Solving RFIC Simulation Tasks Using GPU Computations
Andrei Mihail-Iulian 1, Kula Sebastian 2
1Politehnica University of Bucharest, Electrical Engineering Faculty, Numerical Methods Laboratory, LMN
Splaiul Independentei 313, 060042 Bucharest,Romania
E-mail: iulian@lmn.pub.ro
2Kazimierz Wielki University, Institute of Mechanics and Applied Computer Science
ul. Kopernika 1, 85-074 Bydgoszcz, Poland
Received:
(Received: 12 November 2012; revised: 19 February 2013; accepted: 19 February 2013; published online: 14 March 2013)
DOI: 10.12921/cmst.2013.19.02.89-97
OAI: oai:lib.psnc.pl:433
Abstract:
New generation of General Purpose Graphic Processing Unit (GPGPU) cards with their large computation power allow to approach difficult tasks from Radio Frequency Integrated Circuits (RFICs) modeling area. Using different electromagnetic modeling methods, the Finite Element Method (FEM) and the Finite Integration Technique (FIT), to model Radio Frequency Integrated Circuit (RFIC) devices, large linear equations systems have to be solved. This paper presents the benefits of using Graphic Processing Unit (GPU) computations for solving such systems which are characterized by sparse complex matrices. CUSP is a GPU generic parallel algorithms library for sparse linear algebra and graph computations based on Compute Unified Device Architecture (CUDA). The code is calling iterative methods available in CUSP in order to solve those complex linear equation systems. The tests were performed on various Central Processing Units (CPU) and GPU hardware configurations. The results of these tests show that using GPU computations for solving the linear equations systems, the electromagnetic modeling process of RFIC devices can be accelerated and at the same time a high level of computation accuracy is maintained. Tests were carried out on matrices obtained for an integrated inductor designed for RFICs, and for Micro Stripe (MS) designed for Photonics Integrated Circuit (PIC).
Key words:
GPGPU computing, iterative methods, Photonics Integrated Circuits, Radio Frequency Integrated Circuits
References:
[1] E. Lindholm, J. Nickolls, S. Oberman, J. Montrym, NVIDIA Tesla: A Unified Graphics and Computing Architecture, Micro, IEEE, 39 – 55 (2008).
[2] K.K. Matam, S.R.K. Bharadwaj, and K. Kothapalli, Sparse Matrix Matrix Multiplication on Hybrid CPU+GPU Platforms, in Proc. of 19th Annual International Conference on High Performance Computing (HiPC), Pune, India, (2012).
[3] F. Ellinger, Radio Frequency Integrated Circuits and Technologies, Springer, 2nd edition 2008.
[4] I.-A. Lazar, G. Ciuprina, and D. Ioan, Effective extraction of accurate reduced order models for hf-ic using multi-CPU arhitectures, Inverse Problems in Science and Engineering, 1-13 (2011).
[5] T. Weiland, A discretisation method for the solution of Maxwell’s equations for six-component fields, International Journal of Electronics and Communication AEU 31 116–120 (1977).
[6] G. Ciuprina, D. Ioan, D. Mihalache, Magnetic Hooks in the Finite Integration Technique: A Way Towards Domain Decomposition, Proceedings of the IEEE CEFC, 2008.
[7] G. Ciuprina, D. Ioan, D. Mihalache, and E. Seebacher, Domain partitioning based parametric models for passive on-chip components, Scientific Computing in Electrical Engineering, in the series Mathematics in Industry (J. Roos, L. Costa Eds), Vol. 14, pp. 37-44, Springer, 2010.
[8] LiveLin for MATLAB User’s Guide COMSOL 2011.
[9] B. Gustavsen and A. Semlyen, Rational approximation of frequency domain responses by vector fitting, IEEE Trans. Power Delivery, 1052-1061 (1999).
[10] I.-A. Lazar, M.-I. Andrei, E. Caciulan, G. Ciuprina, and D. Ioan, Parallel algorithms for the efficient extraction of fitting based reduced order models, Proceedings of the 7th International Symposium on ADVANCED TOPICS IN ELECTRICAL ENGINEERING, 1–13 (2011).
[11] T. Davis, Algorithm 832: Umfpack, an unsymmetricpattern multifrontal method, ACM Transactions on Mathematical Software (TOMS), 196-199 (2004).
[12] Y. Saad, Iterative Methods for Sparse Linear Systems, Society for Industrial and Applied Mathematics, 2003. Second edition with corrections.
[13] Y. Saad, and M. H. Schultz, GMRES: A generalized minimal residual algorithm for solving nonsymmetric linear systems, SIAM J. Sci. Stat. Comput., 856-869 (1986).
[14] J. Sanders and E. Kandrot, CUDA by Example: An Introduction to General-Purpose GPU Programming, Addison-Wesley Professional; 1st edition 2010.
[15] D. B. Kirk and W. mei W. Hwu, Programming Massively Parallel Processors: A Hands-on Approach (Applications of GPU Computing Series), Morgan Kaufman Elsevier, 2010.
[16] N. Bell and M. Garland, Cusp: Generic parallel algorithms for sparse matrix and graph computations, 2012. Version 0.3.0.
[17] H. H. J. M. Janssen, J. Niehof, W. H. A. Schilders, Accurate Modeling of Complete Functional RF Blocks: CHAMELEON RF, Scientific Computing in Electrical Engineering, Mathematics in Industry, 81-87 (2007).
[18] S. Kula, Interconnect Elements Propagation Quantities in PIC, Poznan University of Technology Academic Journals, Series Electrical Engineering, Iss. 70, 83-89 (2012).
New generation of General Purpose Graphic Processing Unit (GPGPU) cards with their large computation power allow to approach difficult tasks from Radio Frequency Integrated Circuits (RFICs) modeling area. Using different electromagnetic modeling methods, the Finite Element Method (FEM) and the Finite Integration Technique (FIT), to model Radio Frequency Integrated Circuit (RFIC) devices, large linear equations systems have to be solved. This paper presents the benefits of using Graphic Processing Unit (GPU) computations for solving such systems which are characterized by sparse complex matrices. CUSP is a GPU generic parallel algorithms library for sparse linear algebra and graph computations based on Compute Unified Device Architecture (CUDA). The code is calling iterative methods available in CUSP in order to solve those complex linear equation systems. The tests were performed on various Central Processing Units (CPU) and GPU hardware configurations. The results of these tests show that using GPU computations for solving the linear equations systems, the electromagnetic modeling process of RFIC devices can be accelerated and at the same time a high level of computation accuracy is maintained. Tests were carried out on matrices obtained for an integrated inductor designed for RFICs, and for Micro Stripe (MS) designed for Photonics Integrated Circuit (PIC).
Key words:
GPGPU computing, iterative methods, Photonics Integrated Circuits, Radio Frequency Integrated Circuits
References:
[1] E. Lindholm, J. Nickolls, S. Oberman, J. Montrym, NVIDIA Tesla: A Unified Graphics and Computing Architecture, Micro, IEEE, 39 – 55 (2008).
[2] K.K. Matam, S.R.K. Bharadwaj, and K. Kothapalli, Sparse Matrix Matrix Multiplication on Hybrid CPU+GPU Platforms, in Proc. of 19th Annual International Conference on High Performance Computing (HiPC), Pune, India, (2012).
[3] F. Ellinger, Radio Frequency Integrated Circuits and Technologies, Springer, 2nd edition 2008.
[4] I.-A. Lazar, G. Ciuprina, and D. Ioan, Effective extraction of accurate reduced order models for hf-ic using multi-CPU arhitectures, Inverse Problems in Science and Engineering, 1-13 (2011).
[5] T. Weiland, A discretisation method for the solution of Maxwell’s equations for six-component fields, International Journal of Electronics and Communication AEU 31 116–120 (1977).
[6] G. Ciuprina, D. Ioan, D. Mihalache, Magnetic Hooks in the Finite Integration Technique: A Way Towards Domain Decomposition, Proceedings of the IEEE CEFC, 2008.
[7] G. Ciuprina, D. Ioan, D. Mihalache, and E. Seebacher, Domain partitioning based parametric models for passive on-chip components, Scientific Computing in Electrical Engineering, in the series Mathematics in Industry (J. Roos, L. Costa Eds), Vol. 14, pp. 37-44, Springer, 2010.
[8] LiveLin for MATLAB User’s Guide COMSOL 2011.
[9] B. Gustavsen and A. Semlyen, Rational approximation of frequency domain responses by vector fitting, IEEE Trans. Power Delivery, 1052-1061 (1999).
[10] I.-A. Lazar, M.-I. Andrei, E. Caciulan, G. Ciuprina, and D. Ioan, Parallel algorithms for the efficient extraction of fitting based reduced order models, Proceedings of the 7th International Symposium on ADVANCED TOPICS IN ELECTRICAL ENGINEERING, 1–13 (2011).
[11] T. Davis, Algorithm 832: Umfpack, an unsymmetricpattern multifrontal method, ACM Transactions on Mathematical Software (TOMS), 196-199 (2004).
[12] Y. Saad, Iterative Methods for Sparse Linear Systems, Society for Industrial and Applied Mathematics, 2003. Second edition with corrections.
[13] Y. Saad, and M. H. Schultz, GMRES: A generalized minimal residual algorithm for solving nonsymmetric linear systems, SIAM J. Sci. Stat. Comput., 856-869 (1986).
[14] J. Sanders and E. Kandrot, CUDA by Example: An Introduction to General-Purpose GPU Programming, Addison-Wesley Professional; 1st edition 2010.
[15] D. B. Kirk and W. mei W. Hwu, Programming Massively Parallel Processors: A Hands-on Approach (Applications of GPU Computing Series), Morgan Kaufman Elsevier, 2010.
[16] N. Bell and M. Garland, Cusp: Generic parallel algorithms for sparse matrix and graph computations, 2012. Version 0.3.0.
[17] H. H. J. M. Janssen, J. Niehof, W. H. A. Schilders, Accurate Modeling of Complete Functional RF Blocks: CHAMELEON RF, Scientific Computing in Electrical Engineering, Mathematics in Industry, 81-87 (2007).
[18] S. Kula, Interconnect Elements Propagation Quantities in PIC, Poznan University of Technology Academic Journals, Series Electrical Engineering, Iss. 70, 83-89 (2012).
