Compact data structure and scalable algorithms for the sparse grid technique

Handle URI:
http://hdl.handle.net/10754/597803
Title:
Compact data structure and scalable algorithms for the sparse grid technique
Authors:
Murarasu, Alin; Weidendorfer, Josef; Buse, Gerrit; Butnaru, Daniel; Pflüger, Dirk
Abstract:
The sparse grid discretization technique enables a compressed representation of higher-dimensional functions. In its original form, it relies heavily on recursion and complex data structures, thus being far from well-suited for GPUs. In this paper, we describe optimizations that enable us to implement compression and decompression, the crucial sparse grid algorithms for our application, on Nvidia GPUs. The main idea consists of a bijective mapping between the set of points in a multi-dimensional sparse grid and a set of consecutive natural numbers. The resulting data structure consumes a minimum amount of memory. For a 10-dimensional sparse grid with approximately 127 million points, it consumes up to 30 times less memory than trees or hash tables which are typically used. Compared to a sequential CPU implementation, the speedups achieved on GPU are up to 17 for compression and up to 70 for decompression, respectively. We show that the optimizations are also applicable to multicore CPUs. Copyright © 2011 ACM.
Citation:
Murarasu A, Weidendorfer J, Buse G, Butnaru D, Pflüger D (2011) Compact data structure and scalable algorithms for the sparse grid technique. Proceedings of the 16th ACM symposium on Principles and practice of parallel programming - PPoPP ’11. Available: http://dx.doi.org/10.1145/1941553.1941559.
Publisher:
Association for Computing Machinery (ACM)
Journal:
Proceedings of the 16th ACM symposium on Principles and practice of parallel programming - PPoPP '11
KAUST Grant Number:
UKC0020
Issue Date:
2011
DOI:
10.1145/1941553.1941559
Type:
Conference Paper
Sponsors:
This publication is based on work supported by Award No. UKC0020,made by King Abdullah University of Science and Technology(KAUST).
Appears in Collections:
Publications Acknowledging KAUST Support

Full metadata record

DC FieldValue Language
dc.contributor.authorMurarasu, Alinen
dc.contributor.authorWeidendorfer, Josefen
dc.contributor.authorBuse, Gerriten
dc.contributor.authorButnaru, Danielen
dc.contributor.authorPflüger, Dirken
dc.date.accessioned2016-02-25T12:56:59Zen
dc.date.available2016-02-25T12:56:59Zen
dc.date.issued2011en
dc.identifier.citationMurarasu A, Weidendorfer J, Buse G, Butnaru D, Pflüger D (2011) Compact data structure and scalable algorithms for the sparse grid technique. Proceedings of the 16th ACM symposium on Principles and practice of parallel programming - PPoPP ’11. Available: http://dx.doi.org/10.1145/1941553.1941559.en
dc.identifier.doi10.1145/1941553.1941559en
dc.identifier.urihttp://hdl.handle.net/10754/597803en
dc.description.abstractThe sparse grid discretization technique enables a compressed representation of higher-dimensional functions. In its original form, it relies heavily on recursion and complex data structures, thus being far from well-suited for GPUs. In this paper, we describe optimizations that enable us to implement compression and decompression, the crucial sparse grid algorithms for our application, on Nvidia GPUs. The main idea consists of a bijective mapping between the set of points in a multi-dimensional sparse grid and a set of consecutive natural numbers. The resulting data structure consumes a minimum amount of memory. For a 10-dimensional sparse grid with approximately 127 million points, it consumes up to 30 times less memory than trees or hash tables which are typically used. Compared to a sequential CPU implementation, the speedups achieved on GPU are up to 17 for compression and up to 70 for decompression, respectively. We show that the optimizations are also applicable to multicore CPUs. Copyright © 2011 ACM.en
dc.description.sponsorshipThis publication is based on work supported by Award No. UKC0020,made by King Abdullah University of Science and Technology(KAUST).en
dc.publisherAssociation for Computing Machinery (ACM)en
dc.subjectGPUen
dc.subjectPerformance optimizationen
dc.subjectSparse gridsen
dc.titleCompact data structure and scalable algorithms for the sparse grid techniqueen
dc.typeConference Paperen
dc.identifier.journalProceedings of the 16th ACM symposium on Principles and practice of parallel programming - PPoPP '11en
dc.contributor.institutionTechnische Universitat Munchen, Munich, Germanyen
kaust.grant.numberUKC0020en
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.