Degree Type

Dissertation

Date of Award

2011

Degree Name

Doctor of Philosophy

Department

Electrical and Computer Engineering

First Advisor

Chris C. Chu

Abstract

In the nanometer scale era, placement has become an extremely challenging stage in modern Very-Large-Scale Integration (VLSI) designs. Millions of objects need to be placed legally within a chip region, while both the interconnection and object distribution have to be optimized simultaneously. Due to the extensive use of Intellectual Property (IP) and embedded memory blocks, a design usually contains tens or even hundreds of big macros. A design with big movable macros and numerous standard cells is known as mixed-size design. Due to the big size difference between big macros and standard cells, the placement of mixed-size designs is much more difficult than the standard-cell placement.

This work presents an efficient and high-quality placement tool to handle modern large-scale mixed-size designs. This tool is developed based on a new placement algorithm flow. The main idea is to use the fixed-outline floorplanning algorithm to guide the state-of-the-art analytical placer. This new flow consists of four steps: 1) The objects in the original netlist are clustered into blocks; 2) Floorplanning is performed on the blocks; 3) The blocks are shifted within the chip region to further optimize the wirelength; 4) With big macro locations fixed, incremental placement is applied to place the remaining objects. Several key techniques are proposed to be used in the first two steps. These techniques are mainly focused on the following two aspects: 1) Hypergraph clustering algorithm that can cut down the original problem size without loss of placement Quality of Results (QoR); 2) Fixed-outline floorplanning algorithm that can provide a good guidance to the analytical placer at the global level.

The effectiveness of each key technique is demonstrated by promising experimental results compared with the state-of-the-art algorithms. Moreover, using the industrial mixed-size designs, the new placement tool shows better performance than other existing approaches.

DOI

https://doi.org/10.31274/etd-180810-779

Copyright Owner

Zijun Yan

Language

en

Date Available

2012-04-30

File Format

application/pdf

File Size

155 pages

Share

COinS