Short Course: Phi Coprocessor and Acceleration Techniques for Computational Electromagnetic Methods

Presented by

Wenhua Yu and Xiaoling Yang - 2COMU, Inc.

Abstract

In this short course, we introduce the architecture of Phi Coprocessor that is a new and powerful numerical acceleration platform, programming techniques and acceleration techniques in computational electromagnetic methods. The examples will be in acceleration for the parallel FDTD methods and matrix operations in MoM and FEM. This short course introduces how to modify your serial code to run on Phi Coprocessor in the parallel format. We also introduce how to use Phi Coprocessor to accelerate an existing homegrown or commercial EM software. Phi Coprocessor can be used as a regular CPU and runs the EM code compiled on Intel Xeon E3 and E5 with slight code modifications. Phi Coprocessor IS NOT a GPU and IS a general hardware EM acceleration simulation platform.