Exploiting Multicore Architectures for Physically Based Simulation of Deformable Objects in Virtual Environments

Physically based simulation is an indispensable component of many interactive virtual environments. The main challenge of virtual reality applications is the realtime requirement. Advanced simulation methods as, e.g., the finite elements method (FEM) require significant computational power. However, the performance increases due to higher clock speed are tapering off. Instead, the compute power is increased by replicating processing units, making parallel computing a necessity for all performance demanding applications. In this paper we analyze the runtime and scalability of a dynamic FEM simulation on dualcore and quadcore architectures. Methods for both small and large deformation simulation are tested and parallelized using OpenMP. The algorithms significantly profit from the multicore architectures, with minimal changes to the serial code.