A scheduling policy to save 10% of communication time in parallel fast Fourier transform

The fast Fourier transform (FFT) has applications in almost every frequency related study, for example, image and signal processing, radio astronomy. It is also used as a Poisson operator inversion kernel partial differential equations fluid flows, density functional theory, many-body others. three-dimensional FFT large time complexity . Hence, parallelization to compute such FFTs. Popular libraries perform slab division or pencil decomposition of data. None the existing achieve perfect inverse scaling with cores because requires all-to-all communication clusters hitherto do not have physical connections. Dragonfly, one popular topologies interconnect, supports hierarchical connections among components. We show that if we align Dragonfly topology will better reduce time.