基于GPGPU体系结构的流体动力学仿真
Hydrodynamic Simulations using GPGPU Architectures
摘要:
模拟不同流体的流动可能是一个高度计算密集型的过程,需要大量的资源。最近有很多研究工作针对GPU处理,这可以大大提高不同应用程序的性能,例如平滑粒子流体动力学(SPH),这是最常用的流体动力学模拟。光滑粒子流体动力学(SPH)是计算流体动力学(CFD)中常用的一种数值方法。它是一种可以模拟粒子流以及与结构和高度变形物体相互作用的方法。它将流体替换为一组粒子,这些粒子具有质量、速度和位置等属性,并根据控制动力学进行移动。流体的动力学是基于Navier-Stokes方程。这些描述了流体中连续场的物理性质。SPH近似这些方程使用积分插值,然后数值求解。本文讨论了可用于加速算法的现有技术的现状,并提出了一组可以通过使用不同框架实现的优化。我们还得出结论,性能和精度之间的平衡,使用不同的数值算法,框架和硬件优化。
英文摘要:
Simulating the flow of different fluids can be a highly computational intensive process, which requires large amounts of resources. Recently there has been a lot of research effort directed towards GPU processing, which can greatly increase the performance of different applications, such as Smoothed Particle Hydrodynamics (SPH), which is most commonly used for hydrodynamic simulations. Smoothed particle hydrodynamics (SPH) is a numerical method commonly used in Computational Fluid Dynamics (CFD). It is a method that can simulate particle flow and interaction with structures and highly deformable bodies. It replaces the fluid with a set of particles that carry properties such as mass, speed and position that move according to the governing dynamics. The dynamics of fluids are based on the Navier-Stokes equations. These describe the physical properties of continuous fields in the fluid. SPH approximates these equations using an integral interpolant that is then solved numerically. This article addresses the current state of technologies available that can be used to speed up the algorithm and proposes a set of optimizations that can be achieved by using different frameworks. We also draw conclusions regarding the equilibrium between performance and accuracy, using different numerical algorithms, frameworks and hardware optimizations.
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。