Hydrodynamic Simulations using GPGPU Architectures

TL;DR

This paper explores GPU-based acceleration of hydrodynamic simulations using Smoothed Particle Hydrodynamics (SPH), proposing optimizations and analyzing the balance between performance and accuracy in fluid dynamics modeling.

Contribution

It introduces specific GPU optimization techniques for SPH simulations and evaluates their impact on computational efficiency and accuracy.

Findings

GPU acceleration significantly improves simulation speed.

Optimizations balance performance with numerical accuracy.

Framework and hardware choices affect simulation outcomes.

Abstract

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.