Congestion-Driven Global Routing Based on Timing-Constrained Routing Flexibilities
碩士 === 中華大學 === 資訊工程學系碩士班 === 92 === Because the semiconducting processing technology progresses, and system-on-a-chip design has became a workable design model. Today a chip can accommodate more than hundreds million of transistors. And all of the signals inside a chip must be transmitte...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2004
|
Online Access: | http://ndltd.ncl.edu.tw/handle/96067162656203833288 |
Summary: | 碩士 === 中華大學 === 資訊工程學系碩士班 === 92 === Because the semiconducting processing technology progresses, and system-on-a-chip design has became a workable design model. Today a chip can accommodate more than hundreds million of transistors. And all of the signals inside a chip must be transmitted by routing wire. Hence routing plays a key role in the chip’s world.
Now the congestion of routing wire will generate harmful effect. Hence the problem must be solved during routing plan. And today chips have strict restriction of timing constraint. Thereby timing constraint and congestion must concurrently consider in today’s global routing. Our paper provides an algorithm that can solve the congestion problem and conform to the timing constraint. We use the character of Steiner-point that has timing-constrained location flexibility to improve the congestion problem. It is called initial timing-constrained congestion control. Then we will assign the routing path of every net and the action can also improve the congestion problem. So it is called timing-constrained congestion-driven path assignment. Finally we use simulated-annealing-based rip-up-and-reroute to modify the congestion problem. It uses Steiner-point reassignment and reroute, Steiner-point deletion and reroute, L-pattern flip, and path detour to solve the remaining congestion problems.
Finally we use timing-constrained congestion-driven global routing algorithm and simulated-annealing-based rip-up-and-reroute improve to execute some benchmark files. We can achieve 100% routability.
|
---|