期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2019
卷号：97
期号：18
页码：4850-4859
出版社：Journal of Theoretical and Applied
摘要：We introduce a fluid simulation method using a upsampling-based grid projection and CPU-GPU heterogeneous computing that can explicitly represent and preserve fluid thin sheets. The most important contribution of this paper is that the GPU and grid projection are used to improve the particle-based framework to prevent holes that appear to collapse at thin fluid surfaces or dense points. The proposed framework can handle topology change reliably without numerical diffusion or twisting problems. The anisotropic kernel and PCA (Principal Component Analysis) are performed on the GPU to quickly find the direction of fluid with a thin surface, and CPU-GPU heterogeneous computing technique greatly enhances the efficiency of the candidate position extraction process to calculate the position of a new fluid particle. Our technique is intuitively implemented, allowing for easy parallelization and rapid visualization of the surface of thin sheets of visually detailed water. As a result, we propose a parallel processing framework that can quickly express surface details, such as liquid sheets, that are difficult to represent in traditional particle-based simulations and improve the surface details of the water simulation.
关键词：CPU-GPU heterogeneous computing; Thin sheets of water; Fluid simulation
