ELASTIC AND DEPENDENCY-AWARE NOC MAPPING IN RECONFIGURABLE SYSTEMS

10.6329/CIEE.2016.2.04

Chun-Hsian Huang；Mei-Ru Chen；Yi-Syuan Wu

Task mapping ； reconfigurable systems ； Network-on-Chip

電機工程學刊

23卷2期（2016 / 04 / 01）

63 - 72

英文

To support more efficient parallel processing, the Network-on-Chip (NoC) is being increasingly adopted as a new alternative interconnection architecture. Further, due to the support of partial reconfiguration technique, hardware functionalities in an FPGA device can be dynamically reconfigured on-demand for different applications. By integrating the partial reconfiguration with the NoC infrastructure, the NoC-based reconfigurable system thus becomes an attractive alternative for System-on-Chip (SoC) design. However, this adaptable system design also makes the issue of task mapping on the NoC more complex. As a result, to efficiently map application tasks to the processing elements (PEs) in NoC-based reconfigurable systems has thus become an increasingly important issue. In this work, we try to solve the above issue by proposing an elastic and dependency-aware task (EDA) mapping method. By using the PE reservation heuristic, more applications can be performed on the NoC-based reconfigurable system. Through the dependency- aware task allocation, the communication latencies between the mapped tasks on PEs can be further reduced. Experiments have also showed that, compared to the state-of-the-art task mapping methods, using the proposed EDA method, the total processing time can be reduced by up to 49.98%.