We develop a method that maximizes throughput through the use of vector-based memory I/O and optimized 2D FFT libraries that run on all available physical cores. We also show how to decompose ...