Skip to Main content Skip to Navigation

Modélisation de phénomène électromagnétiques hyperfréquences sur calculateurs parallèles

Abstract : The purpose of this work is the efficient implementation on parallel computers of a formulation for the modeling of unbounded frequency domain microwave problems. Only parallel computation actually enables to mode1 real devices because it reduces the computation time and mainly arranges enough memory. Maxwell's equations are discretized in three dimensions (3D) using finite elements method coupled with absorbing boundary conditions. The formulation has been implemented on parallel computers with shared or distributed memory. The sequential package has been first implemented on a CRAY Cg8 (shared memory). The parallelism is introduced by adding manually compiler directives and changing the data storage mode increases vector performances. The results show that the parallelization of sequential algorithms can give good performances on this type of architecture. The use of parallel-distributed memory computers (cluster of workstations, CRAY T3E) requires a deep modification of the algorithms in order to obtain good performances. For example, the finite element matrix is distributed on al1 processors and the number of messages passing is also minimized. Some examples of answers of realistic devices enlightened with a plane wave are shown. Application to open waveguides is also presented. The computed far field is in good agreement with those obtained by measures.
Complete list of metadata
Contributor : Publications Ampère Connect in order to contact the contributor
Submitted on : Tuesday, April 3, 2007 - 11:07:30 PM
Last modification on : Wednesday, March 31, 2021 - 3:10:05 AM
Long-term archiving on: : Wednesday, April 7, 2010 - 1:01:32 AM


  • HAL Id : tel-00139905, version 1


Christian Vollaire. Modélisation de phénomène électromagnétiques hyperfréquences sur calculateurs parallèles. Autre. Ecole Centrale de Lyon, 1997. Français. ⟨tel-00139905⟩



Record views


Files downloads