Evaluation of the New OASIS Format for Layout Fill Compression

1
Evaluation of the New OASIS Format for Layout
Fill Compression
Supported by Cadence Design Systems Inc., U.S.
National Science Foundation, and the Packard
Foundation
Yu Chen, Andrew B. Kahng, Gabriel Robins,
Alexander Zelikovsky and Yuhong Zheng UCLA,
UCSD, UVA, GSU vlsicad.ucsd.edu
2
Overview

• CMP and Interlevel Dielectric Thickness
• Motivation
• Repetition Operators in GDSII
• Repetition Operators in OASIS their Compression
  Ratios
• Compressible Fill Generation algorithms and
  results
• Fill Compression Problem algorithms and results
• Conclusions Suggestions

3
CMP and Interlevel Dielectric Thickness

• Chemical-Mechanical Planarization (CMP)
• wafer surface planarization
• Uneven features cause polishing pad to deform

Post-CMP ILD thickness
Features

• Interlevel-dielectric (ILD) thickness ? feature
  density
• Insert dummy features to decrease variation

Post-CMP ILD thickness
4
Motivation

• Dummy features grow layout size to terabytes
• ? Need in more compressible format
• New format
• OASIS Open Artwork System Interchange
  Standard
• Our focus
• How to exploit new compression capabilities of
  OASIS?
• How to improve proposed OASIS for further fill
  compression?

5
Repetition Operators in GDSII
AREF array reference
SREF structure reference
6
Repetition Operators in OASIS
Type 1
Type 3
Type 2
Type 5
Type 4
Type 7
Type 6
Type 8
7
Compression Ratio of OASIS Repetitions

• Compression Ratio CR
• size flat OASIS / size compressed OASIS

CR ( , ) T(MN)
CR ( , , ) T(M)
8
Compression Ratio of OASIS Repetitions

• Compression Ratio of Types 4,5 and 8

CR ( , ) 7
CR ( ) 3.5
9
Compressible Fill

• Compressible Fill Generation Problem (CFGP)
• Given a design rule-correct layout, create the
  minimum number of OASIS operators to represent
  area fill features that keep window density
  variation within the given bounds (L,U)

Filled layout with 82 area features
Original layout
10
Greedy Compressible Fill Generation
11
Compressible Fill Generation GZIP
GZIP applied to compressible fill compress slower
and worse than native generator GCF GZIP and
GCF are orthogonal to each other C_Ratio(GCFGZI
P) C_Ratio(GCF)C_Ratio(GZIP)
12
Compressible Fill Generation Results
I Restricted OASIS ( GDSII) II Full
OASIS III Types 1,2,3,4,5 IV Types 4,5
13
Fill Compression Problem

• OASIS Fill Data Compression Problem
• Given a layout containing are fill features,
  represent these area fill features using OASIS
  repetition operators in a way that minimizes the
  resulting data volume

Finding regularities
14
(No Transcript)
15
Exhaustive Search Greedy Fill Compression
16
Fill Compression Results
Rest GDSII Full Full OASIS ESBG
Exhaustive search RSBG Regularity search
17
Fill Compression Results
18
Conclusions Suggestions

• Contributions
• New algorithms for compressible fill generation
• New fill compression algorithms for OASIS format
• Comparison of GDSII and OSIS formats

• Suggestions to enhance OASIS
• Add irregular array OASIS repetition
  combination of 4 5
• Include pseudo-random number generator in OASIS,
  for fast and simple generation of irregular area
  fill