1 Introduction
CHIP Multi-Processors (CMPs) have emerged as ubiquitous solution to attaining performance within thermal constraints. Meanwhile, continued increases toward multicore and manycore scalability have led to interconnection challenges due to wiring, energy, and bandwidth limitations. From a performance perspective, interconnection networks are crucial since they connect core elements and also provide off-chip communication. Previous studies have shown that up to half of the dynamic power dissipation can be associated with interconnection in high-performance processors [1]. In addition, as networks scale up, electrical-only means may not be capable of fulfilling communication requirements because of the needed data rate increments and the distance between cores impacting signal drive implications, thus imposing further restrictions on wire densities [2], [3].