Recently the hardware performance of mobile devices have been extremely increased\nand advanced mobile devices provide multi-cores and high clock speed. In addition,\nmobile devices have advantages in mobility and portability compared with PC and\nConsole, so many games and simulation programs have been developed under mobile\nenvironments. Physically-based simulation is a one of the key issues for deformable\nobject modeling which is widely used to represent the realistic expression of 3D soft\nobjects with tetrahedrons for game and 3D simulation. However, it requires high\ncomputation power to plausibly and realistically represent the physical behaviors and\ninteractions of deformable objects. In this paper, we implemented parallel cloth and\nmass-spring simulation using graphics processing unit (GPU) with OpenCL and multithreaded\ncentral processing unit (CPU) on a mobile device. We applied CPU and GPU\nparallel computing technique into spring force computation and integration methods\nsuch as Euler, Midpoint, 4th-order Runge-Kutta to optimize the computational burden\nof dynamic simulation. The integration methods compute the next step of positions\nand velocities in each node. In this paper, we tested the performance analysis for\nthe spring force calculation and integration method process using CPU only, multithreaded\nCPU, and GPU on mobile device respectively. Our experimental results\nconcluded that the calculation using proposed multi-threaded CPU and GPU multithreaded\nCPU are much faster than using just the CPU only.
Loading....