Cloth Simulation Using AABB Hierarchies and GPU Parallelism

Providing realistic, high-resolution and highfidelityrepresentation of motions ia essential in the clothsimulation problem. In order to make high resolution simulationstractable, several algorithms have been developed thatmanage cloth-object interactions efficiently through specializeddata structures such as AABB trees. However, implementationrestrictions on single CPU architectures impose certain limitson quality and performance in high-demanding simulations,motivating the study of new implementation techniques. Inthis paper we address several critical issues in high resolutioncloth simulation, enabling us to represent and simulateintricate folds and wrinkles. We employ AABB hierarchiesto optimize detection and response in cloth-object collisions.By employing a multi-processor approach on multi-threadedCPU and an emerging multi-core GPU-CUDA architecture, wequantitatively evaluate the workload and computational effortof the cloth simulation application. In addition to this quantitativeperformance evaluation on multi-processor architectureswe illustrate the potential of our approach by presenting avariety of high-quality and high-resolution simulations of clothbehavior under different cloth-object interactions.