Towards Automated Kernel Fusion for the Optimisation of Scientific Applications

Andrew Lamzed-Short; Timothy R. Law; Andrew Mallinson; Gihan R. Mudalige; Stephen A. Jarvis

doi:10.1109/LLVMHPCHiPar51896.2020.00010

Towards Automated Kernel Fusion for the Optimisation of Scientific Applications

Andrew Lamzed-Short^*, Timothy R. Law, Andrew Mallinson, Gihan R. Mudalige, Stephen A. Jarvis

^*Corresponding author for this work

Engineering and Physical Sciences

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

Abstract

In this paper we introduce a novel transformation pass written using LLVM that performs kernel fusion. We demonstrate the correctness and performance of the pass on several example programs inspired by scientific applications of interest. The method achieves up to 4× speedup relative to unfused versions of the programs, and exact performance parity with manually fused versions. In contrast to previous work, it also requires minimal user intervention. Our approach is facilitated by a new loop fusion algorithm capable of interprocedurally fusing both skewed and unskewed loops in different kernels.

Original language	English
Title of host publication	Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020
Subtitle of host publication	The International Conference for High Performance Computing, Networking, Storage and Analysis
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Pages	45-55
Number of pages	11
ISBN (Electronic)	9780738110424
DOIs	https://doi.org/10.1109/LLVMHPCHiPar51896.2020.00010
Publication status	Published - Nov 2020
Event	6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, LLVM-HPC2020 and HiPar 2020 - Virtual, Atlanta, United States Duration: 12 Nov 2020 → …

Publication series

Name	Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis

Conference

Conference	6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, LLVM-HPC2020 and HiPar 2020
Country/Territory	United States
City	Virtual, Atlanta
Period	12/11/20 → …

Bibliographical note

Publisher Copyright:
© 2020 IEEE.

Keywords

HPC
kernel-fusion
LLVM
loop-fusion
optimisation

ASJC Scopus subject areas

Hardware and Architecture
Control and Optimization

Access to Document

10.1109/LLVMHPCHiPar51896.2020.00010

Cite this

Lamzed-Short, A., Law, T. R., Mallinson, A., Mudalige, G. R., & Jarvis, S. A. (2020). Towards Automated Kernel Fusion for the Optimisation of Scientific Applications. In Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis (pp. 45-55). Article 9307048 (Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis). Institute of Electrical and Electronics Engineers (IEEE). https://doi.org/10.1109/LLVMHPCHiPar51896.2020.00010

Lamzed-Short, Andrew ; Law, Timothy R. ; Mallinson, Andrew et al. / Towards Automated Kernel Fusion for the Optimisation of Scientific Applications. Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers (IEEE), 2020. pp. 45-55 (Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis).

@inproceedings{348cbc52ca9d45fbb0f357aa0773623f,

title = "Towards Automated Kernel Fusion for the Optimisation of Scientific Applications",

abstract = "In this paper we introduce a novel transformation pass written using LLVM that performs kernel fusion. We demonstrate the correctness and performance of the pass on several example programs inspired by scientific applications of interest. The method achieves up to 4× speedup relative to unfused versions of the programs, and exact performance parity with manually fused versions. In contrast to previous work, it also requires minimal user intervention. Our approach is facilitated by a new loop fusion algorithm capable of interprocedurally fusing both skewed and unskewed loops in different kernels. ",

keywords = "HPC, kernel-fusion, LLVM, loop-fusion, optimisation",

author = "Andrew Lamzed-Short and Law, {Timothy R.} and Andrew Mallinson and Mudalige, {Gihan R.} and Jarvis, {Stephen A.}",

note = "Publisher Copyright: {\textcopyright} 2020 IEEE.; 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, LLVM-HPC2020 and HiPar 2020 ; Conference date: 12-11-2020",

year = "2020",

month = nov,

doi = "10.1109/LLVMHPCHiPar51896.2020.00010",

language = "English",

series = "Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis",

publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

pages = "45--55",

booktitle = "Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020",

}

Lamzed-Short, A, Law, TR, Mallinson, A, Mudalige, GR & Jarvis, SA 2020, Towards Automated Kernel Fusion for the Optimisation of Scientific Applications. in Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis., 9307048, Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis, Institute of Electrical and Electronics Engineers (IEEE), pp. 45-55, 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, LLVM-HPC2020 and HiPar 2020, Virtual, Atlanta, United States, 12/11/20. https://doi.org/10.1109/LLVMHPCHiPar51896.2020.00010

Towards Automated Kernel Fusion for the Optimisation of Scientific Applications. / Lamzed-Short, Andrew; Law, Timothy R.; Mallinson, Andrew et al.
Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers (IEEE), 2020. p. 45-55 9307048 (Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis).

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

TY - GEN

T1 - Towards Automated Kernel Fusion for the Optimisation of Scientific Applications

AU - Lamzed-Short, Andrew

AU - Law, Timothy R.

AU - Mallinson, Andrew

AU - Mudalige, Gihan R.

AU - Jarvis, Stephen A.

PY - 2020/11

Y1 - 2020/11

N2 - In this paper we introduce a novel transformation pass written using LLVM that performs kernel fusion. We demonstrate the correctness and performance of the pass on several example programs inspired by scientific applications of interest. The method achieves up to 4× speedup relative to unfused versions of the programs, and exact performance parity with manually fused versions. In contrast to previous work, it also requires minimal user intervention. Our approach is facilitated by a new loop fusion algorithm capable of interprocedurally fusing both skewed and unskewed loops in different kernels.

AB - In this paper we introduce a novel transformation pass written using LLVM that performs kernel fusion. We demonstrate the correctness and performance of the pass on several example programs inspired by scientific applications of interest. The method achieves up to 4× speedup relative to unfused versions of the programs, and exact performance parity with manually fused versions. In contrast to previous work, it also requires minimal user intervention. Our approach is facilitated by a new loop fusion algorithm capable of interprocedurally fusing both skewed and unskewed loops in different kernels.

KW - HPC

KW - kernel-fusion

KW - LLVM

KW - loop-fusion

KW - optimisation

UR - http://www.scopus.com/inward/record.url?scp=85100244286&partnerID=8YFLogxK

U2 - 10.1109/LLVMHPCHiPar51896.2020.00010

DO - 10.1109/LLVMHPCHiPar51896.2020.00010

M3 - Conference contribution

AN - SCOPUS:85100244286

T3 - Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis

SP - 45

EP - 55

BT - Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020

PB - Institute of Electrical and Electronics Engineers (IEEE)

T2 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, LLVM-HPC2020 and HiPar 2020

Y2 - 12 November 2020

ER -

Lamzed-Short A, Law TR, Mallinson A, Mudalige GR, Jarvis SA. Towards Automated Kernel Fusion for the Optimisation of Scientific Applications. In Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers (IEEE). 2020. p. 45-55. 9307048. (Proceedings of LLVM-HPC2020 and HiPar 2020 - 6th Workshop on the LLVM Compiler Infrastructure in HPC and Workshop on Hierarchical Parallelism for Exascale Computing, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis). doi: 10.1109/LLVMHPCHiPar51896.2020.00010

Towards Automated Kernel Fusion for the Optimisation of Scientific Applications

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this