Analysis of Schedule and Layout Tuning for Sparse Matrices With Compound Entries on GPUs

J. S. Mueller-Roemer; A. Stork; D. Fellner

doi:10.1111/cgf.13957

Analysis of Schedule and Layout Tuning for Sparse Matrices With Compound Entries on GPUs

J. S. Mueller-Roemer^*, A. Stork, D. Fellner

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

Abstract

Large sparse matrices with compound entries, i.e. complex and quaternionic matrices as well as matrices with dense blocks, are a core component of many algorithms in geometry processing, physically based animation and other areas of computer graphics. We generalize several matrix layouts and apply joint schedule and layout autotuning to improve the performance of the sparse matrix-vector product on massively parallel graphics processing units. Compared to schedule tuning without layout tuning, we achieve speedups of up to 5.5 ×. In comparison to cuSPARSE, we achieve speedups of up to 4.7 ×.

Original language	English
Pages (from-to)	133-143
Number of pages	11
Journal	Computer Graphics Forum
Volume	39
Issue number	6
DOIs	https://doi.org/10.1111/cgf.13957
Publication status	Published - 1 Sept 2020

Keywords

GPGPU
parallel computing
sparse matrix
SpMV

ASJC Scopus subject areas

Computer Graphics and Computer-Aided Design

Fields of Expertise

Information, Communication & Computing

Access to Document

10.1111/cgf.13957Licence: CC BY-NC-ND 4.0

Cite this

@article{f94190015c5540edb626b5cf233d4b5d,

title = "Analysis of Schedule and Layout Tuning for Sparse Matrices With Compound Entries on GPUs",

abstract = "Large sparse matrices with compound entries, i.e. complex and quaternionic matrices as well as matrices with dense blocks, are a core component of many algorithms in geometry processing, physically based animation and other areas of computer graphics. We generalize several matrix layouts and apply joint schedule and layout autotuning to improve the performance of the sparse matrix-vector product on massively parallel graphics processing units. Compared to schedule tuning without layout tuning, we achieve speedups of up to 5.5 ×. In comparison to cuSPARSE, we achieve speedups of up to 4.7 ×.",

keywords = "GPGPU, parallel computing, sparse matrix, SpMV",

author = "Mueller-Roemer, {J. S.} and A. Stork and D. Fellner",

year = "2020",

month = sep,

day = "1",

doi = "10.1111/cgf.13957",

language = "English",

volume = "39",

pages = "133--143",

journal = "Computer Graphics Forum",

issn = "0167-7055",

publisher = "Wiley-Blackwell Publishing Ltd",

number = "6",

}

TY - JOUR

T1 - Analysis of Schedule and Layout Tuning for Sparse Matrices With Compound Entries on GPUs

AU - Mueller-Roemer, J. S.

AU - Stork, A.

AU - Fellner, D.

PY - 2020/9/1

Y1 - 2020/9/1

N2 - Large sparse matrices with compound entries, i.e. complex and quaternionic matrices as well as matrices with dense blocks, are a core component of many algorithms in geometry processing, physically based animation and other areas of computer graphics. We generalize several matrix layouts and apply joint schedule and layout autotuning to improve the performance of the sparse matrix-vector product on massively parallel graphics processing units. Compared to schedule tuning without layout tuning, we achieve speedups of up to 5.5 ×. In comparison to cuSPARSE, we achieve speedups of up to 4.7 ×.

AB - Large sparse matrices with compound entries, i.e. complex and quaternionic matrices as well as matrices with dense blocks, are a core component of many algorithms in geometry processing, physically based animation and other areas of computer graphics. We generalize several matrix layouts and apply joint schedule and layout autotuning to improve the performance of the sparse matrix-vector product on massively parallel graphics processing units. Compared to schedule tuning without layout tuning, we achieve speedups of up to 5.5 ×. In comparison to cuSPARSE, we achieve speedups of up to 4.7 ×.

KW - GPGPU

KW - parallel computing

KW - sparse matrix

KW - SpMV

UR - http://www.scopus.com/inward/record.url?scp=85082838571&partnerID=8YFLogxK

U2 - 10.1111/cgf.13957

DO - 10.1111/cgf.13957

M3 - Article

AN - SCOPUS:85082838571

SN - 0167-7055

VL - 39

SP - 133

EP - 143

JO - Computer Graphics Forum

JF - Computer Graphics Forum

IS - 6

ER -

Analysis of Schedule and Layout Tuning for Sparse Matrices With Compound Entries on GPUs

Abstract

Keywords

ASJC Scopus subject areas

Fields of Expertise

Access to Document

Other files and links

Fingerprint

Cite this