Virtual Systolic Array for QR Decomposition

TitleVirtual Systolic Array for QR Decomposition
Publication TypeConference Paper
Year of Publication2013
AuthorsKurzak, J., P. Luszczek, M. Gates, I. Yamazaki, and J. Dongarra
Conference Name15th Workshop on Advances in Parallel and Distributed Computational Models, IEEE International Parallel & Distributed Processing Symposium (IPDPS 2013)
Date Published05-2013
Conference LocationBoston, MA
Keywordsdataflow programming, message passing, multi-core, QR decomposition, roofline model, systolic array

Systolic arrays offer a very attractive, data-centric, execution model as an alternative to the von Neumann architecture. Hardware implementations of systolic arrays turned out not to be viable solutions in the past. This article shows how the systolic design principles can be applied to a software solution to deliver an algorithm with unprecedented strong scaling capabilities. Systolic array for the QR decomposition is developed and a virtualization layer is used for mapping of the algorithm to a large distributed memory system. Strong scaling properties are discovered, superior to existing solutions.