Universidade Federal de Mato Grosso do Sul

This paper presents the use of GPU as a math co-processor in real-time applications, in special games and physics simulation. Game loop architecture is used to validate the use of GPU such as math and physics co-processor, thus it is shown by this paper a new game loop architecture that employs graphics processors (GPUs) for gen-eral-purpose computation (GPGPU). A critical issue in this field is the process distribution between CPU and GPU. The presented architecture consists in a model for distribution and our implemen-tation showed many advantages in comparison to other approaches without a GPGPU stage. The architecture presented here was mainly designed to support mathematics and physics on the GPU, therefore the GPU stage in the model proposed works to help the CPU such as a math co-processor. Nevertheless, any kind of generic computation can be adapted. The model is implemented in an open source game engine and results obtained using this platform are presented. Keywords:: game loop, GPGPU, co-processor. Author’s Contact