A vector-concurrent application of a conjugate gradient FFT algorithm to electromagnetic radiation and scattering problems

A vector-concurrent form of a nonlinear iterative technique, namely the conjugate gradient fast Fourier transform (CGFFT), is considered. The vectorizable nature of the CGFFT algorithm is exploited by identifying the major processes involved in a given iteration, including the fast Fourier transforms, vector dot products, and norm calculations. Since many operations within the iteration loop are not interdependent, substantial vectorization can be achieved, resulting in improved computational speed. The optimized CGFFT algorithm is used for scattering computations associated with large planar structures. It is shown that the optimization results in considerable improvement in the algorithm´s speed, especially for larger size geometries