1 of 25

2024 ICCAD CAD Contest Problem B: �Power and Timing Optimization Using Multibit Flip-Flop

Problem Description, Benchmarks, and Results

2024-10-30

Sheng-Wei Yang, Jhih-Wei Hsu, Ting Wei Li, �Tzu-Hsuan Chen, Chin-Fang Cindy Shen

© 2024 Synopsys, Inc.

1

2 of 25

Synopsys Confidential Information

CONFIDENTIAL INFORMATION

The information contained in this presentation is the confidential and proprietary information of Synopsys. You are not permitted to disseminate or use any of �the information provided to you in this presentation outside of Synopsys �without prior written authorization.

 

IMPORTANT NOTICE

In the event information in this presentation reflects Synopsys’ future plans, such plans are as of the date of this presentation and are subject to change. Synopsys is not obligated to update this presentation or develop the products with the features and functionality discussed in this presentation. Additionally, Synopsys’ services and products may only be offered and purchased pursuant to an authorized quote and purchase order or a mutually agreed upon written contract with Synopsys.

© 2024 Synopsys, Inc.

2

3 of 25

Outline

  • Problem Description

  • Evaluation Methodology

  • Benchmarks

  • Results

Synopsys Confidential Information

© 2024 Synopsys, Inc.

3

4 of 25

Outline

  • Problem Description

  • Evaluation Methodology

  • Benchmarks

  • Results

Synopsys Confidential Information

© 2024 Synopsys, Inc.

4

5 of 25

Problem Description

Banking

  • Conglomerating multiple single-bit flip-flops into one multibit flip-flop is a useful technique for optimization
    • The movement is usually referred to as “banking”
    • Banking helps save power and area
    • May create dependencies on timing-critical paths

Synopsys Confidential Information

banking

© 2024 Synopsys, Inc.

5

6 of 25

Problem Description

Debanking

  • To take care of timing critical paths, sometimes tool needs to re-evaluate critical paths that end on banked multibit flip-flops.
  • This move is generally referred to as debanking
  • Trade-off between timing and power
    • Banking: Improve power & area. May sacrifice timing.
    • Debanking: Improve timing with power & area penalty.

Synopsys Confidential Information

 

1-bit FF

2-bit FF

Normalized Area

1

1.6

Normalized Power

1

1.7

Normalized Delay

1

2

© 2024 Synopsys, Inc.

6

7 of 25

Problem Description

Specs for Contestants

  • Testcase Input
    • Simplified design with combinational gates and flip-flops
  • Contestants are allowed to:
    • Move or size flip-flops
    • Bank or debank flip-flops
  • Output should satisfy the following constraints:
    • All instances must be placed within die region.
    • All instances must be non-overlapping and on-site.
    • banking and debanking mapping list needs to be provided
    • Nets must remain functionally equivalent
      • No open or short nets.

Synopsys Confidential Information

© 2024 Synopsys, Inc.

7

8 of 25

Outline

  • Problem Description

  • Evaluation Methodology

  • Benchmarks

  • Results

Synopsys Confidential Information

© 2024 Synopsys, Inc.

8

9 of 25

Evaluation Methodology

Total Negative Slacks (TNS)

  •  

Synopsys Confidential Information

© 2024 Synopsys, Inc.

9

10 of 25

Evaluation Methodology

Placement Legality Constraints

  • PlacementRows
    • Start from (<startX>, <startY>)
    • Has a <totalNumOfSites> of cell sites back-to-back placed horizontally
    • cell site: a small rectangle defined in PlacementRows with a dimension of (<siteWidth>, <siteHeight>)
  • On-site: cells with their bottom-left corner aligning to the site grid given from the PlacementRows
  • Cells should not overlap one another

Synopsys Confidential Information

© 2024 Synopsys, Inc.

10

11 of 25

Evaluation Methodology

Cell Density Constraints

  • Placement Density Constraint
    • Design is divided into placement bin units
    • For each bin the total area of the cells could not exceed the described utilization rate
    • The objective is to produce a legalized placement solution
    • Contestants should legalize flip-flops within each bin

Synopsys Confidential Information

© 2024 Synopsys, Inc.

11

12 of 25

Evaluation Methodology

Multiple Objective Optimization: Power, Performance, and Area (PPA)

  •  

Synopsys Confidential Information

© 2024 Synopsys, Inc.

12

13 of 25

Outline

  • Problem Description

  • Evaluation Methodology

  • Benchmarks

  • Results

Synopsys Confidential Information

© 2024 Synopsys, Inc.

13

14 of 25

Benchmarks

  • Testcase1: extracted from an initial opto design with only flip-flop datapath left.
  • Testcase2: extracted from a flatten design with initial opto and legalization done.
  • Testcase3: extracted from a flatten design with initial opto and legalization done.

  • Hidden testcases: design permutations from public testcases

Synopsys Confidential Information

Specs

Testcase1

testcase2

testcase3

hidden1

hidden2

hidden3

hidden4

#Insts

108685

153457

101221

108685

153457

153457

101221

#Nets

104235

153934

104761

104235

153934

153934

104761

#FlipFlops

19876

21164

13016

19879

21164

21164

13016

Weight Factor

411680

519031

336142

411710

519031

519031

366142

© 2024 Synopsys, Inc.

14

15 of 25

Outline

  • Problem Description

  • Evaluation Methodology

  • Benchmarks

  • Results

Synopsys Confidential Information

© 2024 Synopsys, Inc.

15

16 of 25

Results

Synopsys Confidential Information

  • Total Participants: 90 teams
  • Final Submissions: 44 teams
  • Submit with valid scores in at least one testcase: 41 teams

Team

Testcase1

testcase2

testcase3

hidden1

hidden2

hidden3

hidden4

Weighted �final score

1

693986404

745476

675559572

30175076

13212740

52124078

674676431

3107821

2

719733913

811024

693887313

30461537

13388840

54005613

693332444

3219048

3

690084983

835259

678819883

38930436

12555433

51444123

676665833

3258295

4

718444718

770638

691728642

41486870

15729265

52412042

689833099

3417434

5

711752359

814449

714627880

47948025

13958892

53565677

712609301

3494395

6

816288865

821653

803104066

38332727

12895728

59838790

799994828

3537665

7

727709964

719559

755949492

64544400

20782579

59036613

891296821

4108781

© 2024 Synopsys, Inc.

16

17 of 25

Honorable Mentions

Synopsys Confidential Information

© 2024 Synopsys, Inc.

17

18 of 25

Honorable Mention

Best Testcase2 Performance

Team: cadb1008

National Tsing Hua University

Yi-Ting Li

Wuqian Tang

Chi Hsu

Woei-Haur Hung

Prof. Ting-Chi Wang

Prof. Chun-Yao Wang

Synopsys Confidential Information

© 2024 Synopsys, Inc.

18

19 of 25

Honorable Mention

Sixth place

Team: cadb1009

University of Wisconsin Madison

Aditya Das Sarma

Prof. Tsung-Wei Huang

Synopsys Confidential Information

© 2024 Synopsys, Inc.

19

20 of 25

Honorable Mention

Fifth place

Team: cadb1051

The Chinese University of Hong Kong

Bangqi Fu

Yutao Wang

Xingyu Cui

Prof. Evangeline F.Y. Young

Synopsys Confidential Information

© 2024 Synopsys, Inc.

20

21 of 25

Honorable Mention

Fourth place

Team: cadb1038

Fudan University

Benchao Zhu

Jiawei Li

Yang Liu

Prof. Jianli Chen

Synopsys Confidential Information

© 2024 Synopsys, Inc.

21

22 of 25

Winners

Synopsys Confidential Information

© 2024 Synopsys, Inc.

22

23 of 25

Third Place

Team: cadb0017

National Yang Ming Chiao Tung University

Huan-Yuan Chen

Lin Yu-Ruei

Yu-Chia Chen

Prof. Mark Po-Hung Lin

Prof. Hung-Ming Chen

Synopsys Confidential Information

© 2024 Synopsys, Inc.

23

24 of 25

Second Place

Team: cadb0015

National Yang Ming Chiao Tung University

Chi-En Ho

Chieh-Yu Kuo

Wen-Yan Cheng

Prof. Hung-Ming Chen

Synopsys Confidential Information

© 2024 Synopsys, Inc.

24

25 of 25

First Place

Team: cadb0027

Prof. Yao-Wen Chang

Synopsys Confidential Information

Cheng-Yen Li

National Taiwan University

Chuan-Chi Su

Shao-Hsiang Chen

Zheng-Wei Chen

National Central University

© 2024 Synopsys, Inc.

25