Scaling H2 Factorization to Billion-Unknown Three-Dimensional Problems

DOI10.48550/ARXIV.2610.30131MaRDI QIDQ7401830arXivFDO

Authors Tianyu Liang, Chao Chen, James Demmel, Yang Liu

Publication date 21 October 2026

Published in arXiv

Tldr We introduce a distributed-memory $\mathcal{H}^2$ factorization solver that combines spatial box coloring, delayed Schur complement updates, streamed sketching-based interpolative decomposition, and BLAS-3 and shared-memory kernels to solve billion-unknown three-dimensional problems efficiently. This is the first demonstrated strong recursive skeletonization factorization for three-dimensional volume integral equations at this scale.

Full work available at URL https://doi.org/10.48550/ARXIV.2610.30131

Mathematics Subject Classification ID

Direct numerical methods for linear systems and matrix inversion (65F05) Numerical methods for low-rank matrix approximation; matrix compression (65F55) Numerical methods for integral equations (65R20) Parallel numerical computation (65Y05)

arXiv classification

math-phmath.NA

Cites work

• ``Compress and Eliminate” Solver for Symmetric Positive Definite Sparse Matrices
• A recursive skeletonization factorization based on strong admissibility
• Accuracy Directly Controlled Fast Direct Solutions of General ${\cal H}^2$-Matrices and Its Application to Electrically Large Integral-Equation-Based Electromagnetic Analysis
• Accuracy Directly Controlled Fast Direct Solutions of General ${\cal H}^2$-Matrices and Its Application to Electrically Large Integral-Equation-Based Electromagnetic Analysis
• An O(N) distributed-memory parallel direct solver for planar integral equations
• Efficient Algorithms for Computing a Strong Rank-Revealing QR Factorization
• Fast Direct Solvers
• Fast hierarchical solvers for sparse matrices using extended sparsification and low-rank approximation
• FMM-LU: A Fast Direct Solver for Multiscale Boundary Integral Equations in Three Dimensions
• Introduction to hierarchical matrices with applications.
• Low-rank approximation and regression in input sparsity time
• Numerical instability of calculating inverse of spatial covariance matrices
• On the Compression of Low Rank Matrices
• Parallelization of the inverse fast multipole method with an application to boundary element method
• Randomized algorithms for the low-rank approximation of matrices
• Scalable Linear Time Dense Direct Solver for 3-D Problems Without Trailing Sub-Matrix Dependencies
• The inverse fast multipole method: using a fast approximate direct dolver as a preconditioner for dense linear systems

This page was built for publication: Scaling H2 Factorization to Billion-Unknown Three-Dimensional Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7401830)