elib
DLR-Header
DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Sparse matrix-vector multiplication on GPGPU clusters: A new storage format and a scalable implementation

Kreutzer, Moritz and Hager, Georg and Wellein, Gerhard and Fehske, Holger and Basermann, Achim and Bishop, Alan R. (2012) Sparse matrix-vector multiplication on GPGPU clusters: A new storage format and a scalable implementation. In: Bisher bei IEEE Xplore (online, URL s.u.); Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops (IPDPS 2012), 1696 -1702. IEEE Conference Publications. Workshop on Large-Scale Parallel Processing to be held at the IEEE International Parallel and Distributed Processing Symposium 2012, 21.-25. Mai 2012, Shanghai, China. ISBN 978-1-4673-0974-5.

[img]
Preview
PDF
153kB

Official URL: http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6270844

Abstract

Sparse matrix-vector multiplication (spMVM) is the dominant operation in many sparse solvers. We investigate performance properties of spMVM with matrices of various sparsity patterns on the nVidia “Fermi” class of GPGPUs. A new “padded jagged diagonals storage” (pJDS) format is proposed which may substantially reduce the memory overhead intrinsic to the widespread ELLPACK-R scheme while making no assumptions about the matrix structure. In our test scenarios the pJDS format cuts the overall spMVM memory footprint on the GPGPU by up to 70%, and achieves 91% to 130% of the ELLPACK-R performance. Using a suitable performance model we identify performance bottlenecks on the node level that invalidate some types of matrix structures for efficient multi-GPGPU parallelization. For appropriate sparsity patterns we extend previous work on distributed-memory parallel spMVM to demonstrate a scalable hybrid MPI-GPGPU code, achieving efficient overlap of communication and computation.

Item URL in elib:https://elib.dlr.de/75140/
Document Type:Conference or Workshop Item (Speech)
Title:Sparse matrix-vector multiplication on GPGPU clusters: A new storage format and a scalable implementation
Authors:
AuthorsInstitution or Email of AuthorsAuthor's ORCID iD
Kreutzer, MoritzErlangen Regional Computing CenterUNSPECIFIED
Hager, GeorgErlangen Regional Computing CenterUNSPECIFIED
Wellein, GerhardErlangen Regional Computing CenterUNSPECIFIED
Fehske, HolgerErnst-Moritz-Arndt University of GreifswaldUNSPECIFIED
Basermann, AchimDLR-KP, SC-VSSUNSPECIFIED
Bishop, Alan R.Los Alamos National LaboratoryUNSPECIFIED
Date:2012
Journal or Publication Title:Bisher bei IEEE Xplore (online, URL s.u.); Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops (IPDPS 2012)
Refereed publication:Yes
Open Access:Yes
Gold Open Access:No
In SCOPUS:No
In ISI Web of Science:No
Page Range:1696 -1702
Publisher:IEEE Conference Publications
ISBN:978-1-4673-0974-5
Status:Published
Keywords:Parallel sparse matrix-vector multiplication, multi-core processors, GPGPUs, new storage formats
Event Title:Workshop on Large-Scale Parallel Processing to be held at the IEEE International Parallel and Distributed Processing Symposium 2012
Event Location:Shanghai, China
Event Type:international Conference
Event Dates:21.-25. Mai 2012
Organizer:IEEE
HGF - Research field:Aeronautics, Space and Transport (old)
HGF - Program:Space (old)
HGF - Program Themes:W SY - Technik für Raumfahrtsysteme
DLR - Research area:Space
DLR - Program:W SY - Technik für Raumfahrtsysteme
DLR - Research theme (Project):W - Vorhaben SISTEC (old)
Location: Köln-Porz
Institutes and Institutions:Institut of Simulation and Software Technology
Institut of Simulation and Software Technology > Distributed Systems and Component Software
Deposited By: Basermann, Dr.-Ing. Achim
Deposited On:01 Mar 2012 08:08
Last Modified:31 Jul 2019 19:35

Repository Staff Only: item control page

Browse
Search
Help & Contact
Information
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.