Acoustic simulation based on wave equation solver can achieve accurate results. Due to the computation complexity. however, its practical application has been limited. One solution is to exploit the parallel processing power of modern GPUs to accelerate runtime computation. Nonetheless, GPU implementation of wave equation solvers is a non-trivial task, and the main difficulty lies in maximizing the parallelism of computation while mak-ing sure it fits into available GPU memory.In this paper, we studied the major wave equation solving techniques and matching layer technology. based on which we implement the acoustic simulation pipeline on GPUs and optimize the implementation on the Geforce GTX285. We also propose techniques to solve the inter-GPU communication and load balancing and implemented the whole pipeline on a multi-GPU platform. This enables simulation of larger scenes. Finally, we propose a solution of integrating out techniques into 3D game engine. |