Adaptively restarted block Krylov subspace methods with low-synchronization skeletons

DOI10.1007/s11075-022-01437-1zbMath1522.65057arXiv2208.09899OpenAlexW4313421141MaRDI QIDQ6157451

Kathryn Lund

Publication date: 11 May 2023

Published in: Numerical Algorithms (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/2208.09899

zbMATH Keywords

stabilityKrylov subspace methodsArnoldi methodhigh-performance computingblock methodsGram-Schmidtloss of orthogonalitycommunication-avoiding methodslow-synchronization methods

Mathematics Subject Classification ID

Computational methods for sparse matrices (65F50) Iterative numerical methods for linear systems (65F10) Parallel numerical computation (65Y05) Software, source code, etc. for problems pertaining to linear algebra (15-04) Orthogonalization in numerical linear algebra (65F25)

Cites Work

• Unnamed Item
• Backward error analysis of the AllReduce algorithm for Householder QR decomposition
• An adaptive \(s\)-step conjugate gradient algorithm with dynamic basis updating.
• A note on the error analysis of classical Gram-Schmidt
• Roundoff error analysis of the CholeskyQR2 algorithm
• Updating the QR decomposition of block tridiagonal and block Hessenberg matrices
• The block grade of a block Krylov space
• A hybrid block GMRES method for nonsymmetric systems with multiple right-hand sides
• Ritz and pseudo-Ritz values using matrix polynomials
• Block Gram-Schmidt algorithms and their stability properties
• Block Krylov subspace methods for functions of matrices
• Rounding error analysis of the classical Gram-Schmidt orthogonalization process
• Convergence properties of block GMRES and matrix polynomials
• Enlarged Krylov Subspace Conjugate Gradient Methods for Reducing Communication
• Analysis of the Rational Krylov Subspace Projection Method for Large-Scale Algebraic Riccati Equations
• Rational Krylov approximation of matrix functions: Numerical methods and optimal pole selection
• Communication-optimal Parallel and Sequential QR and LU Factorizations
• The university of Florida sparse matrix collection
• The Stability of Block Variants of Classical Gram--Schmidt
• A Storage-Efficient $WY$ Representation for Products of Householder Transformations
• The Adaptive $s$-Step Conjugate Gradient Method
• Communication lower bounds and optimal algorithms for numerical linear algebra
• Accuracy and Stability of Numerical Algorithms
• Preparing sparse solvers for exascale computing
• Block Krylov Subspace Methods for Functions of Matrices II: Modified Block FOM
• Strategies for the Vectorized Block Conjugate Gradients Method
• Hardware-Oriented Krylov Methods for High-Performance Computing
• Block Modified Gram--Schmidt Algorithms and Their Analysis
• Bits and Bugs
• On Improving Linear Solver Performance: A Block Variant of GMRES
• Low synchronization Gram–Schmidt and generalized minimal residual algorithms
• Randomized Gram--Schmidt Process with Application to GMRES

This page was built for publication: Adaptively restarted block Krylov subspace methods with low-synchronization skeletons