Multistage Mixed Precision Iterative Refinement

TL;DR

This paper introduces a multistage mixed precision iterative refinement method for solving linear systems that adaptively switches between precision levels to balance performance and accuracy, reducing unnecessary high-precision computations.

Contribution

It develops a novel multistage algorithm that combines existing mixed precision approaches with adaptive switching, improving usability and efficiency in solving linear systems.

Findings

Reduces computational costs by adaptively switching precisions.

Avoids refactorization in higher precision in some scenarios.

Confirmed benefits through extensive numerical experiments.

Abstract

Low precision arithmetic, in particular half precision floating point arithmetic, is now available in commercial hardware. Using lower precision can offer significant savings in computation and communication costs with proportional savings in energy. Motivated by this, there has been a renewed interest in mixed precision iterative refinement for solving linear systems $Ax=b$, and new variants of GMRES-based iterative refinement have been developed. Each particular variant with a given combination of precisions leads to different condition number-based constraints for convergence of the backward and forward errors, and each has different performance costs. The constraints for convergence given in the literature are, as an artifact of the analyses, often overly strict in practice, and thus could lead a user to select a more expensive variant when a less expensive one would have sufficed. In this work, we develop a multistage mixed precision iterative refinement solver which aims to combine existing mixed precision approaches to balance performance and accuracy and improve usability. For an initial combination of precisions, the algorithm begins with the least expensive approach and convergence is monitored via inexpensive computations with quantities produced during the iteration. If slow convergence or divergence is detected using particular stopping criteria, the algorithm switches to use a more expensive, but more reliable variant. A novel aspect of our approach is that, unlike existing implementations, our algorithm first attempts to use ``stronger'' solvers for the solution update before resorting to increasing the precision(s). In some scenarios, this can avoid the need to refactorize the matrix in higher precision. We perform extensive numerical experiments on random dense problems and problems from real applications which confirm the benefits of the multistage approach.