Out-of-core GPU 2D-shift-FFT algorithm for ultra-high-resolution hologram generation

Abstract

We propose a novel out-of-core GPU algorithm for 2D-Shift-FFT (i.e., 2D-FFT with FFT-shift) to generate ultra-high-resolution holograms. Generating an ultra-high-resolution hologram requires a large complex matrix (e.g., 100K2) with a size that typically exceeds GPU memory. To handle such a large-scale hologram plane with limited GPU memory, we employ a 1D-FFT based 2D-FFT computation method. We transpose the column data to have a continuous memory layout to improve the column-wise 1D-FFT stage performance in both the data communication and GPU computation. We also combine the FFT-shift and transposition steps to reduce and hide the workload. To maximize the GPU utilization efficiency, we exploit the concurrent execution ability of recent heterogeneous computing systems. We also further optimize our method’s performance with our cache-friendly chunk generation algorithm and pinned-memory buffer approach. We tested our method on three computing systems having different GPUs and various sizes of complex matrices. Compared to the conventional implementation based on the state-of-the-art GPU FFT library (i.e., cuFFT), our method achieved up to 3.24 and 3.06 times higher performance for a large-scale complex matrix in single- and double-precision cases, respectively. To assess the benefits offered by the proposed approach in an actual application, we applied our method to the layer-based CGH process. As a result, it reduced the time required to generate an ultra-high-resolution hologram (e.g., 100K2) up to 28% compared to the use of the conventional algorithm. These results demonstrate the efficiency and usefulness of our method.

Large‐scale 3D fast Fourier transform computation on a GPU

“…However, only part of the system can be a page-locked region, which limits the problem size that previous algorithms could treat. Lee and others [25] proposed a pinned-memory buffer approach to handle large-scale data that exceeded the maximum size of the page-locked memory. Instead of managing the entire input data in the pinned-memory space, they constructed small-sized pinned-memory buffers were constructed.…”

“…We employed the pinned buffer concept of Lee and others [25] and adapted it to 3D-FFT problem. We also present a novel data rearrangement (transposition) method for the 3D-FFT case to efficiently utilize the pinned buffer.…”

Taming data locality for task scheduling under memory constraint in runtime systems

“…Out-of-core computations on GPUs. The limited memory of GPUs have motivated many studies on how to efficiently access data stored outside the GPU memory for several specific applications such as stencil computations [3], sorting problems [4], large scale graph processing [5] or graphic computations [6]. The solution generally consists in building data blocks that each fit within the GPU memory.…”

Efficient mesh‐based realistic computer‐generated hologram synthesis with polygon resolution adjustment

“…Because high-resolution CGH synthesis has a large amount of computation, it is essential to speed it up, and many studies have been conducted. Lee [13] accelerated FFT by processing FFT-shift and FFT simultaneously using FFT-shift performed forward and backward in FFT. Matsushima [12] proposed an algorithm to speed up occlusion culling by limiting the area calculated by diffracted angles in mesh-based CGH.…”