Header

UZH-Logo

Maintenance Infos

Adaptive Sampling and Rendering of Fluids on the GPU


Zhang, Y; Solenthaler, B; Pajarola, R (2008). Adaptive Sampling and Rendering of Fluids on the GPU. In: Proceedings Symposium on Point-Based Graphics, Los Angeles, 10 August 2008 - 11 August 2008, 137-146.

Abstract

In this paper, we propose a novel GPU-friendly algorithm for the Smoothed Particle Hydrodynamics (SPH) simulation for weakly compressible fluids. The major goal of our algorithm is to implement a GPU-based SPH simulation that can simulate and render a large number of particles at interactive speed. Additionally, our algorithm exhibits the following three features. Firstly, our algorithm supports adaptive sampling of the fluids. Particles can be split into several sub-particles in geometrically complex regions to provide a more accurate simulation. At the same time, nearby particles deep inside the fluids are merged to a single particle to reduce the number of particles. Secondly, the fluids are visualized by directly computing the intersection between ray and an isosurface defined by the surface particles. A dynamic particle grouping algorithm and equation solver are employed to quickly find the ray-isosurface intersection. Thirdly, based on the observation that the SPH simulation is a naturally parallel algorithm, the whole SPH simulation, including the adaptive sampling of the fluids as well as surface particle rendering, is executed on the GPU to fully utilize the computational power and parallelism of modern graphics hardware. Our experimental data shows that we can simulate about 50K adaptively sampled particles, or up to 120K particles in the fixed sampling case at a rate of approximately 20 time steps per second.

Abstract

In this paper, we propose a novel GPU-friendly algorithm for the Smoothed Particle Hydrodynamics (SPH) simulation for weakly compressible fluids. The major goal of our algorithm is to implement a GPU-based SPH simulation that can simulate and render a large number of particles at interactive speed. Additionally, our algorithm exhibits the following three features. Firstly, our algorithm supports adaptive sampling of the fluids. Particles can be split into several sub-particles in geometrically complex regions to provide a more accurate simulation. At the same time, nearby particles deep inside the fluids are merged to a single particle to reduce the number of particles. Secondly, the fluids are visualized by directly computing the intersection between ray and an isosurface defined by the surface particles. A dynamic particle grouping algorithm and equation solver are employed to quickly find the ray-isosurface intersection. Thirdly, based on the observation that the SPH simulation is a naturally parallel algorithm, the whole SPH simulation, including the adaptive sampling of the fluids as well as surface particle rendering, is executed on the GPU to fully utilize the computational power and parallelism of modern graphics hardware. Our experimental data shows that we can simulate about 50K adaptively sampled particles, or up to 120K particles in the fixed sampling case at a rate of approximately 20 time steps per second.

Statistics

Downloads

305 downloads since deposited on 29 Jan 2009
10 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Conference or Workshop Item (Paper), refereed, original work
Communities & Collections:03 Faculty of Economics > Department of Informatics
Dewey Decimal Classification:000 Computer science, knowledge & systems
Language:English
Event End Date:11 August 2008
Deposited On:29 Jan 2009 12:09
Last Modified:10 Aug 2017 19:41

Download

Preview Icon on Download
Preview
Filetype: PDF
Size: 2MB