🤖 AI Summary
This study addresses the efficient implementation of the three-dimensional finite-difference time-domain (3D FDTD) algorithm in multi-GPU environments, systematically evaluating the impact of various communication strategies on the overhead introduced by convolutional perfectly matched layers (CPML). Through comparative analysis of host-mediated versus GPU-direct communication and ghost-cell extension techniques, the work demonstrates that GPU-direct communication is the dominant optimization, whereas extending ghost regions yields marginal gains. Experimental results show single-GPU performance of 28.89–32.9 billion output points per second with CPML overhead below 1%. GPU-direct communication achieves a 2.46–2.76× speedup over host-mediated transfer. Strong scaling yields a 1.51× acceleration with two GPUs, and four GPUs successfully overcome single-card memory limitations to simulate extremely large grids.
📝 Abstract
In this paper we describe a communication-strategy study for multi-GPU three-dimensional finite-difference time-domain computation with convolutional perfectly matched layer boundary conditions using CUDA. The metrics used to determine the most effective implementation include runtime, throughput in millions of output points per second, strong-scaling efficiency, CPML overhead, host-staged versus direct GPU-to-GPU exchange speedup, and enlarged-ghost speedup. On a single NVIDIA Quadro RTX 6000 GPU, the CPML implementation sustains 2,889--3,290 million output points per second with less than 1\% boundary-layer overhead, providing the single-GPU baseline for the multi-GPU study. The results show that direct GPU-to-GPU peer exchange is the dominant optimization with a 2.46--2.76$\times$ speedup over host-staged exchange, while enlarged ghost regions give only modest benefits because the reduced communication frequency is partly offset by redundant computation and additional memory traffic. On NVIDIA Quadro RTX 8000 GPUs, the implementation gives up to a 1.51$\times$ speedup on two GPUs for the tested strong-scaling cases, while four GPUs enable larger grids that approach or exceed single-GPU memory capacity.