Acceleration of Bullet Physics Physical Simulation Library Using GPU and Demonstration on a Set-Top Box Platform

Abstract

Bullet physics library is an open source, professional physics library which does the 3D collision detection, applies laws of physics on the rigid bodies and performs the rigid body and soft body dynamics. It is used in games, movies for simulating the rigid bodies according to the laws of physics. Bullet uses the inbuilt feature for continuous tracking and updating the world transform of the colliding and the moving object. The main aim of the paper is utilization of GPU for visual effects through bullet physics.GPU computing is the use of a GPU (graphics processing unit) together with a CPU to accelerate general-purpose scientific and engineering applications [2]. CPU in association with GPU proves to be powerful combination as CPU contains optimized cores for serial processing while GPU consist of 100s of cores which can be used for parallel computation. GPU can act as an accelerator for compute intensive processing by offloading those tasks on to GPU [1].In this paper we intent to harness this unused processing power of the GPU for performing the complex graphics operations as well as the compute intensive operations of the bullet physics library. An application is developed using shaders for running it on the graphics processing unit and integrate the bullet physics library into it. The GPU used for this process is Nvidia graphics card, Quadro NVS 295. Finally, the code is ported onto the Pandaboard and Snowball for ensuring it's compatibility with the orly platform.