HI-FFT: Heterogeneous Parallel In-Place Algorithm for Large-Scale 2D-FFT

Fast Fourier Transform (FFT) is a fundamental operation for 2D data in various applications. To accelerate large-scale 2D-FFT computation, we propose Heterogeneous parallel In-place algorithm, HI-FFT. Our novel work decomposition method makes it possible to run our algorithm on the original (i.e., in-place), unlike prior algorithms that require additional memory space out-of-place) guarantee independence among sub-tasks. also removes duplicated operations out-of-place approaches. Using method, introduced an in-place heterogeneous utilizes both multi-core CPU and GPU simultaneously. maximize utilization efficiency of computing resources, priority-based dynamic scheduling method.We compared performance seven different algorithms, including ours, problems whose sizes varied from 20K2 120K2. As result, found achieved up 2.92 4.42 times higher than conventional homogeneous based state-of-the-art libraries, respectively. Also, showed 2.27 while requiring two less space. check benefit HI-FFT actual application, applied CGH (Computer Generated Holography) process. We successfully reduces hologram generation time. These results demonstrate advantage approach computation.