Parallel Cloth Simulation with Adaptive Mesh Refinement and Coarsening Using OpenMP on Fujitsu HPC2500(Parallel Processing Applications)

概要

論文の詳細を見る
In this paper, we propose a parallel implementation of simulating shape as well as force acting on cloth using particle based model on Fujitsu HPC2500 marchine. Inherent degree of parallelism, iterative nature of particle based model and high computational cost make it feasible to employ the parallelization using OpenMP. Since each particle is linked with its neighbors only, parallel model just needs to resolve some reduction and synchronization issues. In many cases, some regions of cloth mesh are more responsible for simulation while other regions have insignificant role. The efficient solution is to adaptively increase or decrease the mesh resolution in particular regions. Adaptive mesh refinement combined with coarsening (AMRC) reduces the simulation time of sequential algorithm but faster application demands its parallel implementation. Mesh density varies during execution due to AMRC and requires the dynamic redistribution of load. For this purpose, a load balancing scheme based on 'Active-Lists' has been developed to simulate the cloth. This scheme is compared with the dynamic scheduling constructs available in OpenMP. Result describes that we have succeeded to parallelize the cloth simulation with a satisfactory efficiency.
一般社団法人情報処理学会の論文
2004-10-15