Multiobjective optimization of deadspace, a critical resource for 3D-IC integration

In 3D-IC integration and its implied resource optimization, a particularly critical resource is deadspace - regions between floorplan blocks. Deadspace is required for through-silicon via (TSV) planning and other related design tasks, but the effective use of this limited and highly-contested resource requires effort. While most previous work focuses on a single design issue at a time, we propose a lightweight multiobjective deadspace-optimization methodology that simultaneously optimizes interconnect, IR-drop, clock-tree size and maximal temperature. This methodology repeatedly re-evaluates design quality during early chip planning and uses resulting information to guide further optimization. Experimental results indicate that constructing an appropriate deadspace distribution improves design tradeoffs and is effective in practice.