arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2401.08586cs.DC
收藏

采用基于网格的相对坐标的 GPU 加速混合精度光滑粒子流体动力学框架

A GPU accelerated mixed-precision Smoothed Particle Hydrodynamics framework with cell-based relative coordinates

Zirui Mao, Xinyi Li, Shenyang Hu, Ganesh Gopalakrishnan, Ang Li

AI总结:

本文提出 GPU 加速的混合精度 SPH 框架,并用 FP16 RCLL 算法与内存优化加速 NNPS,在百万粒子示例中实现显著性能提升。

摘要:

光滑粒子流体动力学(SPH)对于在各类应用中模拟复杂的大变形问题至关重要,且需要大量计算能力。SPH 计算时间的很大一部分用于最近邻粒子搜索(NNPS)过程。尽管先进的 NNPS 算法已被开发以提高 SPH 效率,但现代计算硬件所能带来的潜在效率提升仍未得到充分探索。本研究考察了 GPU 并行架构、GPU 上的低精度计算以及 GPU 内存管理对 NNPS 效率的影响。我们的方法采用 GPU 加速的混合精度 SPH 框架,对 NNPS 使用低精度 16 位浮点数(FP16),同时保持其他组件的高精度。为确保 NNPS 中 FP16 的准确性,我们提出了一种基于相对坐标的链表(RCLL)算法,用于存储背景网格内粒子的 FP16 相对坐标。测试结果显示,相对于基于 CPU 的 NNPS 算法,出现了三轮显著加速。第一轮来自 GPU 并行计算,效率提升最高达 1000 倍。第二轮通过低精度 GPU 计算实现,其中所提出的基于 FP16 的 RCLL 算法在 GPU 上相比基于 FP64 的方法带来 1.5 倍的效率提升。通过优化 GPU 内存带宽利用,FP16 RCLL 算法的效率可进一步提高 2.7 倍,这一点在一个包含 100 万粒子的示例中得到证明。我们的代码发布于 https://github.com/pnnl/lpNNPS4SPH。

英文摘要:

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 management 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 testing results show three significant speedup rounds for CPU-based NNPS algorithms. The first comes from parallel GPU computations, with up to a 1000x efficiency gain. The second is achieved through low-precision GPU computing, where the proposed FP16-based RCLL algorithm offers a 1.5x efficiency improvement over the FP64-based approach on GPUs. By optimizing GPU memory bandwidth utilization, the efficiency of the FP16 RCLL algorithm can be further boosted by 2.7x, as demonstrated in an example with 1 million particles. Our code is released at https://github.com/pnnl/lpNNPS4SPH.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑