Evaluating LULESH Kernels on OpenCL FPGA

Zheming Jin; Hal Finkel

doi:10.1007/978-3-030-17227-5_15

Evaluating LULESH Kernels on OpenCL FPGA

Zheming Jin, Hal Finkel

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

3 Scopus citations

Abstract

FPGAs are becoming promising heterogeneous computing components for high-performance computing. In this paper, we evaluate the resource utilizations, performance, and performance per watt of our implementations of the LULESH kernels in OpenCL on an Arria10-based FPGA platform. LULESH is a complex proxy application in the CORAL benchmark suite. We choose two representative kernels “CalcFBHourglassForceForElems” and “EvalEOSForElems” from the application in our study. Compared with the baseline implementations, our optimizations improve the performance by a factor of 1.65X and 2.96X for the two kernels on the FPGA, respectively. Using directives for accelerator programming, we also evaluate the performance of the kernels on an Intel Xeon 16-core CPU and an Nvidia K80 GPU. We find that the FPGA, constrained by the memory bandwidth, can perform 1.05X to 3.4X better than the CPU and GPU for small problem sizes. For the first kernel, the performance per watt on the FPGA is 1.59X and 7.1X higher than that on an Intel Xeon 16-core CPU and an Nvidia K80 GPU, respectively. For the second kernel, the performance per watt on the GPU is 1.82X higher than that on the FPGA. However, the performance per watt on the FPGA is 1.77X higher than that on the CPU.

Original language	English
Title of host publication	Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings
Editors	Christian Hochberger, Andreas Koch, Pedro Diniz, Brent Nelson, Roger Woods
Publisher	Springer Verlag
Pages	199-213
Number of pages	15
ISBN (Print)	9783030172268
DOIs	https://doi.org/10.1007/978-3-030-17227-5_15
State	Published - 2019
Externally published	Yes
Event	15th International Symposium on Applied Reconfigurable Computing, ARC 2019 - Darmstadt, Germany Duration: Apr 9 2019 → Apr 11 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11444 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th International Symposium on Applied Reconfigurable Computing, ARC 2019
Country/Territory	Germany
City	Darmstadt
Period	04/9/19 → 04/11/19

Funding

Acknowledgments. The research was supported by the U.S. Department of Energy, Office of Science, under contract DE-AC02-06CH11357 and made use of the Argonne Leadership Computing Facility, a DOE Office of Science User Facility.

Keywords

FPGA
Kernel optimizations
LULESH
OpenCL

Access to Document

10.1007/978-3-030-17227-5_15

Cite this

Jin, Z., & Finkel, H. (2019). Evaluating LULESH Kernels on OpenCL FPGA. In C. Hochberger, A. Koch, P. Diniz, B. Nelson, & R. Woods (Eds.), Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings (pp. 199-213). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11444 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-17227-5_15

Jin, Zheming ; Finkel, Hal. / Evaluating LULESH Kernels on OpenCL FPGA. Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings. editor / Christian Hochberger ; Andreas Koch ; Pedro Diniz ; Brent Nelson ; Roger Woods. Springer Verlag, 2019. pp. 199-213 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{b7b7fc353b7a4e29aff8fa216a7d61a0,

title = "Evaluating LULESH Kernels on OpenCL FPGA",

abstract = "FPGAs are becoming promising heterogeneous computing components for high-performance computing. In this paper, we evaluate the resource utilizations, performance, and performance per watt of our implementations of the LULESH kernels in OpenCL on an Arria10-based FPGA platform. LULESH is a complex proxy application in the CORAL benchmark suite. We choose two representative kernels “CalcFBHourglassForceForElems” and “EvalEOSForElems” from the application in our study. Compared with the baseline implementations, our optimizations improve the performance by a factor of 1.65X and 2.96X for the two kernels on the FPGA, respectively. Using directives for accelerator programming, we also evaluate the performance of the kernels on an Intel Xeon 16-core CPU and an Nvidia K80 GPU. We find that the FPGA, constrained by the memory bandwidth, can perform 1.05X to 3.4X better than the CPU and GPU for small problem sizes. For the first kernel, the performance per watt on the FPGA is 1.59X and 7.1X higher than that on an Intel Xeon 16-core CPU and an Nvidia K80 GPU, respectively. For the second kernel, the performance per watt on the GPU is 1.82X higher than that on the FPGA. However, the performance per watt on the FPGA is 1.77X higher than that on the CPU.",

keywords = "FPGA, Kernel optimizations, LULESH, OpenCL",

author = "Zheming Jin and Hal Finkel",

note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Switzerland AG.; 15th International Symposium on Applied Reconfigurable Computing, ARC 2019 ; Conference date: 09-04-2019 Through 11-04-2019",

year = "2019",

doi = "10.1007/978-3-030-17227-5\_15",

language = "English",

isbn = "9783030172268",

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

publisher = "Springer Verlag",

pages = "199--213",

editor = "Christian Hochberger and Andreas Koch and Pedro Diniz and Brent Nelson and Roger Woods",

booktitle = "Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings",

}

Jin, Z & Finkel, H 2019, Evaluating LULESH Kernels on OpenCL FPGA. in C Hochberger, A Koch, P Diniz, B Nelson & R Woods (eds), Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11444 LNCS, Springer Verlag, pp. 199-213, 15th International Symposium on Applied Reconfigurable Computing, ARC 2019, Darmstadt, Germany, 04/9/19. https://doi.org/10.1007/978-3-030-17227-5_15

Evaluating LULESH Kernels on OpenCL FPGA. / Jin, Zheming; Finkel, Hal.
Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings. ed. / Christian Hochberger; Andreas Koch; Pedro Diniz; Brent Nelson; Roger Woods. Springer Verlag, 2019. p. 199-213 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11444 LNCS).

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

TY - GEN

T1 - Evaluating LULESH Kernels on OpenCL FPGA

AU - Jin, Zheming

AU - Finkel, Hal

PY - 2019

Y1 - 2019

N2 - FPGAs are becoming promising heterogeneous computing components for high-performance computing. In this paper, we evaluate the resource utilizations, performance, and performance per watt of our implementations of the LULESH kernels in OpenCL on an Arria10-based FPGA platform. LULESH is a complex proxy application in the CORAL benchmark suite. We choose two representative kernels “CalcFBHourglassForceForElems” and “EvalEOSForElems” from the application in our study. Compared with the baseline implementations, our optimizations improve the performance by a factor of 1.65X and 2.96X for the two kernels on the FPGA, respectively. Using directives for accelerator programming, we also evaluate the performance of the kernels on an Intel Xeon 16-core CPU and an Nvidia K80 GPU. We find that the FPGA, constrained by the memory bandwidth, can perform 1.05X to 3.4X better than the CPU and GPU for small problem sizes. For the first kernel, the performance per watt on the FPGA is 1.59X and 7.1X higher than that on an Intel Xeon 16-core CPU and an Nvidia K80 GPU, respectively. For the second kernel, the performance per watt on the GPU is 1.82X higher than that on the FPGA. However, the performance per watt on the FPGA is 1.77X higher than that on the CPU.

AB - FPGAs are becoming promising heterogeneous computing components for high-performance computing. In this paper, we evaluate the resource utilizations, performance, and performance per watt of our implementations of the LULESH kernels in OpenCL on an Arria10-based FPGA platform. LULESH is a complex proxy application in the CORAL benchmark suite. We choose two representative kernels “CalcFBHourglassForceForElems” and “EvalEOSForElems” from the application in our study. Compared with the baseline implementations, our optimizations improve the performance by a factor of 1.65X and 2.96X for the two kernels on the FPGA, respectively. Using directives for accelerator programming, we also evaluate the performance of the kernels on an Intel Xeon 16-core CPU and an Nvidia K80 GPU. We find that the FPGA, constrained by the memory bandwidth, can perform 1.05X to 3.4X better than the CPU and GPU for small problem sizes. For the first kernel, the performance per watt on the FPGA is 1.59X and 7.1X higher than that on an Intel Xeon 16-core CPU and an Nvidia K80 GPU, respectively. For the second kernel, the performance per watt on the GPU is 1.82X higher than that on the FPGA. However, the performance per watt on the FPGA is 1.77X higher than that on the CPU.

KW - FPGA

KW - Kernel optimizations

KW - LULESH

KW - OpenCL

UR - https://www.scopus.com/pages/publications/85064543083

U2 - 10.1007/978-3-030-17227-5_15

DO - 10.1007/978-3-030-17227-5_15

M3 - Conference contribution

AN - SCOPUS:85064543083

SN - 9783030172268

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

SP - 199

EP - 213

BT - Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings

A2 - Hochberger, Christian

A2 - Koch, Andreas

A2 - Diniz, Pedro

A2 - Nelson, Brent

A2 - Woods, Roger

PB - Springer Verlag

T2 - 15th International Symposium on Applied Reconfigurable Computing, ARC 2019

Y2 - 9 April 2019 through 11 April 2019

ER -

Jin Z, Finkel H. Evaluating LULESH Kernels on OpenCL FPGA. In Hochberger C, Koch A, Diniz P, Nelson B, Woods R, editors, Applied Reconfigurable Computing - 15th International Symposium, ARC 2019, Proceedings. Springer Verlag. 2019. p. 199-213. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-17227-5_15

Evaluating LULESH Kernels on OpenCL FPGA

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this