Reproducibility strategies for parallel preconditioned conjugate gradient

DOI10.1016/J.CAM.2019.112697MaRDI QIDQ2297141zbMATH OpenOpenAlexWikidataFDO

Authors Roman Iakymchuk, M. Wiesenberger, Enrique S. Quintana-Ortí, María Barreda, José I. Aliaga

Publication date 18 February 2020

Published in Journal of Computational and Applied Mathematics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/j.cam.2019.112697

zbMATH Keywords

reproducibilityaccuracyhigh-performance computingpreconditioned conjugate gradientfloating-point expansionlong accumulator

Mathematics Subject Classification ID

Parallel numerical computation (65Y05) Complexity and performance of numerical algorithms (65Y20) Modes of computation (nondeterministic, parallel, interactive, probabilistic, etc.) (68Q10)

Recommendations

• First steps towards more numerical reproducibility
• Accuracy and effectiveness of preconditioned conjugate gradient algorithms for large and ill-conditioned problems
• Reproducible and accurate matrix multiplication
• Using accurate arithmetics to improve numerical reproducibility and stability in parallel applications
• Efficiency of reproducible level 1 BLAS

Cites work

• scientific article; zbMATH DE number 3303655 (Why is no real title available?)
• A set of level 3 basic linear algebra subprograms
• Accelerating the solution of linear systems by iterative refinement in three precisions
• Accuracy and Stability of Numerical Algorithms
• Accurate Sum and Dot Product
• Accurate floating-point summation. II: Sign, \(K\)-fold faithful and rounding to nearest
• Basic Linear Algebra Subprograms for Fortran Usage
• Emulation of a FMA and Correctly Rounded Sums: Proved Algorithms Using Rounding to Odd
• Error-free transformations of matrix multiplication by using fast routines of matrix multiplication and its applications
• Handbook of Floating-Point Arithmetic
• High-Precision Anchored Accumulators for Reproducible Floating-Point Summation
• MPFR
• Parallel Reproducible Summation
• Parallel iterative methods for sparse linear systems
• Reproducible and accurate matrix multiplication
• The exact dot product as basic tool for long interval arithmetic

Cited in

(3)

• Compensated summation and dot product algorithms for floating-point vectors on parallel architectures: error bounds, implementation and application in the Krylov subspace methods
• General framework for deriving reproducible Krylov subspace algorithms: BiCGStab case
• Infinite-precision inner product and sparse matrix-vector multiplication using Ozaki scheme with Dot2 on manycore processors

Describes a project that uses

Uses Software

• BLAS
• MPFR
• mctoolbox
• LBNL
• HPL
• ExBLAS
• FELTOR

This page was built for publication: Reproducibility strategies for parallel preconditioned conjugate gradient

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2297141)