Different implementations on a massively parallel computer system of a semi-Lagrangian method within the numerical weather forecast model HIRLAM are presented. In principle semi-Lagrangian methods on massively paralle...
详细信息
Different implementations on a massively parallel computer system of a semi-Lagrangian method within the numerical weather forecast model HIRLAM are presented. In principle semi-Lagrangian methods on massively parallel architectures result in irregular communications, i.e., communications between arbitrary processors. It is shown that the fastest implementation increases the total execution time per time step with an acceptable amount in relation to the advantage of applying a semi-Lagrangian method.< >
作者：
FUCHS, CKOPP, GSCHWAB, AJInstitute of Electric Energy Systems and High-Voltage Technology University of Karlsruhe
Kaiserstrasse 12 76128 Karlsruhe Germany Chistophs Fuchs was born in Sinsheim
germany in 1967. He received the Dipl.-Ing degree in electrical engineering at the University of Karlsruhe in 1992. In April 1993 he joined the Institute of Electric Energy Systems and High-Voltage Technolgoy. His current research interests include analytical and numerical calculation of electromagnetic fields in EMC applications and high-performance computing. Gerhard Kopp was born in Karlsruhe
Germany in 1968. He is studying physics at the University of Karlsruhe and is now working towards his diploma. His thesis deals with numeical calculations of electromagnetic fields. Professor Dr.-Ing. A. J. Schwab graduated in electrical engineering at the Universiyt of Karlsruhe. After receiving his Ph.D. degree he worked as a post-doctoral fellow at MIT in 1970/71. He held appointments at the Universities of Darmstadt and dormund and
eventually was appointed EE Full Professor at the University of Karlsruhe. In 1989 he was appointed Director of the ABB Corporate Research Center at Heidelberg Germany. Since 1994 he has worked again as Professor and Director of the Institute for Electric Energy Systems and High-Voltage Technology at the University of Karlsruhe. He is the author of three books on high-voltage measurements electromagnetic compatibility and field theory concepts that have been published in German English Russian an Chinese. He is a member of the German national society of electrical engineers VDE a member of the Conférence Internationale des Grands Réseaux Electriques (CIGRE) Fellow of the IEEE and at present Chairman of the IEEE Germany Section.
This paper presents a new method in TLM for very efficient computation of highly conducting thin shielding walls. Firstly, the impulse response of a finite conducting sheet is computed analytically via inverse Laplace...
详细信息
This paper presents a new method in TLM for very efficient computation of highly conducting thin shielding walls. Firstly, the impulse response of a finite conducting sheet is computed analytically via inverse Laplace transform. Subsequently, a recursive convolution integral is formulated with the analytical impulse response. This allows a very efficient and accurate calculation of the diffusion process through imperfectly conducting walls.
To construct numerical schemes of he Godunov type for solving magnetohydrodynamical (MHD) problems, an approximate method of solving the MHD Riemann problem is required in order to calculate the time-averaged fluxes a...
详细信息
To construct numerical schemes of he Godunov type for solving magnetohydrodynamical (MHD) problems, an approximate method of solving the MHD Riemann problem is required in order to calculate the time-averaged fluxes at the interfaces of numerical zones, Such an MHD Riemann solver is presented here which treats all waves emanating from the initial discontinuity as themselves discontinuous. Thus shock jump conditions are used for rarefactions, which limits the applicability of this work to weak rarefactions, the case most important for computation. The solutions from our approximate MHD Riamnn solver consist of two fast waves (either shock or rarefaction) two rotational discontinuities, two rarefaction waves (either shock or rarefaction), and one contact discontinuity for a general MHD Riemann problem. In order to display rotational discontinuities, a three-component model is necessary. Only under very limited circumstances is there no rotational discontinuity involved and thus the two component approximation may be used in the MHD Riemann problem. The solutions of the MHD Riemann problem in the shock tube problem which generates the compound wave in the earlier work contain two fast rarefaction waves, two slow shocks, one contact discontinuity, and one rotational discontinuity in our formalism. (C) 1994 Academic Press, Inc.
作者：
DAI, WLWOODWARD, PRSchool of Physics and Astronomy
Supercomputer Institute Army High Performance Computing Research Center University of Minnesota 1100 Washington Avenue South Minneapolis Minnesota 55415
An extension of the piecewise parabolic method to treat multi-dimensional ideal magnetohydrodynamical equations is presented in this paper. The multidimensional scheme is constructed from a one-dimensional functioning...
详细信息
An extension of the piecewise parabolic method to treat multi-dimensional ideal magnetohydrodynamical equations is presented in this paper. The multidimensional scheme is constructed from a one-dimensional functioning code based on the dimensional splitting method originally suggested by Strang. The functioning code is built upon a nonlinear Riemann solver for ideal MHD equations recently developed by the authors. The correctness of the scheme is tested in the steepening of waves in both one- and two-dimensional situations and in Various MHD shock-tube problems which involve all the discontinuities in ideal MHD. The robust character of the scheme is demonstrated in the shock-tube problems and in the interaction between MHD shocks and a cloud. The results of these problems show that the scheme keeps the principal advantages of a high-order Godunov scheme: robust operation in the presence of very strong waves, thin shock fronts with little attendant noise generation, and thin contact discontinuity, (C) 1994 Academic Press. Inc.
作者：
DAI, WLWOODWARD, PRUniversity of Minnesota
School of Physics and Astronomy Army High Performance Computing Research Center Supercomputer Institute 1100 Washington Avenue South Minneapolis Minnesota 55415
Interactions between magnetohydrodynamical (MHD) discontinuities are studied through numerical simulations for the set of one-dimensional MHD equations. The interactions include the impact of a shock on a contact disc...
详细信息
Interactions between magnetohydrodynamical (MHD) discontinuities are studied through numerical simulations for the set of one-dimensional MHD equations. The interactions include the impact of a shock on a contact discontinuity, the collision of two shocks, and the catchup of a shock over another shock. The shocks involved in the interactions may be very strong. Each shock in an interaction may be either a fast or a slow shock.
作者：
KIMSEY, KDOLSON, MA[a] U.S. Army Research Laboratory
Weapons Technology Directorate Aberdeen Proving Ground MD 21005-5066 USA
[b] Army High Performance Computing Research Center Computer Sciences Corporation USA
This paper discusses a parallel algorithm and data structures for implementing multimaterial, two-step Eulerian finite difference solution schemes on hypercube architectures. Selected problems in impact dynamics have ...
详细信息
This paper discusses a parallel algorithm and data structures for implementing multimaterial, two-step Eulerian finite difference solution schemes on hypercube architectures. Selected problems in impact dynamics have been modeled on the Connection Machine model CM5, and the results are compared with computational results reported in the literature, as well as direct comparison with experimental data.
This paper details the electrical design of a multichip module ( MCM ) which represents a single node of a multiprocessor digital system. After a brief overview on the architecture, the paper discusses the key perform...
详细信息
This paper details the electrical design of a multichip module ( MCM ) which represents a single node of a multiprocessor digital system. After a brief overview on the architecture, the paper discusses the key performance related ingredients of the MCM such as optimum chip placement, waveform degradation, delay and noise budgets. It is the conclusion of the paper that the physical and electrical design of an MCM be conducted in parallel so as to ensure the electrical performance of the module with minimum redesign.
We outline a unified approach for building a library of collective communication operations that performs well on a cross-section of problems encountered in real applications. The target architecture is a two-dimensio...
详细信息
We outline a unified approach for building a library of collective communication operations that performs well on a cross-section of problems encountered in real applications. The target architecture is a two-dimensional mesh with worm-hole routing, but the techniques also apply to higher dimensional meshes and hypercubes. We stress a general approach, addressing the need for implementations that perform well for various sized vectors and grid dimensions, including non-power-of-two grids. This requires the development of general techniques for building hybrid algorithms. Finally, the approach also supports collective communication within a group of nodes, which is required by many scalable algorithms. Results from the Intel Paragon system are included.< >
暂无评论