Development of real-time audio processing applications for GPGPUs is not without challenges. Parallel processing of audio signals is often constrained by serial dependencies within or between algorithms. On GPGPUs, insufficient data pressure further limits the attainable performance improvements, as it causes inactivity of the GPU cores. In this paper, we analyze the limits of audio processing on GPGPUs and present an approach based on event-driven scheduling, that maximizes data pressure to favour performance improvements. We also present recent enhancements of Audio n-Genie, an open-source development environment for audio-processing applications. By combining Audio n-Genie and the proposed approach, we show that it is possible to increase audio processing speed-up.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.