High-speed computer-generated hologram based on resource optimization for block-based parallel processing

A huge amount of computation is required to generate a hologram using a computer. In order to speed up the computer-generated-hologram (CGH) operation, we use a parallel programming technique using a general purpose graphic processing unit (GPGPU). In this paper, we propose three techniques to improve CGH performance in the condition using GPU. The first is to remove the memory bottleneck by allocating shared memory and a dedicated thread for this process, and the second is to optimize the block allocation within the GPU using a hologram pixel-based method. The third is to increase the computation time by minimizing the idle region by using multiple threads of host processor and device. When these three techniques were implemented in the GTX 1080Ti GPU, it took 25.05 ms to generate the HD digital hologram with 10 K object points, and compared to the previous research, the performance improvement was at least 1.56 times up to 216.71 times.