SPADIX: A Highly Efficient Accelerator for Solving 3-D Partial Differential Equations
Authors: 
Jiajun Li, Yue DengAuthors Info & Claims
Pages 2797 - 2809
Abstract
Solving partial differential equations (PDEs) holds immense significance in numerous scientific and engineering fields. While analytical solutions to PDEs are often restricted to simple cases, numerical methods offer powerful techniques to approximate solutions for complex PDEs. Previous works have proposed customized accelerators to address the compute- and memory-intensive aspects of numerical PDE solvers. However, these approaches primarily focus on 2-D PDEs and encounter challenges in scaling to support 3-D PDEs due to increased complexity and computational demands. In this article, we introduce Spadix, a highly efficient hardware accelerator designed for numerical 3-D PDE solvers. Spadix leverages a customized processing element (PE) array architecture specifically tailored to the compute and data access patterns in 3-D PDEs. The PE incorporates techniques such as temporal and spatial data reuse to minimize data accesses, enhancing overall performance and energy efficiency. Additionally, Spadix supports the checkerboard method for numerical PDE solvers, which exhibits a faster convergence rate compared to the Jacobi method without compromising parallelism. Our evaluation demonstrates that Spadix achieves an average <inline-formula> <tex-math notation="LaTeX">$8.4\times $ </tex-math></inline-formula> speedup with <inline-formula> <tex-math notation="LaTeX">$9.2\times $ </tex-math></inline-formula> energy reduction over NVIDIA RTX3090 GPU and a <inline-formula> <tex-math notation="LaTeX">$9.7\times $ </tex-math></inline-formula> speedup with <inline-formula> <tex-math notation="LaTeX">$3.2\times $ </tex-math></inline-formula> energy reduction over Alrescha, the state-of-the-art PDE-solving accelerator.
References
[1]
L. C. Evans, Partial Differential Equations, vol. 19. Providence, RI, USA: Amer. Math. Soc., 2022.
[2]
C. L. Fefferman, J. C. Robinson, J. L. R. Diez, and J. L. Rodrigo, Partial Differential Equations in Fluid Mechanics. Cambridge, U.K.: Cambridge Univ. Press, 2018, vol. 452.
[3]
M. M. Bhatti, M. Marin, A. Zeeshan, and S. I. Abdelsalam, “Recent trends in computational fluid dynamics,” Front. Phys., vol. 8, Oct. 2020, Art. no.
[4]
K. Hutter and Y. Wang, Fluid and Thermodynamics. Cham, Switzerland: Springer, 2016.
[5]
T. Rabczuk, H. Ren, and X. Zhuang, “A nonlocal operator method for partial differential equations with application to electromagnetic waveguide problem,” Comput., Mater. Continu, vol. 59, no. 1, pp. 31–55, 2019.
[6]
E. Samaniegoet al., “An energy approach to the solution of partial differential equations in computational mechanics via machine learning: Concepts, implementation and applications,” Comput. Methods Appl. Mech. Eng., vol. 362, Apr. 2020, Art. no.
[7]
J. Mu and B. Kim, “29.2 A 21×21 dynamic-precision bit-serial computing graph accelerator for solving partial differential equations using finite difference method,” in Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC), 2021, pp. 406–408.
[8]
M. N. Özi¸sik, H. R. Orlande, M. J. Colaço, and R. M. Cotta, Finite Difference Methods in Heat Transfer. Boca Raton, FL, USA: CRC Press, 2017.
[9]
J. N. Reddy, Introduction to the Finite Element Method. New York, NY, USA: McGraw-Hill Educ., 2019.
[10]
F. Moukalled, L. Mangani, M. Darwish, F. Moukalled, L. Mangani, and M. Darwish, The Finite Volume Method. Berlin, Germany: Springer, 2016.
[11]
J. Li, Y. Zhang, H. Zheng, and K. Wang, “FDMAX: An elastic accelerator architecture for solving partial differential equations,” in Proc. 50th Annu. Int. Symp. Comput. Architect., 2023, pp. 1–12. [Online]. Available: https://doi.org/10.1145/3579371.3589083
[12]
J. Mu, C. Yu, T. T.-H. Kim, and B. Kim, “A scalable bit-serial computing hardware accelerator for solving 2D/3D partial differential equations using finite difference method,” in Proc. IEEE 48th Eur. Solid State Circuits Conf. (ESSCIRC), 2022, pp. 353–356.
[13]
B. Feinberg, U. K. R. Vengalam, N. Whitehair, S. Wang, and E. Ipek, “Enabling scientific computing on memristive accelerators,” in Proc. ACM/IEEE 45th Annu. Int. Symp. Comput. Architect. (ISCA), 2018, pp. 367–382.
[14]
B. Asgari, R. Hadidi, T. Krishna, H. Kim, and S. Yalamanchili, “ALRESCHA: A lightweight reconfigurable sparse-computation accelerator,” in Proc. IEEE Int. Symp. High Perform. Comput. Architect. (HPCA), 2020, pp. 249–260.
[15]
Y. Chi, J. Cong, P. Wei, and P. Zhou, “SODA: Stencil with optimized dataflow architecture,” in Proc. IEEE/ACM Int. Conf. Comput.-Aided Design (ICCAD), 2018, pp. 1–8.
[16]
G. B. Folland, Introduction to Partial Differential Equations. Princeton, NJ, USA: Princeton Univ. Press, 2020.
[17]
“Nvidia-developer-blog of work.,” Nvidia. 2016. [Online]. Available: https://github.com/NVIDIA-developer-blog/code-samples/tree/master/series/cuda-cpp/finite-difference
[18]
M. ElMaghrbay, R. Ammar, and S. Rajasekaran, “Fast GPU algorithms for implementing the red-black gauss-Seidel method for solving partial differential equations,” in Proc. IEEE Symp. Comput. Commun. (ISCC), 2013, pp. 000269–000274.
[19]
“Measuring processor power: TDP vs. ACP of work.,” Intel. 2011. [Online]. Available: https://www.intel.com/content/dam/doc/white-paper/resources-xeon-measuring-processor-power-paper.pdf
[20]
M. Horowitz. “Energy table for 45nm process of work.,” 2012. [Online]. Available: https://sites.google.com/site/seecproject
[21]
A. Nicholls and B. Honig, “A rapid finite difference algorithm, utilizing successive over-relaxation to solve the poisson–Boltzmann equation,” J. Comput. Chem., vol. 12, no. 4, pp. 435–445, 1991.
[22]
T. Namiki, “A new FDTD algorithm based on alternating-direction implicit method,” IEEE Trans. Microw. Theory Techn., vol. 47, no. 10, pp. 2003–2007, Oct. 1999.
[23]
J. H. Bramble, Multigrid Methods. Boca Raton, FL, USA: CRC Press, 2019.
[24]
J. Blechschmidt and O. G. Ernst, “Three ways to solve partial differential equations with neural networks—A review,” GAMM-Mitteilungen, vol. 44, no. 2, 2021, Art. no.
[25]
Y. Khoo, J. Lu, and L. Ying, “Solving parametric PDE problems with artificial neural networks,” Eur. J. Appl. Math., vol. 32, no. 3, pp. 421–435, 2021.
[26]
M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” J. Comput. Phys., vol. 378, pp. 686–707, Feb. 2019.
[27]
D. Michéa and D. Komatitsch, “Accelerating a three-dimensional finite-difference wave propagation code using GPU graphics cards,” Geophys. J. Int., vol. 182, no. 1, pp. 389–402, 2010.
[28]
M. Giles, E. László, I. Reguly, J. Appleyard, and J. Demouth, “GPU implementation of finite difference solvers,” in Proc. 7th Workshop High Perform. Comput. Finance, 2014, pp. 1–8.
[29]
G. R. Markall, D. A. Ham, and P. H. J. Kelly, “Towards generating optimised finite element solvers for GPUs from high-level specifications,” Procedia Comput. Sci., vol. 1, no. 1, pp. 1815–1823, 2010. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1877050910002048
[30]
A. Balevic, L. Rockstroh, A. Tausendfreund, S. Patzelt, G. Goch, and S. Simon, “Accelerating simulations of light scattering based on finite-difference time-domain method with general purpose GPUs,” in Proc. 11th IEEE Int. Conf. Comput. Sci. Eng., 2008, pp. 327–334.
[31]
A. Dziekonski, P. Sypek, A. Lamecki, and M. Mrozowski, “Finite element matrix generation on a GPU,” Prog. Electromagn. Res., vol. 128, pp. 249–265, May 2012.
[32]
T. Okimura, T. Sasayama, N. Takahashi, and S. Ikuno, “Parallelization of finite element analysis of nonlinear magnetic fields using GPU,” IEEE Trans. Mag., vol. 49, no. 5, pp. 1557–1560, May 2013.
[33]
Z. Fu, T. J. Lewis, R. M. Kirby, and R. T. Whitaker, “Architecting the finite element method pipeline for the GPU,” J. Comput. Appl. Math., vol. 257, pp. 195–211, Feb. 2014.
[34]
P. Huthwaite, “Accelerated finite element elastodynamic simulations using the GPU,” J. Comput. Phys., vol. 257, pp. 687–707, Jan. 2014.
[35]
J. Martínez-Frutos and D. Herrero-Pérez, “Efficient matrix-free GPU implementation of fixed grid finite element analysis,” Finite Elements Anal. Design, vol. 104, pp. 61–71, Oct. 2015.
[36]
G. Rivera and T. Chau-Wen, “Tiling Optimizations for 3D scientific computations,” in Proc. ACM/IEEE Conf. Supercomput., 2000, pp. 1–23.
[37]
D.-M. Dang, C. Christara, and K. R. Jackson. “A parallel implementation on GPUs of ADI finite difference methods for parabolic PDEs with applications in finance.,” 2010. [Online]. Available: https://ssrn.com/abstract=1580057
[38]
B. Hanindhito, D. Gourounas, A. Fathi, D. Trenev, A. Gerstlauer, and L. K. John, “GAPS: GPU-acceleration of PDE solvers for wave simulation,” in Proc. 36th ACM Int. Conf. Supercomput., 2022, pp. 1–13.
[39]
J. P. Durbano and F. E. Ortiz, “FPGA-based acceleration of the 3D finite-difference time-domain method,” in Proc. 12th Annu. IEEE Symp. Field-Program. Custom Comput. Mach., 2004, pp. 156–163.
[40]
Y. Ishigaki, Y. Tomioka, T. Shibata, and H. Kitazawa, “An FPGA implementation of 3D numerical simulations on a 2D SIMD array processor,” in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), 2015, pp. 938–941.
[41]
H. M. Waidyasooriya and M. Hariyama, “FPGA-based deep-pipelined architecture for FDTD acceleration using OpenCL,” in Proc. IEEE/ACIS 15th Int. Conf. Comput. Inf. Sci. (ICIS), 2016, pp. 1–6.
[42]
A. Bocco, Y. Durand, and F. De Dinechin, “SMURF: Scalar multiple-precision unum RISC-V floating-point accelerator for scientific computing,” in Proc. Conf. Next Gener. Arithmetic, 2019, pp. 1–8.
[43]
E. Ipek, “Memristive accelerators for dense and sparse linear algebra: From machine learning to high-performance scientific computing,” IEEE Micro, vol. 39, no. 1, pp. 58–61, Feb. 2019.
[44]
Y. Huang, N. Guo, M. Seok, Y. Tsividis, and S. Sethumadhavan, “Analog computing in a modern context: A linear algebra accelerator case study,” IEEE Micro, vol. 37, no. 3, pp. 30–38, Jun. 2017.
[45]
Y. Huang, N. Guo, M. Seok, Y. Tsividis, K. Mandli, and S. Sethumadhavan, “Hybrid analog-digital solution of nonlinear partial differential equations,” in Proc. 50th Annu. IEEE/ACM Int. Symp. Microarchitect. (MICRO), 2017, pp. 665–678.
Index Terms
- SPADIX: A Highly Efficient Accelerator for Solving 3-D Partial Differential Equations
Index terms have been assigned to the content through auto-classification.
Recommendations
Two-dimensional differential transform for partial differential equations
The differential transform is a numerical method for solving differential equations. In this paper, we present the definition and operation of the two-dimensional differential transform. A distinctive feature of the differential transform is its ability ...
A new framework for solving partial differential equations using semi-analytical explicit RK(N)-type integrators
This paper develops a new explicit semi-analytical approach to solving PDEs based on the variation-of-constants formula, or the equivalent integration equation. These new schemes avoid the discretization of spatial derivatives. Therefore, the accuracy ...
Finite difference approximations for two-sided space-fractional partial differential equations
Fractional order partial differential equations are generalizations of classical partial differential equations. Increasingly, these models are used in applications such as fluid flow, finance and others. In this paper we examine some practical ...
Comments
Please enable JavaScript to view thecomments powered by Disqus.Information & Contributors
Information
Published In
0278-0070 © 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See https://www.ieee.org/publications/rights/index.html for more information.
Publisher
IEEE Press
Publication History
Published: 27 March 2024
Qualifiers
- Research-article
Contributors
Other Metrics
Bibliometrics & Citations
Bibliometrics
Article Metrics
- 0Total Citations
- 0Total Downloads
- Downloads (Last 12 months)0
- Downloads (Last 6 weeks)0
Reflects downloads up to 22 Sep 2024
Other Metrics
Citations
View Options
View options
Get Access
Login options
Check if you have access through your login credentials or your institution to get full access on this article.
Sign in