Non-Uniform FFT (NUFFT) Algorithms for Antenna Radiation

We present and discuss the parallel implementation on GPUs of three types of NUFFTs in CUDA language: NER (Non-Equispaced Results), or type-1, NUFFTs dealing with equally spaced samples and returning non-equally spaced results; NED (Non-Equispaced Data), or type-2, NUFFTs dealing with non-equally spaced samples and returning equally spaced results; Type-3 NUFFTs, dealing with non-equally spaced samples and returning non-equally spaced results.The computational and accuracy performance of the three implementations are analyzed for the three cases with reference to array antenna radiation problems. In particular, the timing performance is assessed against that of a parallel implementation of the same algorithm on multicore CPUs.