Original title: Performance of parallel QR factorization methods on the NVIDIA Grace CPU Superchip
Authors: Břichňáč, V. ; Šístek, Jakub
Document type: Papers
Conference/Event: Programs and Algorithms of Numerical Mathematics /22./, Hejnice (CZ), 20240623
Year: 2025
Language: eng
Abstract: This article studies several algorithms for QR factorization based on hierarchical Householder reflectors organized into elimination trees, which are particularly suited for tall-and-skinny matrices and allow parallelization. We examine the effect of various parameters on the performance of the tree-based algorithms. The work is accompanied with a custom implementation that utilizes a task-based runtime system (OpenMP or StarPU). The same algorithm is implemented in the PLASMA library. The performance evaluation is done on the recent NVIDIA Grace CPU Superchip.
Keywords: NVIDIA Grace CPU; QR factorization; task-based programming
Project no.: GA23-06159S (CEP)
Funding provider: GA ČR
Host item entry: Programs and Algorithms of Numerical Mathematics 22 : Proceedings of Seminar, ISBN 978-80-85823-74-5
Note: Související webová stránka: http://dx.doi.org/10.21136/panm.2024.03

Institution: Institute of Mathematics AS ČR (web)
Document availability information: Fulltext is available in the digital repository of the Academy of Sciences.
Original record: https://hdl.handle.net/11104/0368021

Permalink: http://www.nusl.cz/ntk/nusl-684970


The record appears in these collections:
Research > Institutes ASCR > Institute of Mathematics
Conference materials > Papers