PNNL

Abstract

Smoothed Particle Hydrodynamics (SPH) is essential for modeling complex large-deformation problems across various applications, requiring significant computational power. A major portion of SPH computation time is dedicated to the Nearest Neighboring Particle Search (NNPS) process. While advanced NNPS algorithms have been developed to enhance SPH efficiency, the potential efficiency gains from modern computation hardware remain underexplored. This study investigates the impact of GPU parallel architecture, low-precision computing on GPUs, and GPU memory utilization on NNPS efficiency. Our approach employs a GPU-accelerated mixed-precision SPH framework, utilizing low-precision float-point 16 (FP16) for NNPS while maintaining high precision for other components. To ensure FP16 accuracy in NNPS, we introduce a Relative Coordinated-based Link List (RCLL) algorithm, storing FP16 relative coordinates of particles within background cells. Our findings emphasize the substantial superiority of GPUs over CPUs, with efficiency gains of up to 1000x. The mixed-precision SPH, based on the computationally intensive O(N^2) all-list NNPS algorithm, achieves a 5x efficiency boost through FP16 computation. In contrast, the mixed-precision SPH, based on the computationally lighter O(N) RCLL algorithm, yields a 1.5x efficiency gain over standard FP64, primarily limited by memory usage. With optimized GPU memory utilization strategies, the FP16 RCLL algorithm can further enhance efficiency by 2.7x in an example involving 1 million particles. The proposed mixed-precision SPH framework based on the RCLL algorithm excels in both accuracy and efficiency, particularly for high-resolution problems with a large number of particles.