Use of GPUs in room acoustic modeling and auralization
Lauri Savioja, Dinesh Manocha, Ming C. Lin · 2010
All room acoustic modeling techniques are computationally demanding due to the underlying complexity of sound propagation. Traditionally, all the modeling methods have been implemented for the central processing unit (CPU) of the computer. Recent trends in terms of programmable many-core processors or graphics processing units (GPU) is resulting in a paradigm shift. Modern GPUs are massively parallel processors with a different memory hierarchy, whose peak computational performance can be 10X more than that of current CPUs. In this paper, we give an overview of GPU architectures and address issues in designing suitable algorithms that map well to GPU architectures. The room acoustic modeling and auralization techniques are especially discussed in the light of possible GPU-based implementations. We also give a brief overview of recent methods for geometric and numeric sound propagation that offer one order of magnitude speedup over CPU-based algorithms.