Novel hardware acceleration techniques for finite difference time domain methods

In this paper, we introduce the architecture of Phi Coprocessor, programming techniques and acceleration techniques in the finite different time domain (FDTD) methods. Phi Coprocessor can be used as a regular CPU and run the EM code optimized for regular CPUs such as Intel Xeon E5 or AMD Opteron 6300 with slight code modifications. The examples will be for the acceleration of the parallel FDTD methods.