Multilevel communication optimal LU and QR factorizations for hierarchical platforms

Grigori, Laura and Jacquelin, Mathias and Khabou, Amal (2013) Multilevel communication optimal LU and QR factorizations for hierarchical platforms. [MIMS Preprint]

Warning
There is a more recent version of this item available.
[thumbnail of ipdps.pdf] PDF
ipdps.pdf

Download (437kB)

Abstract

This study focuses on the performance of two classical dense linear algebra algorithms, the LU and the QR factorizations, on multilevel hierarchical platforms. We first introduce a performance model called Hierarchical Cluster Platform (HCP), encapsulating the characteristics of such platforms. The focus is set on reducing the communication requirements of studied algorithms at each level of the hierarchy. Lower bounds on communication are therefore extended with respect to the \HCP model. We then introduce multilevel LU and QR algorithms tailored for those platforms, and provide a detailed performance analysis. We also provide a set of performance predictions showing the need for such algorithms on large platforms.

Item Type: MIMS Preprint
Additional Information: submitted to the 28th IEEE International Parallel & Distributed Processing Symposium
Uncontrolled Keywords: QR, LU, exascale, hierarchical platforms
Subjects: MSC 2010, the AMS's Mathematics Subject Classification > 65 Numerical analysis
Depositing User: Amal Khabou
Date Deposited: 25 Oct 2013
Last Modified: 08 Nov 2017 18:18
URI: https://eprints.maths.manchester.ac.uk/id/eprint/2031

Available Versions of this Item

Actions (login required)

View Item View Item