ACES Publication Search





There are 13 search results for:



Title: APPLICATION OF INTEGRAL EQUATION AND HYBRID TECHNIQUES TO THE PARALLEL COMPUTATION OF ELECTROMAGNETIC FIELDS IN A DISTRIBUTED MEMORY ENVIRONMENT
Abstract: This paper describes the parallelization of the method of moments and hybrid code FEKO for execution on massively parallel supercomputers with a distributed memory as well as on clusters of connected workstations. The parallel implementation of the different phases of the solution process, e.g. matrix fill, solution of the system of linear equations, and near- and far-field computation is discussed in detail. Several results for different applications are given and the achieved performance is presented
Author(s): Ulrich Jakobus, Ulrich Jakobus
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 1737 KB

Title: RUNNING SUPERNEC ON THE 22 PROCESSOR IBM-SP2 AT SOUTHAMPTON UNIVERSITY
Abstract: SuperNEC (SNEC) is an object-oriented version of NEC-2 which has been modified to execute on a network of distributed memory processors. The matrix filling, solving and pattern computation routines are capable of running in parallel. A number of structures have been simulated using this code on the 22 processor IBM-SP2 machine at Southampton University. The principal problem studied was the DC-3 at 90 MHz. LU decomposition and an iterative matrix solution scheme were used in the study. The simulation time for this structure (which includes 3-D radiation patterns) dropped from 2.5 hours on a single processor to about 17 minutes when simulated on 12 processors using LU decomposition. Execution times are about half of these times when using the iterative solver. The far field patterns obtained from the simulation are compared with measured data and show good agreement. The largest problem tackled on the IBM machine was the DC-3 simulated at 160 MHz. This problem requires 17035 segments and was simulated in 5.3 hours on 21 processors
Author(s): D. C. Nitch, A. P.C. Fourie, J. S. Reeve, D. C. Nitch, A. P.C. Fourie, J. S. Reeve
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 458 KB

Title: HIGHLY PARALLEL IMPLEMENTATION OF THE 3D INTEGRAL EQUATION ASYMPTOTIC PHASE METHOD FOR ELECTROMAGNETIC SCATTERING
Abstract: In this paper, we discuss the implementation of the 3D Integral Equation-Asymptotic Phase (IE-AP) method using the parallel architecture IBM RS/6000 SP. The IE-AP method is a hybrid numerical/asymptotic approach for electromagnetic scattering that attempts to reduce the number of unknowns required to accurately model electrically large structures. The IE-AP method will be described, and results will be reported for the parallel matrix fill implementation, and the relative performance of the PESSL and PETSc toolkits for parallel matrix solution.
Author(s): Xianneng Shen, Aaron W. Davis, Keith R. Aberegg, Andrew F. Peterson, Xianneng Shen, Aaron W. Davis, Keith R. Aberegg, Andrew F. Peterson
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 651 KB

Title: PERFORMANCE OPTIMIZATION OF AN INTEGRAL EQUATION CODE FOR JET ENGINE SCATTERING ON CRAY-C90
Abstract: The numerical solution of Maxwell’s equations is a computationally intensive task and use of high-performance parallel computing facilities is necessary for the larger class of practical problems in scattering, propagation and antenna modeling. It is therefore necessary to carefully consider algorithm optimizations aimed at improving the code’s run time performance on the computing platform employed. Although some performance improvement can be derived from compiler-level optimizations, further speed-up may involve manual effort in algorithm restructuring, data layout, and parallelization. This paper focuses on the manual optimizations used to improve the performance of a moment method code for the analysis of a cylindrically periodic structure, as is the case with a jet engine. We describe the steps taken which resulted in nearly two orders of magnitude improvement over the original version of the code. A 16-processor shared-memory CRAY-C90 vector supercomputer was employed. Our optimization took advantage of SSD its solid-state storage, enabled better loop vectorizations, parallelized the matrix_fill routine, and called appropriate CRAY-C90 library routines
Author(s): Mikhail Smelyanskiy, Edward S. Davidson, John L. Volakis, Mikhail Smelyanskiy, Edward S. Davidson, John L. Volakis
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 868 KB

Title: OPTIMISATION AND LARGE SCALE COMPUTATION IN INTEGRAL EQUATION SCATTERING ANALYSES
Abstract: The kinds of difficulties posed by large scattering computations change as larger problems are addressed. Unfavourable cost scalings make the performance of small, core, portions of code dominant, and require that they, and the overall code structure, be optimised for large scale computation. This is discussed in the context of rcs and scattering computations of multi-wavelength bodies using a time domain integral equation treatment. Examples presented include the NASA almond evaluated at 25 wavelengths long, and an assembly of 101 spherical scatterers of ~1/2 wavelength diameter each, occupying a volume of side ~250 wavelengths.
Author(s): S. J. Dodson, S. P. Walker, M. J. Bluck, S. J. Dodson, S. P. Walker, M. J. Bluck
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 2480 KB

Title: PERFORMANCE MODELING OF THE FINITE-DIFFERENCE TIME-DOMAIN METHOD ON PARALLEL SYSTEMS
Abstract: As high-performance parallel codes are developed or ported to new architectures, it is often diffi-cult to quantify the causes of performance problems. Models of program performance can provide users with insight into the effect of system and program parame-ters on performance, can help programmers tune appli-cations, and can help programmers make decisions about processor allocation. This paper introduces a modeling technique applied to the Finite-Difference Time-Domain (FDTD) algorithm. The technique models the performance of an existing application in terms of the size of the problem being solved and the number of processors. The models show that for sufficiently large problem sizes the algorithm performs well. However, for smaller prob-lem sizes or when too many processors are used, the mod-els show that parallel overheads become significant.
Author(s): James E. Lumpp, Jr, Shashi K. Mazumdar, Stephen D. Gedney, James E. Lumpp, Jr, Shashi K. Mazumdar, Stephen D. Gedney
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 1077 KB

Title: IMPLEMENTATION AND APPLICATION OF A FD-TD SIMULATION TOOL FOR THE ANALYSIS OF COMPLEX 3D STRUCTURES
Abstract: This paper presents information about the develop-ment of an electromagnetic analysis tool “LC” which integrates the Finite-Difference Time-Domain (FD-TD) method with an interactive Graphical User In-terface (GUI). The paper will discuss the program implementation and design; many issues in the implementation have surfaced, concerning the problem of producing a graphical model editor/simulator that runs efficiently on various hardware systems. The paper will also explore how the solver’s capabilities aid design engineers when investigating and solving packaging and interconnect design issues as well as the program’s application to engineering problems
Author(s): Gary Haussmann, Melinda Piket-May, Roger J. Gravrok, Gary Haussmann, Melinda Piket-May, Roger J. Gravrok
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 662 KB

Title: Optimizing the Parallel Implementation of a Finite Difference Time Domain Code on a Multi-user Network of Workstations
Abstract: The implementation of a parallel, three dimensional, finite difference time domain (FDTD) computer program is considered and applied to a test scattering problem on a multi-user network of desktop workstations. The computation has primarily been done on a local area network (LAN) using six identical HP 9000/715 workstations (i.e. a homogeneous environment) with the Parallel Virtual Machine (PVM) software being employed as the communications harness.In this paper the sequential and parallel FDTD approaches are reviewed. We investigate the factors which cause a reduction in efficiency in the latter, such as host allocation and load balancing. We propose a task migration process, which is efficient for the FDTD algorithm, as a partial solution. The advantages of this approach are discussed and further developments based on available computational resources are suggested
Author(s): J. V. Mullan, C. J. Gillan, V. F. Fusco, J. V. Mullan, C. J. Gillan, V. F. Fusco
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 710 KB

Title: PARALLEL COMPUTATION OF LARGE-SCALE ELECTROMAGNETIC FIELD DISTRIBUTIONS
Abstract: Some experience of the use of high-frequency electromagnetics software on parallel computers is reported. Types of such computers are reviewed and approaches to the parallelisation of existing serial software are discussed. A practical large-scale problem is presented involving the modelling in very fine detail of electromagnetic penetration into biological systems. This was tested on state-of-the-art parallel computers and important practical and strategic aspects of the experience derived are discussed. It was found that considerable programmer effort was required to optimise the software to use the computer architecture effectively, but that efficient acceleration of the run-times of typical computational tasks could be achieved, provided that the tasks were large and were partitioned optimally.
Author(s): Peter S. Excell, Adam D. Tinniswood, Kathleen Haigh-Hutchinson, Peter S. Excell, Adam D. Tinniswood, Kathleen Haigh-Hutchinson
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 2818 KB

Title: A DEDICATED TLM ARRAY PROCESSOR
Abstract: for TLM. Limitations introduced by the The transmission line matrix (TLM) method is introduced and the specific issue of computational efficiency is discussed. The implementation of TLM on parallel computers is studied leading to the creation of a highly efficient processor designed specifically connection strategies employed by most parallel architectures are overcome through the use of a novel data routing architecture. The basic idea is extended to include stub loaded and three-dimensional TLM. The development of a prototype processor is discussed and potential applications are given
Author(s): D. Stothard, S. C. Pomeroy, D. Stothard, S. C. Pomeroy
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 827 KB

Title: BANDWIDTH REDUCED FULL-WAVE SIMULATION OF LOSSLESS AND THIN PLANAR MICROSTRIP CIRCUITS
Abstract: We present a full-wave, high-performance, numerical scheme for the analysis of planar microstrip circuits which is based on an efficient electromagnetic formulation of the field problem and on the bandwidth reduction of the discretized sparse matrix. The above mentioned electromagnetic efficiency is at-tained by considering a Mixed Potential Integral Equa-tion (MPIE) with the kernel expressed by closed-form spatial-domain Green’s functions; as a consequence, the reaction integrals are evaluated by using just one-dimensional numerical integration over a finite spatial do-main. Moment method discretization of the MPIE leads to the corresponding matrix problem. The accurate analysis of the matrix properties shows that a sparsity of 70-85% in the discretized linear system can be routinely enforced without significantly altering the solution accuracy. A new scheme for the sparse matrix bandwidth reduc-tion, particularly tailored for electromagnetic problems, can be therefore introduced, leading to considerable re-ductions of the simulation time. Results are presented demonstrating that the use of a bandwidth reduction strategy coupled with efficient problem-matched Green’s functions allows as to obtain speed-ups in simulation time of more than one order of magnitude with respect to stan-dard state of the art implementations.
Author(s): A. Caproni, F. Cervelli, M. Mongiardo, L. Tarricone, F. Malucelli, A. Caproni, F. Cervelli, M. Mongiardo, L. Tarricone, F. Malucelli
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 1033 KB

Title: SPEEDUP USING A MODAL FREQUENCY METHOD FOR FINITE ELEMENT ANALYSIS OF A DUAL-MODE MICROWAVE FILTER
Abstract: Computer time required for finite element analysis of microwave filters is reduced by more than an order of magnitude by using modal frequency rather than direct frequency methods. In the conventional direct fre-quency method, the number of unknowns is equal to the number of edge degrees of freedom. Instead, the new modal frequency method first computes the 3D modes and then uses them as basis functions, thereby greatly reduc-ing the number of degrees of freedom. The two methods are applied to the European benchmark problem of a dual-mode microwave filter. The modal frequency meth-od obtains essentially the same results as the direct fre-quency method, but when analyzing 201 frequencies it yields a speedup factor of 15.
Author(s): John R. Brauer, John R. Brauer
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 927 KB

Title: A FAST MEI SCHEME FOR THE COMPUTATION OF SCATTERING BY VERY LARGE CYLINDERS
Abstract: A fast scheme for measured equation of invariance (MEI) method is presented in this paper. The scheme combines a strategic technique of the interpolation and extrapolation of MEI coefficients with a special algorithm of cyclic block band matrix to fast solve the scattering problems of very large conducting cylinders. The circumferential dimension of scattering objects could exceed 10,000 wavelength. Computational speed could be 2-3 order faster than conventional MEI method. The fast scheme is especially applicable to scattering problems of very large conducting objects in which other numerical methods may fail.
Author(s): Y. Liu, K. Lan, K. K. Mei, E. K. N. Yung, Y. Liu, K. Lan, K. K. Mei, E. K. N. Yung
File Type: Journal Paper
Issue:Volume: 13      Number: 2      Year: 1998
Download Link:Click here to download PDF     File Size: 578 KB