Evaluation of a performance portable lattice Boltzmann code using OpenCL

Simon N McIntosh-Smith; Dan Curran

doi:10.1145/2664666.2664668

Evaluation of a performance portable lattice Boltzmann code using OpenCL

Simon N McIntosh-Smith, Dan Curran

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

4 Citations (Scopus)

Abstract

With the advent of many-core computer architectures such as GPGPUs from NVIDIA and AMD, and more recently Intel’s Xeon Phi, ensuring performance portability of HPC codes is potentially becoming more complex. In this work we have focused on one important application area — struc- tured grid codes — and investigated techniques exploiting OpenCL to enable performance portability across a diverse range of high-end many-core architectures. In particular we have chosen to investigate 3D lattice Boltzmann codes (D3Q19 BGK). We have developed an OpenCL versions of these codes in order to provide cross-platform functional portability, and compared the performance of this OpenCL version to optimized native versions on each target platform, including hybrid OpenMP/AVX versions on CPUs and Xeon Phi, and CUDA versions on NVIDIA GPUs. Results show that, contrary to conventional wisdom, using OpenCL it is possible to achieve a high degree of performance portabil- ity, at least for 3D lattice Boltzmann codes, using a set of straightforward techniques. The performance portable code in OpenCL is also highly competitive with the best per- formance using the native parallel programming models on each platform.

Original language	English
Title of host publication	Proceedings of the International Workshop on OpenCL 2013 & 2014
Place of Publication	New York, NY
Publisher	Association for Computing Machinery (ACM)
Number of pages	12
ISBN (Electronic)	978-1-4503-3007-7
DOIs	https://doi.org/10.1145/2664666.2664668
Publication status	Published - May 2014
Event	International Workshop on OpenCL (IWOCL) - Bristol, United Kingdom Duration: 12 May 2014 → 13 May 2014

Publication series

Name	ICPS
Publisher	ACM
Volume	2664668

Conference

Conference	International Workshop on OpenCL (IWOCL)
Country/Territory	United Kingdom
City	Bristol
Period	12/05/14 → 13/05/14

Keywords

Many-core, heterogeneous, GPU, Xeon Phi, structured grid, lattice Boltzmann, performance portability

Access to Document

10.1145/2664666.2664668

http://dl.acm.org/citation.cfm?id=2664668

Handle.net

Persistent link

AESArch - Simon McIntosh-Smith joining as a C-I
McIntosh-Smith, S. N. (Principal Investigator)
1/12/13 → 1/04/15
Project: Research
Future-proof massively-parallel execution of multi-block applications
McIntosh-Smith, S. N. (Principal Investigator)
1/10/13 → 31/03/17
Project: Research

HPC (High Performance Computing) and HTC (High Throughput Computing) Facilities
Alam, S. R. (Manager), Williams, D. A. G. (Manager), Eccleston, P. E. (Manager) & Greene, D. (Manager)
Facility/equipment: Facility

Cite this

@inproceedings{78d225de283e46e0a37c04038ded2568,

title = "Evaluation of a performance portable lattice Boltzmann code using OpenCL",

abstract = "With the advent of many-core computer architectures such as GPGPUs from NVIDIA and AMD, and more recently Intel{\textquoteright}s Xeon Phi, ensuring performance portability of HPC codes is potentially becoming more complex. In this work we have focused on one important application area — struc- tured grid codes — and investigated techniques exploiting OpenCL to enable performance portability across a diverse range of high-end many-core architectures. In particular we have chosen to investigate 3D lattice Boltzmann codes (D3Q19 BGK). We have developed an OpenCL versions of these codes in order to provide cross-platform functional portability, and compared the performance of this OpenCL version to optimized native versions on each target platform, including hybrid OpenMP/AVX versions on CPUs and Xeon Phi, and CUDA versions on NVIDIA GPUs. Results show that, contrary to conventional wisdom, using OpenCL it is possible to achieve a high degree of performance portabil- ity, at least for 3D lattice Boltzmann codes, using a set of straightforward techniques. The performance portable code in OpenCL is also highly competitive with the best per- formance using the native parallel programming models on each platform.",

keywords = "Many-core, heterogeneous, GPU, Xeon Phi, structured grid, lattice Boltzmann, performance portability",

author = "McIntosh-Smith, {Simon N} and Dan Curran",

year = "2014",

month = may,

doi = "10.1145/2664666.2664668",

language = "English",

series = "ICPS",

publisher = "Association for Computing Machinery (ACM)",

booktitle = "Proceedings of the International Workshop on OpenCL 2013 & 2014",

address = "United States",

note = "International Workshop on OpenCL (IWOCL) ; Conference date: 12-05-2014 Through 13-05-2014",

}

McIntosh-Smith, SN & Curran, D 2014, Evaluation of a performance portable lattice Boltzmann code using OpenCL. in Proceedings of the International Workshop on OpenCL 2013 & 2014., 2, ICPS, vol. 2664668, Association for Computing Machinery (ACM), New York, NY, International Workshop on OpenCL (IWOCL), Bristol, United Kingdom, 12/05/14. https://doi.org/10.1145/2664666.2664668

Evaluation of a performance portable lattice Boltzmann code using OpenCL. / McIntosh-Smith, Simon N; Curran, Dan.
Proceedings of the International Workshop on OpenCL 2013 & 2014. New York, NY: Association for Computing Machinery (ACM), 2014. 2 (ICPS; Vol. 2664668).

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Evaluation of a performance portable lattice Boltzmann code using OpenCL

AU - McIntosh-Smith, Simon N

AU - Curran, Dan

PY - 2014/5

Y1 - 2014/5

N2 - With the advent of many-core computer architectures such as GPGPUs from NVIDIA and AMD, and more recently Intel’s Xeon Phi, ensuring performance portability of HPC codes is potentially becoming more complex. In this work we have focused on one important application area — struc- tured grid codes — and investigated techniques exploiting OpenCL to enable performance portability across a diverse range of high-end many-core architectures. In particular we have chosen to investigate 3D lattice Boltzmann codes (D3Q19 BGK). We have developed an OpenCL versions of these codes in order to provide cross-platform functional portability, and compared the performance of this OpenCL version to optimized native versions on each target platform, including hybrid OpenMP/AVX versions on CPUs and Xeon Phi, and CUDA versions on NVIDIA GPUs. Results show that, contrary to conventional wisdom, using OpenCL it is possible to achieve a high degree of performance portabil- ity, at least for 3D lattice Boltzmann codes, using a set of straightforward techniques. The performance portable code in OpenCL is also highly competitive with the best per- formance using the native parallel programming models on each platform.

AB - With the advent of many-core computer architectures such as GPGPUs from NVIDIA and AMD, and more recently Intel’s Xeon Phi, ensuring performance portability of HPC codes is potentially becoming more complex. In this work we have focused on one important application area — struc- tured grid codes — and investigated techniques exploiting OpenCL to enable performance portability across a diverse range of high-end many-core architectures. In particular we have chosen to investigate 3D lattice Boltzmann codes (D3Q19 BGK). We have developed an OpenCL versions of these codes in order to provide cross-platform functional portability, and compared the performance of this OpenCL version to optimized native versions on each target platform, including hybrid OpenMP/AVX versions on CPUs and Xeon Phi, and CUDA versions on NVIDIA GPUs. Results show that, contrary to conventional wisdom, using OpenCL it is possible to achieve a high degree of performance portabil- ity, at least for 3D lattice Boltzmann codes, using a set of straightforward techniques. The performance portable code in OpenCL is also highly competitive with the best per- formance using the native parallel programming models on each platform.

KW - Many-core, heterogeneous, GPU, Xeon Phi, structured grid, lattice Boltzmann, performance portability

U2 - 10.1145/2664666.2664668

DO - 10.1145/2664666.2664668

M3 - Conference Contribution (Conference Proceeding)

T3 - ICPS

BT - Proceedings of the International Workshop on OpenCL 2013 & 2014

PB - Association for Computing Machinery (ACM)

CY - New York, NY

T2 - International Workshop on OpenCL (IWOCL)

Y2 - 12 May 2014 through 13 May 2014

ER -

Evaluation of a performance portable lattice Boltzmann code using OpenCL

Abstract

Publication series

Conference

Keywords

Access to Document

Handle.net

Fingerprint

Projects

AESArch - Simon McIntosh-Smith joining as a C-I

Future-proof massively-parallel execution of multi-block applications

Equipment

HPC (High Performance Computing) and HTC (High Throughput Computing) Facilities

Cite this