Performance analysis and optimization of the JOREK code for many-core CPUs

T.B. Fehér; M. Hölzl; G. Latu; G.T.A. Huijsmans

doi:10.48550/arXiv.1810.04413

Performance analysis and optimization of the JOREK code for many-core CPUs

T.B. Fehér, M. Hölzl, G. Latu, G.T.A. Huijsmans

Research output: Contribution to journal › Article › Academic

48 Downloads (Pure)

Abstract

This report investigates the performance of the JOREK code on the Intel Knights Landing and Skylake processor architectures. The OpenMP scaling of the matrix construction part of the code was analyzed and improved synchronization methods were implemented. A new switch was implemented to control the number of threads used for the linear equation solver independently from other parts of the code. The matrix construction subroutine was vectorized, and the data locality was also improved. These steps led to a factor of two speedup for the matrix construction.

Original language	English
Article number	1810.04413
Number of pages	15
Journal	arXiv
Volume	2018
DOIs	https://doi.org/10.48550/arXiv.1810.04413
Publication status	Published - 1 Oct 2018

Keywords

Computer Science - Performance

Access to Document

10.48550/arXiv.1810.04413

published versionFinal published version, 732 KB

Cite this

@article{63268e16fa8a40d9a748d9ddb5e41a67,

title = "Performance analysis and optimization of the JOREK code for many-core CPUs",

abstract = "This report investigates the performance of the JOREK code on the Intel Knights Landing and Skylake processor architectures. The OpenMP scaling of the matrix construction part of the code was analyzed and improved synchronization methods were implemented. A new switch was implemented to control the number of threads used for the linear equation solver independently from other parts of the code. The matrix construction subroutine was vectorized, and the data locality was also improved. These steps led to a factor of two speedup for the matrix construction.",

keywords = "Computer Science - Performance",

author = "T.B. Feh{\'e}r and M. H{\"o}lzl and G. Latu and G.T.A. Huijsmans",

year = "2018",

month = oct,

day = "1",

doi = "10.48550/arXiv.1810.04413",

language = "English",

volume = "2018",

journal = "arXiv",

issn = "2331-8422",

publisher = "Cornell University Library",

}

TY - JOUR

T1 - Performance analysis and optimization of the JOREK code for many-core CPUs

AU - Fehér, T.B.

AU - Hölzl, M.

AU - Latu, G.

AU - Huijsmans, G.T.A.

PY - 2018/10/1

Y1 - 2018/10/1

N2 - This report investigates the performance of the JOREK code on the Intel Knights Landing and Skylake processor architectures. The OpenMP scaling of the matrix construction part of the code was analyzed and improved synchronization methods were implemented. A new switch was implemented to control the number of threads used for the linear equation solver independently from other parts of the code. The matrix construction subroutine was vectorized, and the data locality was also improved. These steps led to a factor of two speedup for the matrix construction.

AB - This report investigates the performance of the JOREK code on the Intel Knights Landing and Skylake processor architectures. The OpenMP scaling of the matrix construction part of the code was analyzed and improved synchronization methods were implemented. A new switch was implemented to control the number of threads used for the linear equation solver independently from other parts of the code. The matrix construction subroutine was vectorized, and the data locality was also improved. These steps led to a factor of two speedup for the matrix construction.

KW - Computer Science - Performance

U2 - 10.48550/arXiv.1810.04413

DO - 10.48550/arXiv.1810.04413

M3 - Article

SN - 2331-8422

VL - 2018

JO - arXiv

JF - arXiv

M1 - 1810.04413

ER -

Performance analysis and optimization of the JOREK code for many-core CPUs

Abstract

Keywords

Access to Document

Fingerprint

Cite this