Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.
To the best of my knowledge, FDTD is much more sensitive to memory speed than number crunching power of CPU. The calculations to be done for each cell is substantially less than what is for FE. I would not use parallel processing (multicomputer) for calulations of each cell. Rather I would utilize some type of domain decomposition for that, each being done in a separate computer and the final processing would be done in a single computer.
For each cell, you had better use multiple threads. You can find good papers in IEEE on both topics.
Thank you for your reply. In fact, I do not want to use parallel processing (multicomputer) for calulations of each cell as you said. I just plan to separate the large problem into several small parts, and solve these small parts by the conventional FDTD. Each part is solved by one comupter,respectively.
And I only want to know how to separate the large body, communicate every computer and share the data.