MoC-based Realization of Multi-core Speech Encoders

摘要

In this paper, we demonstrate a teehnique for mapping a multimedia streaming application into a mesh NoC using an example of speech encoder named SPEEX. To decrease the size of the target mesh, we use an algorithm for merging functional blocks using various metrics, such as core code size or execution time. We propose and test three algorithms for core mapping. According to the presented experimental results, the process of assigning the functional block into the NoC mesh is strongly influenced by the selected strategy.