• Corpus ID: 246240616

Mixed Precision GMRES-based Iterative Refinement with Recycling

  title={Mixed Precision GMRES-based Iterative Refinement with Recycling},
  author={Eda Aydin Oktay and Erin C. Carson},
With the emergence of mixed precision capabilities in hardware, iterative refinement schemes for solving linear systems Ax = b have recently been revisited and reanalyzed in the context of three or more precisions. These new analyses show that under certain constraints on condition number, the LU factorization of the matrix can be computed in low precision without affecting the final accuracy. Another promising technique is GMRES-based iterative refinement, which, in contrast to the standard… 


Five-Precision GMRES-based iterative refinement
The five-precision GMRES-based iterative refinement (GMRES-IR5) algorithm has the potential to solve relatively badly conditioned problems in less time and memory than GM RES-IR3, thanks to the use of lower precision arithmetic in the GMRES iterations.
A New Analysis of Iterative Refinement and Its Application to Accurate Solution of Ill-Conditioned Sparse Linear Systems
A GMRES-based iterative refinement method that makes use of the computed LU factors as preconditioners and can succeed where standard refinement fails, and that it can provide accurate solutions to systems with condition numbers of order $u^{-1}$ and greater.
Accelerating Restarted GMRES With Mixed Precision Arithmetic
The generalized minimum residual method (GMRES) is a commonly used iterative Krylov solver for sparse, non-symmetric systems of linear equations and theoretical results linking the convergence of finite precision GMRES with classical Gram-Schmidt with reorthogonalization and its infinite precision counterpart are provided.
Accelerating the Solution of Linear Systems by Iterative Refinement in Three Precisions
The results suggest that on architectures for which half precision is efficiently implemented it will be possible to solve certain linear systems up to twice as fast and to greater accuracy, as well as recommending a standard solver that uses LU factorization in single precision.
Multistage Mixed Precision Iterative Refinement
A novel aspect of this approach is that, unlike existing implementations, this algorithm first attempts to use “stronger” GMRES-based solvers for the solution update before resorting to increasing the precision(s).
A survey of subspace recycling iterative methods
This survey concerns subspace recycling methods, a popular class of iterative methods that enable effective reuse of subspace information in order to speed up convergence and find good initial
Recycling Krylov Subspaces for Sequences of Linear Systems
This work proposes and analyzes two methods that significantly reduce the total number of matrix-vector products required to solve all systems and can reduce the cost of solving subsequent systems in the sequence by recycling selected subspaces generated for previous systems.
Truncation Strategies for Optimal Krylov Subspace Methods
A method is described that reveals which subspaces of the Krylov space were important for convergence thus far and exactly how important they are and is used to select which subspace to keep for orthogonalizing future search directions.
GMRES with Deflated Restarting
  • R. Morgan
  • Computer Science
    SIAM J. Sci. Comput.
  • 2002
The deflation of small eigenvalues can greatly improve the convergence of restarted GMRES and it is demonstrated that using harmonic Ritz vectors is important because then the whole subspace is a Krylov subspace that contains certain important smaller subspaces.
The Worst-Case GMRES for Normal Matrices
This paper completely characterize the worst-case GMRES-related quantities in the next-to-last iteration step and evaluates the standard bound in terms of explicit polynomials involving the matrix eigenvalues involving the determinants of the GMRES residual norm.