GPU-based implementation of reverb effect

Graphics processing units (GPUs) have been increasingly used in different engineering applications in the past several years. One of the main application areas is digital signal processing, especially in the domain of audio, image, and video processing. Fast Fourier Transform (FFT) is often core part of those processing algorithms, and it is efficiently implemented on the GPUs through available libraries. In this paper, we focus on audio signal processing, and present an implementation of convolution reverb effect on the NVIDIA GPUs using CUDA. Implemented effect is tested with carefully chosen set of examples and we observed significant speedups over reference, CPU implementation. Finally, the results are briefly discussed with emphasize on benefits and drawbacks of using GPUs in such application.