Audio on the GPU: Real-Time Time Domain Audio Convolution on Graphics Cards

ASU Author/Contributor (non-ASU co-authors, if there are any, appear on document)
Andrew Keith LaChance (Creator)
Appalachian State University (ASU )
Web Site:
Rahman Tashakkori

Abstract: The architecture of CPUs has shifted in recent years from increased speed to more cores on the chips. With this change, more developers are focusing on parallelism; however, many developers have not taken advantage of a common hardware component that specializes in parallel applications: the Graphics Processing Unit (GPU). By writing code to execute on GPUs, developers have been able to gain increased performance over the traditional CPU in many problem domains, including signal processing. Time domain convolution is an important component of signal processing. Currently, the fastest process to perform convolution is frequency domain multiplication. In addition to being more complex, inconsistencies such as missing data are difficult to solve in the frequency domain. It has been shown that executing frequency domain multiplication on GPUs improves performance, but there is no research for time domain convolution on GPUs. This thesis provides two algorithms that implement time domain convolution on GPUs: one algorithm is for computing convolution all at once and another is designed for real time computation and playing the results. The results from this thesis indicate that using the GPU significantly reduces processing time for time domain convolution.

Additional Information

LaChance, A.K. (2011). Audio on the GPU: Real-Time Time Domain Audio Convolution on Graphics Cards. Unpublished master’s thesis. Appalachian State University, Boone, NC.
Language: English
Date: 2011
Convolution, GPU, Graphics Card, Audio, Real Time