Advantages of modal decomposition for efficient scaling of parallel FDTD computations
| Autor: | Dmitry A. Gorodetsky, Philip A. Wilsey |
|---|---|
| Rok vydání: | 2011 |
| Předmět: |
Speedup
Computer science Modal analysis Computation Finite-difference time-domain method Computer Science Applications Computational science Matrix decomposition Parallel processing (DSP implementation) Modeling and Simulation Electronic engineering Overhead (computing) Transient (computer programming) Electrical and Electronic Engineering |
| Zdroj: | International Journal of Numerical Modelling: Electronic Networks, Devices and Fields. 24:507-514 |
| ISSN: | 0894-3370 |
| DOI: | 10.1002/jnm.795 |
| Popis: | When a scheme such as FDTD is used to perform full-wave electromagnetic simulation of electrically large objects, it can result in excessive computation times. Parallel processing is one means to reduce the computation time. A shortcoming of the parallel approach is the communication overhead that is responsible for reduced speedup efficiency but inevitable at every time step. An alternative to reduce the communication is to look at FDTD from the modal analysis point of view. With modal analysis we can gain some very useful insights into the operation of the algorithm, which we pay for with a relatively expensive setup time. We consider parallelizing the modal analysis algorithm with transient and steady-state excitation scenarios. Copyright © 2011 John Wiley & Sons, Ltd. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|