Abstract
In this paper, we perform an extensive benchmarking and analysis of the performance and scalability of our software tool called CFD suite, which implements the AI-based domain-specific method for accelerating CFD (computation fluid dynamic) simulations proposed by us recently. By exploring various computing platforms containing both CPUs and GPUs, this analysis helps select suitable platforms for training and inference stages across heterogeneous execution environments. We propose and investigate two modes of utilizing the proposed decomposition of the AI model at the inference stage – either by calling each sub-model one by one (on GPUs) with reduced memory requirements or by performing pipeline predictions (on CPUs with large RAM) to improve the overall performance. It is shown that for the whole inference stage (including overheads), due to the pipeline execution and excluding overheads for data transfers through PCIe, the speedup provided by two Intel Xeon Gold CPUs (Skylake) is 2.4 times higher than for V100 GPU.
The authors are grateful to the byteLAKE company for their substantive support. The project financed under the program of the Minister of Science and Higher Education under the name “Regional Initiative of Excellence” in the years 2019–2022 project number 020/RID/2018/19 the amount of financing 12,000,000 PLN.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Archibald, R., et al.: Integrating deep learning in domain sciences at exascale. ar**v preprint ar**v:2011.11188v1 (2020)
BEM overview, September 2022. https://kdm.wcss.pl/wiki/Bem_overview
Bhatt, D., Zhang, B.W., Zuckerman, D.M.: Steady-state simulations using weighted ensemble path sampling. J. Chem. Phys. 133(1), 014110 (2010)
Chen, D., Hu, F., Nian, G., Yang, T.: Deep residual learning for nonlinear regression. Entropy 22(2), 193 (2020)
Paul, E.L., Atiemo-Obeng, V., Kresta, S.M. (eds.): Wiley, Hoboken (2004)
Horovod: Home, May 2022. https://horovod.ai/
Kim, B., et al.: Deep fluids: a generative network for parameterized fluid simulations. ar**v preprint ar**v:1806.02071v2 (2019)
Maulik, R., Sharma, H., Patel, S., Lusch, B., Jennings, E.: Accelerating RANS turbulence modeling using potential flow and machine learning. ar**v preprint ar**v:1910.10878 (2019)
Mattson, P., et al.: MLPerf training benchmark. ar**v preprint ar**v:1910.01500v3 (2020)
MixIT: the enterprise mixing analysis tool. https://mixing-solution.com/
MLPerf Benchmarks, April 2022. https://www.nvidia.com/en-us/data-center/resources/mlperf-benchmarks/
Mostafazadeh, B., et al.: Unsteady Navier-Stokes computations on GPU architectures. In: 23rd AIAA Computational Fluid Dynamics Conference (2017)
Obiols-Sales, O., Vishnu, A., Malaya, N., Chandramowlishwaran, A.: CFDNet: a deep learning-based accelerator for fluid simulations. In: Proceedings of the 34th ACM International Conference on Supercomputing (ICS 2020), pp. 1–12. ACM (2020)
OpenFOAM: the open source CFD toolbox, May 2022. https://www.openfoam.com
The Performance of MLPerf as a Ubiquitous Benchmark is Lacking, 8 April 2022. https://www.nextplatform.com/2022/04/08/the-performance-of-mlperf-as-a-ubiquitous-benchmark-is-lacking
Reddi, V.J., et al.: MLPerf inference benchmark. ar**v preprint ar**v:1911.02549v2 (2020)
Rojek, K., et al.: Adaptation of fluid model EULAG to graphics processing unit architecture. Concurr. Comput. Pract. Exp. 27(4), 937–957 (2015)
OpenVINO, May 2022. https://docs.openvino.ai
Rojek, K., Halbiniak, K., Kuczynski, L.: CFD code adaptation to the FPGA architecture. Int. J. High Perform. Comput. Appl. 35(1), 33–46 (2021)
Rojek, K., Wyrzykowski, R., Gepner, P.: AI-accelerated CFD simulation based on OpenFOAM and CPU/GPU computing. In: Paszynski, M., Kranzlmüller, D., Krzhizhanovskaya, V.V., Dongarra, J.J., Sloot, P.M.A. (eds.) ICCS 2021. LNCS, vol. 12743, pp. 373–385. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-77964-1_29
Sadrehaghighi, I.: Basics of Computer Architecture as Relates to CFD, January 2022. https://www.researchgate.net/publication/339212886
Sadrehaghighi, I.: Artificial Intelligence (AI) and Deep Learning for CFD, January 2022. https://www.researchgate.net/publication/339795951
Sanchez-Gonzalez, A., et al.: Learning to simulate complex physics with graph networks. ar**v preprint ar**v:3394.45567 (2020)
Szustak, L., et al.: Adaptation of MPDATA heterogeneous stencil computation to Intel Xeon Phi coprocessor. Sci. Program. (2015). https://doi.org/10.1155/2015/642705
Tompson, J., Schlachter, K., Sprechmann, P., Perlin, K.H.: Accelerating eulerian fluid simulation with convolutional networks. In: ICML 2017: Proceedings of the 34th International Conference on Machine Learning, vol. 70, pp. 3424–3433 (2017)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Rojek, K., Wyrzykowski, R. (2023). Performance and Scalability Analysis of AI-Accelerated CFD Simulations Across Various Computing Platforms. In: Singer, J., Elkhatib, Y., Blanco Heras, D., Diehl, P., Brown, N., Ilic, A. (eds) Euro-Par 2022: Parallel Processing Workshops. Euro-Par 2022. Lecture Notes in Computer Science, vol 13835. Springer, Cham. https://doi.org/10.1007/978-3-031-31209-0_17
Download citation
DOI: https://doi.org/10.1007/978-3-031-31209-0_17
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-31208-3
Online ISBN: 978-3-031-31209-0
eBook Packages: Computer ScienceComputer Science (R0)