Algorithms for Simultaneous Exploration of Multi-Domain Transforms for Design Closure in Emerging Technologies

University Of Illinois At Chicago, Chicago IL

Investigators

Abstract

Proposal No: 0811855 PI name: Dutt, Shantanu S Title: Algorithms for Simultaneous Exploration of Multi-Domain Transforms for Design Closure in Emerging Technologies Institution: University of Illinois at Chicago ABSTRACT In this project the PIs plan to accomplish the following goals. (a) develop network-flow based meta-structures and algorithms that simultaneously consider multiple physical synthesis and physical design transformations, and does so simultaneously across all problem parts of a placed circuit, in order to produce a near-optimal constraint-satisfying transformation on the circuit so that design closure is quickly achieved; (b) The physical design transformations will include replacement and global rerouting performed together - an important consideration with little work on it so far; (c) the algorithms developed will address constraint metrics like performance, power, signal integrity, routability, wire length, power density and temperature; (d) the PI will develop discretized network flow techniques (network flow is a continuous optimization method) in order to achieve the twin benefits of near-optimal solutions and fast run-times for discrete optimization problems of the above type. A successful completion of this research project will result in tools that will enable VLSI designers to achieve high-quality large SOC designs for emerging technologies in tractable times. This will contribute significantly to making these new technologies viable, and in turn will positively impact a myriad of new applications of these advanced technologies in areas ranging from medicine to sensor-based environmental monitoring. Graduate and undergraduate students will be trained in this area via research participation and a course in VLSI CAD which will include topics on logic and physical synthesis. Graduate student training will include Ph.D. students whose theses will be based on the topics in this project.