arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2409.03079math.NAcs.NA
收藏

关于 s 步 GMRES 的向后稳定性

On the backward stability of s-step GMRES

Erin Carson, Yuxin Ma

AI总结:

本文提出改进分析框架以分离 QR 与最小二乘舍入误差,研究 s-step GMRES 的向后稳定性,并通过修正 s-step Arnoldi 过程在更大块大小下保持精度。

摘要:

通信(即数据移动)是经典 Krylov 子空间方法求解器在现代计算机架构上性能的关键瓶颈。为此,人们提出了这些方法的避免通信变体;它们在精确算术下等价,但在有限精度下可能不稳定。在本文中,我们研究 s-step GMRES(也称为 communication-avoiding GMRES,避免通信 GMRES)的向后稳定性。与 [A. Buttari、N. J. Higham、T. Mary 和 B. Vieublé,2024 年预印本] 提出的“模块化框架”相比,我们给出了一个改进框架,通过分离 QR 分解和最小二乘问题求解中舍入误差的影响,简化对 s-step GMRES 的分析;该框架将标准 GMRES(s=1)作为特例包含在内。这个新框架的关键优势在于,可以清楚看出正交化方法如何影响向后误差,而且当修改 GMRES 中使用的正交化方法时,除正交化本身外无需重新评估其他内容。利用该框架,我们分析了采用常用块正交化方法的 s-step GMRES:块修正 Gram–Schmidt 算法和重正交化块经典 Gram–Schmidt 算法。一个例子说明了 s-step GMRES 与经典 s-step Arnoldi 过程配合时产生的不稳定性,并展示了常用解决策略的局限。为解决这一问题,我们提出一种修正的 s-step Arnoldi 过程,它允许使用大得多的块大小 s,同时保持令人满意的精度;我们的数值实验证实了这一点。

英文摘要:

Communication, i.e., data movement, is a critical bottleneck for the performance of classical Krylov subspace method solvers on modern computer architectures. Variants of these methods which avoid communication have been introduced, which, while equivalent in exact arithmetic, can be unstable in finite precision. In this work, we address the backward stability of $s$-step GMRES, also known as communication-avoiding GMRES. Compared to the ``modular framework'' proposed in [A.~Buttari, N.~J.~Higham, T.~Mary, \& B.~Vieublé. Preprint in 2024.], we present an improved framework for simplifying the analysis of $s$-step GMRES, which includes standard GMRES ($s=1$) as a special case, by isolating the effects of rounding errors in the QR factorization and the solution of the least squares problem. The key advantage of this new framework is that it is evident how the orthogonalization method affects the backward error, and it is not necessary to re-evaluate anything other than the orthogonalization itself when modifying the orthogonalization used in GMRES. Using this framework, we analyze $s$-step GMRES with popular block orthogonalization methods: block modified Gram--Schmidt and reorthogonalized block classical Gram--Schmidt algorithms. An example illustrates the resulting instability of $s$-step GMRES when paired with the classical $s$-step Arnoldi process and shows the limitations of popular strategies for resolving this instability. To address this issue, we propose a modified $s$-step Arnoldi process that allows for much larger block size $s$ while maintaining satisfactory accuracy, as confirmed by our numerical experiments.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑