Particle-Based Fluid Simulation on the GPU

Abstract

Abstract. Large scale particle-based fluid simulation is important to both the scientific and computer graphics communities. In this paper, we explore the effectiveness of implementing smoothed particle hydrodynamics on the streaming architecture of a GPU. A dynamic quadtree structure is proposed to accelerate the computation of inter-particle forces. Our method readily extends to higher dimensions without undue increase in memory or computation costs. We show that a GPU implementation runs nearly an order of magnitude faster than our CPU version for large problem sizes.

SPH simulation of oil displacement in cavity-fracture structures

“…Employment of SPH for a unified Lagrangian formulation for both elastic wall and immiscible fluids rather than some coupled models is quite promising, but the challenging problem of fluid structure interaction (Dowell and Hall, 2001) has to be solved in the first place. Although there have been massive parallel computation and lots of other advanced computation technologies, such as graphics processor unit (Hegeman et al, 2006;Yang et al, 2007;Chen et al, 2009), it has to be said that a large scale three-dimensional simulation of oil recovery with realistic structures would still be beyond the reach at the moment, and maybe the multi-scale method using SPH together with some average treatments would be more feasible.…”

Six Degrees-of-Freedom Haptic Interaction with Fluids

“…The only exception is the work presented in [1], which is not entirely computed in realtime, limiting the interaction possibilities. The higher performance is achieved through the use of the SPH model and our efficient GPU implementation, and performing better than [9] and [10] (12 and 1.7 times faster, respectively), and close to [11] (1.2 times slower), which focus exclusively on the optimization of a particle-based simulation of fluids. The different force and torque plots of section 6.4 provide a visual representation of the force feedback.…”

“…In order to accelerate the computation of the simulations, research has focused on the implementation of the SPH algorithm on GPU, being naturally parallel. Early attempts [8] [9] were not fully GPU-based. In [10], the entire simulation runs on GPU, by using a texture representation of a grid space subdivision for neighbor computation purposes.…”

Multi-agent traffic simulation with CUDA

“…The GPGPU toolkits have been widely adopted by researchers and industry alike. Todays GPUs are used in many fields, for example in molecular dynamics [18], gas and fluid dynamics [16], astro-physics [10,11], for coupled map lattices [14] genetic programming [23], graph algorithms [12] as well as DNA sequencing [24] or even database queries [15]. Most of these examples bear in common, that they involve very computation-intensive operations and are known as being highly adaptable to a SIMD architecture.…”