear algebra literature. If you have an example you would like to add, please reach out.

Demo Codes

• HPS demo codes: Collection of Matlab demo codes for using the HPS method for two-dimensional Laplace and Helmholtz problems (including free-space Helmholtz with well-conditioned FEM-BEM coupling).
• Fast direct solver Minitutorial: P.G. Martinsson and A. Gillman presented a minitutorial at SIAM CSE 2025. This repo contains the talks and demo codes for both parts of the session.