A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation

Joseph Huber; Weile Wei; Giorgis Georgakoudis; Johannes Doerfert; Oscar Hernandez

doi:10.1007/978-3-030-85262-7_10

A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation

Joseph Huber, Weile Wei, Giorgis Georgakoudis, Johannes Doerfert, Oscar Hernandez

Programming Systems

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Scopus citations

Abstract

This paper presents a methodology for using LLVM-based tools to tune the DCA++ (dynamical cluster approximation) application that targets the new ARM A64FX processor. The goal is to describe the changes required for the new architecture and generate efficient single instruction/multiple data (SIMD) instructions that target the new Scalable Vector Extension instruction set. During manual tuning, the authors used the LLVM tools to improve code parallelization by using OpenMP SIMD, refactored the code and applied transformation that enabled SIMD optimizations, and ensured that the correct libraries were used to achieve optimal performance. By applying these code changes, code speed was increased by 1.98 × and 78 GFlops were achieved on the A64FX processor. The authors aim to automatize parts of the efforts in the OpenMP Advisor tool, which is built on top of existing and newly introduced LLVM tooling.

Original language	English
Title of host publication	OpenMP
Subtitle of host publication	Enabling Massive Node-Level Parallelism - 17th International Workshop on OpenMP, IWOMP 2021, Proceedings
Editors	Simon McIntosh-Smith, Bronis R. de Supinski, Jannis Klinkenberg
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	142-155
Number of pages	14
ISBN (Print)	9783030852610
DOIs	https://doi.org/10.1007/978-3-030-85262-7_10
State	Published - 2021
Event	17th International Workshop on OpenMP, IWOMP 2021 - Bristol, United Kingdom Duration: Sep 14 2021 → Sep 16 2021

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12870 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th International Workshop on OpenMP, IWOMP 2021
Country/Territory	United Kingdom
City	Bristol
Period	09/14/21 → 09/16/21

Funding

Notice: This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy. gov/downloads/doe-public-access-plan). This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344 (LLNL-CONF-819815). This work was supported by the Scientific Discovery through Advanced Computing (SciDAC) program funded by US Department of Energy, Office of Science, Advanced Scientific Computing Research (ASCR) and Basic Energy Sciences (BES) Division of Materials Sciences and Engineering. This research was also supported by the Exas-cale Computing Project (17-SC-20-SC), a collaborative effort of the US Department of Energy Office of Science and the National Nuclear Security Administration, in particular its subproject on Scaling OpenMP with LLVM for Exascale performance and portability (SOLLVE).

Funders	Funder number
U.S. Department of Energy
Office of Science
Basic Energy Sciences
National Nuclear Security Administration
Advanced Scientific Computing Research
Lawrence Livermore National Laboratory	DE-AC52-07NA27344, LLNL-CONF-819815
Division of Materials Sciences and Engineering	17-SC-20-SC

Keywords

Compilers
Feedback
HPC tools
LLVM
OpenMP
SIMD

Access to Document

10.1007/978-3-030-85262-7_10

Cite this

Huber, J., Wei, W., Georgakoudis, G., Doerfert, J., & Hernandez, O. (2021). A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation. In S. McIntosh-Smith, B. R. de Supinski, & J. Klinkenberg (Eds.), OpenMP: Enabling Massive Node-Level Parallelism - 17th International Workshop on OpenMP, IWOMP 2021, Proceedings (pp. 142-155). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12870 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-85262-7_10

Huber, Joseph ; Wei, Weile ; Georgakoudis, Giorgis et al. / A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation. OpenMP: Enabling Massive Node-Level Parallelism - 17th International Workshop on OpenMP, IWOMP 2021, Proceedings. editor / Simon McIntosh-Smith ; Bronis R. de Supinski ; Jannis Klinkenberg. Springer Science and Business Media Deutschland GmbH, 2021. pp. 142-155 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{66ec6ee79c204ae593d3813f861eeeb1,

title = "A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation",

abstract = "This paper presents a methodology for using LLVM-based tools to tune the DCA++ (dynamical cluster approximation) application that targets the new ARM A64FX processor. The goal is to describe the changes required for the new architecture and generate efficient single instruction/multiple data (SIMD) instructions that target the new Scalable Vector Extension instruction set. During manual tuning, the authors used the LLVM tools to improve code parallelization by using OpenMP SIMD, refactored the code and applied transformation that enabled SIMD optimizations, and ensured that the correct libraries were used to achieve optimal performance. By applying these code changes, code speed was increased by 1.98 × and 78 GFlops were achieved on the A64FX processor. The authors aim to automatize parts of the efforts in the OpenMP Advisor tool, which is built on top of existing and newly introduced LLVM tooling.",

keywords = "Compilers, Feedback, HPC tools, LLVM, OpenMP, SIMD",

author = "Joseph Huber and Weile Wei and Giorgis Georgakoudis and Johannes Doerfert and Oscar Hernandez",

note = "Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG.; 17th International Workshop on OpenMP, IWOMP 2021 ; Conference date: 14-09-2021 Through 16-09-2021",

year = "2021",

doi = "10.1007/978-3-030-85262-7_10",

language = "English",

isbn = "9783030852610",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "142--155",

editor = "Simon McIntosh-Smith and {de Supinski}, {Bronis R.} and Jannis Klinkenberg",

booktitle = "OpenMP",

}

Huber, J, Wei, W, Georgakoudis, G, Doerfert, J & Hernandez, O 2021, A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation. in S McIntosh-Smith, BR de Supinski & J Klinkenberg (eds), OpenMP: Enabling Massive Node-Level Parallelism - 17th International Workshop on OpenMP, IWOMP 2021, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12870 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 142-155, 17th International Workshop on OpenMP, IWOMP 2021, Bristol, United Kingdom, 09/14/21. https://doi.org/10.1007/978-3-030-85262-7_10

A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation. / Huber, Joseph; Wei, Weile; Georgakoudis, Giorgis et al.
OpenMP: Enabling Massive Node-Level Parallelism - 17th International Workshop on OpenMP, IWOMP 2021, Proceedings. ed. / Simon McIntosh-Smith; Bronis R. de Supinski; Jannis Klinkenberg. Springer Science and Business Media Deutschland GmbH, 2021. p. 142-155 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12870 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation

AU - Huber, Joseph

AU - Wei, Weile

AU - Georgakoudis, Giorgis

AU - Doerfert, Johannes

AU - Hernandez, Oscar

PY - 2021

Y1 - 2021

N2 - This paper presents a methodology for using LLVM-based tools to tune the DCA++ (dynamical cluster approximation) application that targets the new ARM A64FX processor. The goal is to describe the changes required for the new architecture and generate efficient single instruction/multiple data (SIMD) instructions that target the new Scalable Vector Extension instruction set. During manual tuning, the authors used the LLVM tools to improve code parallelization by using OpenMP SIMD, refactored the code and applied transformation that enabled SIMD optimizations, and ensured that the correct libraries were used to achieve optimal performance. By applying these code changes, code speed was increased by 1.98 × and 78 GFlops were achieved on the A64FX processor. The authors aim to automatize parts of the efforts in the OpenMP Advisor tool, which is built on top of existing and newly introduced LLVM tooling.

AB - This paper presents a methodology for using LLVM-based tools to tune the DCA++ (dynamical cluster approximation) application that targets the new ARM A64FX processor. The goal is to describe the changes required for the new architecture and generate efficient single instruction/multiple data (SIMD) instructions that target the new Scalable Vector Extension instruction set. During manual tuning, the authors used the LLVM tools to improve code parallelization by using OpenMP SIMD, refactored the code and applied transformation that enabled SIMD optimizations, and ensured that the correct libraries were used to achieve optimal performance. By applying these code changes, code speed was increased by 1.98 × and 78 GFlops were achieved on the A64FX processor. The authors aim to automatize parts of the efforts in the OpenMP Advisor tool, which is built on top of existing and newly introduced LLVM tooling.

KW - Compilers

KW - Feedback

KW - HPC tools

KW - LLVM

KW - OpenMP

KW - SIMD

UR - http://www.scopus.com/inward/record.url?scp=85115447703&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-85262-7_10

DO - 10.1007/978-3-030-85262-7_10

M3 - Conference contribution

AN - SCOPUS:85115447703

SN - 9783030852610

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 142

EP - 155

BT - OpenMP

A2 - McIntosh-Smith, Simon

A2 - de Supinski, Bronis R.

A2 - Klinkenberg, Jannis

PB - Springer Science and Business Media Deutschland GmbH

T2 - 17th International Workshop on OpenMP, IWOMP 2021

Y2 - 14 September 2021 through 16 September 2021

ER -

Huber J, Wei W, Georgakoudis G, Doerfert J, Hernandez O. A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation. In McIntosh-Smith S, de Supinski BR, Klinkenberg J, editors, OpenMP: Enabling Massive Node-Level Parallelism - 17th International Workshop on OpenMP, IWOMP 2021, Proceedings. Springer Science and Business Media Deutschland GmbH. 2021. p. 142-155. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-85262-7_10

A Case Study of LLVM-Based Analysis for Optimizing SIMD Code Generation

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this