Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs

Johannes Mueller-Roemer, André Stork, Dieter W. Fellner

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-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 5:5x. In comparison to cuSPARSE, we achieve speedups of up to 4:7x.
Original languageEnglish
Title of host publicationVision, Modeling, and Visualization
PublisherUniversity of Rostock
ISBN (Electronic)978-3-03868-098-7
Publication statusPublished - 2019

Publication series

NameVision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)
PublisherUniversity of Rostock

Keywords

  • Lead Topic: Digitized Work
  • Research Area: (Interactive) simulation (SIM)
  • General Purpose Computation on Graphics Processing Unit (GPGPU)
  • GPU computing
  • Linear systems
  • Code generation

Fields of Expertise

  • Information, Communication & Computing

Cite this

Mueller-Roemer, J., Stork, A., & Fellner, D. W. (2019). Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs. In Vision, Modeling, and Visualization (Vision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)). University of Rostock.

Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs. / Mueller-Roemer, Johannes; Stork, André ; Fellner, Dieter W. .

Vision, Modeling, and Visualization . University of Rostock, 2019. (Vision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)).

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Mueller-Roemer, J, Stork, A & Fellner, DW 2019, Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs. in Vision, Modeling, and Visualization . Vision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019), University of Rostock.
Mueller-Roemer J, Stork A, Fellner DW. Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs. In Vision, Modeling, and Visualization . University of Rostock. 2019. (Vision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)).
Mueller-Roemer, Johannes ; Stork, André ; Fellner, Dieter W. . / Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs. Vision, Modeling, and Visualization . University of Rostock, 2019. (Vision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)).
@inproceedings{a0cd2909bc244ea489d9be439a41ea01,
title = "Joint Schedule and Layout Autotuning 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 5:5x. In comparison to cuSPARSE, we achieve speedups of up to 4:7x.",
keywords = "Lead Topic: Digitized Work, Research Area: (Interactive) simulation (SIM), General Purpose Computation on Graphics Processing Unit (GPGPU), GPU computing, Linear systems, Code generation, Digitized Work, (Interactive) simulation (SIM), GPU computing, Linear systems, Code generation",
author = "Johannes Mueller-Roemer and Andr{\'e} Stork and Fellner, {Dieter W.}",
year = "2019",
language = "English",
series = "Vision, Modeling, and Visualization / von/by Schulz, Hans-J\{\"o}rg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)",
publisher = "University of Rostock",
booktitle = "Vision, Modeling, and Visualization",

}

TY - GEN

T1 - Joint Schedule and Layout Autotuning for Sparse Matrices with Compound Entries on GPUs

AU - Mueller-Roemer, Johannes

AU - Stork, André

AU - Fellner, Dieter W.

PY - 2019

Y1 - 2019

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 5:5x. In comparison to cuSPARSE, we achieve speedups of up to 4:7x.

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 5:5x. In comparison to cuSPARSE, we achieve speedups of up to 4:7x.

KW - Lead Topic: Digitized Work

KW - Research Area: (Interactive) simulation (SIM)

KW - General Purpose Computation on Graphics Processing Unit (GPGPU)

KW - GPU computing

KW - Linear systems

KW - Code generation

KW - Digitized Work

KW - (Interactive) simulation (SIM)

KW - GPU computing

KW - Linear systems

KW - Code generation

M3 - Conference contribution

T3 - Vision, Modeling, and Visualization / von/by Schulz, Hans-J\örg [Ed.] [et al.]. - European Association for Computer Graphics (Eurographics): University of Rostock. - 978-3-03868-098-7 (ISBN). - (2019)

BT - Vision, Modeling, and Visualization

PB - University of Rostock

ER -