AdEle+: An Adaptive Congestion-and-Energy-Aware Elevator Selection for Partially Connected 3D NoCs

Unlike conventional 2D Networks-on-Chip (NoCs), 3D NoCs offer a scalable and energy efficient on-chip communication. Vertical die stacking of is enabled using inter-layer Through-Silicon Via (TSV) links. However, TSV technology suffers from low reliability high fabrication costs. To mitigate these costs, Partially Connected (PC-3DNoCs), which use fewer vertical links, have been introduced. with links (a.k.a. elevators), elevator-less routers will to send their traffic nearby elevators for traffic, increasing the load congestion at potentially reducing performance. Therefore, it important that choose balance among available elevators. address this problem, paper presents an ad aptive congestion- energy-aware xmlns:xlink="http://www.w3.org/1999/xlink">ele vator-selection algorithm, called AdEle+. AdEle+ employs offline multi-objective simulated-annealing-based optimization find good elevator subsets routers. In addition, during loads, uses adaptive online selection algorithm select subset dynamically manage on Moreover, in circumstances, switches minimal distance improve efficiency. Compared state-of-the-art algorithms under various PC-3DNoC configurations patterns, reduces average latency by 9.5% up 11.2% while hardware overhead 10.1% due its