M
Mark Hoemmen
Researcher at Sandia National Laboratories
Publications - 45
Citations - 2253
Mark Hoemmen is an academic researcher from Sandia National Laboratories. The author has contributed to research in topics: Iterative method & QR decomposition. The author has an hindex of 20, co-authored 45 publications receiving 2037 citations. Previous affiliations of Mark Hoemmen include University of California, Berkeley.
Papers
More filters
Journal ArticleDOI
Communication-optimal Parallel and Sequential QR and LU Factorizations
TL;DR: Two parallel and sequential dense QR factorization algorithms that are both optimal (up to polylogarithmic factors) in the amount of communication they perform, and just as stable as Householder QR are presented.
Posted Content
Communication-optimal parallel and sequential QR and LU factorizations
TL;DR: In this article, the authors present parallel and sequential dense QR factorization algorithms that are both optimal (up to polylogarithmic factors) in the amount of communication they perform, and just as stable as Householder QR.
Communication-avoiding krylov subspace methods
James Demmel,Mark Hoemmen +1 more
TL;DR: This thesis aims to take s steps of a KSM for the same communication cost as 1 step, which would be optimal, and proposes techniques for developing communication-avoiding versions of nonsymmetric Lanczos iteration and Method of Conjugate Gradients for solving linear systems.
Proceedings ArticleDOI
Minimizing communication in sparse matrix solvers
TL;DR: This work reorganizes the sparse-matrix kernel to compute a set of matrix-vector products at once and reorganizing the rest of the algorithm accordingly, resulting in a new variant of GMRES that gets speedups of up to 4.3x over standard GMRES, without sacrificing convergence rate or numerical stability.
Proceedings ArticleDOI
Avoiding communication in sparse matrix computations
TL;DR: This paper focuses on an alternative building block for sparse iterative solvers, the "matrix powers kernel" [x, Ax, A2x, ..., Akx], and shows that by organizing computations around this kernel, it can achieve near-minimal communication costs.