摘要：The Network-on-Chip (NoC) approach is a promising solution to the increasing complexity of on-chip communication problems due to its high scalability. A NoC architecture design with ultra-low latency and high throughput is critical in order to support a wide range of applications. In this paper, we propose novel spatial-based NoC resource allocation algorithms to reduce the communication congestions so that efficiently improve the overall performance. We propose both offline (e.g. probability-based searching, genetic-based searching) and online (e.g. online greedy, one-corner, four corner) algorithms for NoC resource allocation. Based on our experiment results, the proposed algorithms show strong capability of improving the NoC performance.
关键词：Network-on-chip;multi/may-core processors;NoC resource allocation;2D compaction algorithms;spatial-based global optimization
