K-way partitioning under timing, pin, and area constraints

Abstract

Circuit partitioning is a very extensively studied problem. Our proposed methodology easily extends to multiple constraints that are very dominant in the design of large scale VLSI systems. In this paper we formulate the problem as a nonlinear program (NLP). The NLP is solved for the objective of minimum cutset size under the constraints of pins, area, and timing. We have tested the unified framework for area, timing, and pin constraints. The NLP is solved using the commercial LP/NLP solver MINOS. We have done extensive testing using large scale RT level benchmarks and have shown that our methods can be used for exploring the design space for obtaining constraint satisfying system designs. We also provide extensions for solving system design problems where a choice between multiple technologies, packaging components, performance, cost, yield, and more can be the constraints for design related decisions.