1 / 8

An Improved Two-Way Partitioning Algorithm with Stable Performance

An Improved Two-Way Partitioning Algorithm with Stable Performance. Chung-Kuan Cheng Yen-Chuen A. Wei. Motivation. KL based Algorithms highly sensitive to initial conditions Can be solved by multiple runs of the Algorithm

ophelia
Download Presentation

An Improved Two-Way Partitioning Algorithm with Stable Performance

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. An Improved Two-Way Partitioning Algorithm with Stable Performance Chung-Kuan Cheng Yen-Chuen A. Wei

  2. Motivation • KL based Algorithms highly sensitive to initial conditions • Can be solved by multiple runs of the Algorithm • Probability of finding good initial conditions drop exponentially with circuit size

  3. Tools(1) – Ratio Cut • Two way partitioning Algorithm • Finds partitioning with best ratio using RAB = CAB / (|A| x |B|) • Partitioning not necessarily even

  4. Tools(2) - FM • Fiduccia-Mattheyses Partitioning • Move node with best gain that satisfies the balance condition from one partition to the other until all are locked

  5. Step One - Divide • Divide the design into two partitions using ratio cut • Keep applying ratio cut to any partition larger than #Nodes/g • Authors suggest g = 50

  6. Step Two – Conquer • Form Supernodes • Partition Supernodes using FM number_of_reps times using size constraints • Authors suggest 10 for number_of_reps • Choose best results from FM

  7. Step Three – Clean Up • Expand the Supernodes back out • Apply one more run of FM

  8. Experimental Results

More Related