A Non-static Data Layout Enhancing Parallelism and Vectorization in Sparse Grid Algorithms

Handle URI:
http://hdl.handle.net/10754/597350
Title:
A Non-static Data Layout Enhancing Parallelism and Vectorization in Sparse Grid Algorithms
Authors:
Buse, Gerrit; Pfluger, Dirk; Murarasu, Alin; Jacob, Riko
Abstract:
The name sparse grids denotes a highly space-efficient, grid-based numerical technique to approximate high-dimensional functions. Although employed in a broad spectrum of applications from different fields, there have only been few tries to use it in real time visualization (e.g. [1]), due to complex data structures and long algorithm runtime. In this work we present a novel approach inspired by principles of I/0-efficient algorithms. Locally applied coefficient permutations lead to improved cache performance and facilitate the use of vector registers for our sparse grid benchmark problem hierarchization. Based on the compact data structure proposed for regular sparse grids in [2], we developed a new algorithm that outperforms existing implementations on modern multi-core systems by a factor of 37 for a grid size of 127 million points. For larger problems the speedup is even increasing, and with execution times below 1 s, sparse grids are well-suited for visualization applications. Furthermore, we point out how a broad class of sparse grid algorithms can benefit from our approach. © 2012 IEEE.
Citation:
Buse G, Pfluger D, Murarasu A, Jacob R (2012) A Non-static Data Layout Enhancing Parallelism and Vectorization in Sparse Grid Algorithms. 2012 11th International Symposium on Parallel and Distributed Computing. Available: http://dx.doi.org/10.1109/ISPDC.2012.34.
Publisher:
Institute of Electrical and Electronics Engineers (IEEE)
Journal:
2012 11th International Symposium on Parallel and Distributed Computing
KAUST Grant Number:
UK-C0020
Issue Date:
Jun-2012
DOI:
10.1109/ISPDC.2012.34
Type:
Conference Paper
Sponsors:
This publication is based on work supported by Award No.UK-C0020, made by King Abdullah University of Science andTechnology (KAUST).
Appears in Collections:
Publications Acknowledging KAUST Support

Full metadata record

DC FieldValue Language
dc.contributor.authorBuse, Gerriten
dc.contributor.authorPfluger, Dirken
dc.contributor.authorMurarasu, Alinen
dc.contributor.authorJacob, Rikoen
dc.date.accessioned2016-02-25T12:31:18Zen
dc.date.available2016-02-25T12:31:18Zen
dc.date.issued2012-06en
dc.identifier.citationBuse G, Pfluger D, Murarasu A, Jacob R (2012) A Non-static Data Layout Enhancing Parallelism and Vectorization in Sparse Grid Algorithms. 2012 11th International Symposium on Parallel and Distributed Computing. Available: http://dx.doi.org/10.1109/ISPDC.2012.34.en
dc.identifier.doi10.1109/ISPDC.2012.34en
dc.identifier.urihttp://hdl.handle.net/10754/597350en
dc.description.abstractThe name sparse grids denotes a highly space-efficient, grid-based numerical technique to approximate high-dimensional functions. Although employed in a broad spectrum of applications from different fields, there have only been few tries to use it in real time visualization (e.g. [1]), due to complex data structures and long algorithm runtime. In this work we present a novel approach inspired by principles of I/0-efficient algorithms. Locally applied coefficient permutations lead to improved cache performance and facilitate the use of vector registers for our sparse grid benchmark problem hierarchization. Based on the compact data structure proposed for regular sparse grids in [2], we developed a new algorithm that outperforms existing implementations on modern multi-core systems by a factor of 37 for a grid size of 127 million points. For larger problems the speedup is even increasing, and with execution times below 1 s, sparse grids are well-suited for visualization applications. Furthermore, we point out how a broad class of sparse grid algorithms can benefit from our approach. © 2012 IEEE.en
dc.description.sponsorshipThis publication is based on work supported by Award No.UK-C0020, made by King Abdullah University of Science andTechnology (KAUST).en
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)en
dc.subjectcacheen
dc.subjectparallelen
dc.subjectpermutationen
dc.subjectSIMDen
dc.subjectsparse gridsen
dc.titleA Non-static Data Layout Enhancing Parallelism and Vectorization in Sparse Grid Algorithmsen
dc.typeConference Paperen
dc.identifier.journal2012 11th International Symposium on Parallel and Distributed Computingen
dc.contributor.institutionTechnische Universitat Munchen, Munich, Germanyen
dc.contributor.institutionEidgenossische Technische Hochschule Zurich, Zurich, Switzerlanden
kaust.grant.numberUK-C0020en
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.