43
1 1 Jason Cong Majid Sarrafzadeh ISPD 2000: April 10-12 C Incremental Physical Design Jason Cong UCLA Majid Sarrafzadeh Northwestern 2 Jason Cong Majid Sarrafzadeh ISPD 2000: April 10-12 C Outline l PART I: Introduction & Motivation l PART II: Partitioning, Floorplanning, and Placemenet l PART III: Routing l PART IV: Conclusion

Incremental Physical Design - VAST labcadlab.cs.ucla.edu/~cong/slides/ispd2k.pdfIncremental Physical Design Jason Cong UCLA Majid Sarrafzadeh Northwestern 2 Jason Cong ISPD 2000: April

Embed Size (px)

Citation preview

1

1

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Physical Design

Jason Cong UCLA

Majid Sarrafzadeh Northwestern

2

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Outline

l PART I: Introduction & Motivation

l PART II: Partitioning, Floorplanning, and Placemenet

l PART III: Routing

l PART IV: Conclusion

2

3

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

PART I:Introduction & Motivation

4

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Introduction & Motivation

l Concurrent optimization:l problems are getting complex

l need quick evaluation of a number of alternatives

l Existing algorithms are “incremental”l what can we say about the quality of the final solution?

l when to do, and how much, incremental “moves”?

l A more fundamental concept: Given a solution, assessits quality

l Design for change

l The right data structures

3

5

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Current Methods

l Ad-hoc (low-temp annealing or rip-up and reroute)

l No or very little understanding of the solution quality

l Painful and slow updates

l May have undesired global impact

l There has been very little done on the topic

l (We provide a reasonable list of references as a startingpoints: data structures, layout algorithms, synthesisalgorithms, etc)

l Today’s presentation is meant to serve as a motivator

6

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

PART II:Partitioning, Floorplanning, and

Placement

4

7

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Partitioning

8

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning

l An original (hyper)graph and its near-optimal partitioningresults are given.

l An incrementally changed (hyper)graph is obtained byadding/deleting vertices/edges from the original graph.

l Incremental partitioning problem is to find the near-optimal partitioning for the changed (hyper)graphwithout doing from the scratch.

l Incremental partitioning should be much faster than theoriginal partitioning.

l It is preferable to use the existing partitioning results forthe original (hyper)graph.

5

9

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Original Graph

10

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

One Vertex Changed (e.g., resized)

6

11

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

One Edge Added (logic restructuring)

12

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

3rd Vertex Changed

7

13

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

0-proximity Vertices

0-proximity

0-proximity

0-proximity

0-proximity

0-proximity

14

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

1-proximity Vertices

0-proximity

0-proximity

0-proximity

1-proximity

1-proximity

0-proximity

0-proximity1-proximity

8

15

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

2-proximity Vertices

0-proximity

0-proximity

0-proximity

1-proximity

1-proximity

0-proximity

2-proximity

1-proximity 0-proximity

16

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Partitioning Problem on the OriginalGraph

9

17

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Partitioning Problem on theChanged Graph (inf-threshold)

0-proximity

0-proximity

0-proximity

1-proximity

Changed Vertices

0-proximity

18

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

0-Threshold IncrementalPartitioning

0-proximity

0-proximity

0-proximity

0-proximity

0-proximity

10

19

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

1-Threshold IncrementalPartitioning

0-proximity

0-proximity

0-proximity

1-proximity

1-proximity

0-proximity

0-proximity

20

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 1%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

1.2

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Net-cut results normalized to infinity-threshold results.

11

21

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 1%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Runtime normalized to infinity-threshold results.

22

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 5%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

1.2

1.4

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Net-cut results normalized to infinity-threshold results.

12

23

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 5%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

1.2

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Runtime normalized to infinity-threshold results.

24

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 10%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Net-cut results normalized to infinity-threshold results.

13

25

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 10%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

1.2

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Runtime normalized to infinity-threshold results.

26

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 20%Change In the Original Graph

0

0.5

1

1.5

2

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Net-cut results normalized to infinity-threshold results.

14

27

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 20%Change In the Original Graph

0

0.2

0.4

0.6

0.8

1

1.2

ibm01 ibm02 ibm03 ibm04 ibm05

0-threshold 1-threshold

2-threshold inf-threshold

Runtime normalized to infinity-threshold results.

28

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 1%Change In the Original Graph

15

29

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 5%Change In the Original Graph

30

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 10%Change In the Original Graph

16

31

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Partitioning with 20%Change In the Original Graph

32

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Some more ideas

l Localized proximity

l Net-list dependent proximity

l New/ targeted algorithms (not “locked full partitioner ”)for incremental partitioning

l Create an initial solution that is suitable for change

17

33

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Floorplanning

34

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Floorplanning

l Do not resize the floorplan if not needed or if not needed“badly” (based on an enhanced sizing tree) [Crenshaw-Sarrafzadeh]

Time Ratio Plot

1

10

100

1000

10000

Vec1x Vec5x Vec10x Vec50x Vec100x

Vector File

Ex10

Ex20

Ex50

18

35

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

A slack based sizing tree(for quick decision making)

w+ - slack in widthh+ - slack in height

Horizontal critical pathVertical critical path

6

1

2 3

4

5

7

56

105

4

5

5

5

87

8

1011

14

7

7

8-7

2

1 6

53w+=0h+= 2w+=0

h+= 0

w+=0h+= 0

w+=3h+= 0

w+=0h+= 2

w+=0h+= 0

w+=0h+= 0

4

w+=0h+= 0

w+=0h+= 0

w+= 1h+= 0

w+=2h+= 0

10-8

TSSmodule width height

1 3 22 0 23 0 44 3 05 2 06 0 0

36

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Speed-up after a series of moves(adding buffers, sizing, gate duplication

User Times for Finding Area Improving Moves

0.00

100.00

200.00

300.00

400.00

500.00

600.00

700.00

800.00

900.00

fract

-40

fract

-50

fract

-60

fract

-70

fract

-90

stru

ct-9

0

stru

ct-1

00

stru

ct-1

20

stru

ct-1

30

stru

ct-1

50

primary

1-70

primary

1-80

primary

1-90

primary

1-100

primary

1-130

primary

2-70

primary

2-80

primary

2-90

primary

2-120

primary

2-150

benchmarks

CP

U r

un

tim

e (s

)

slack-aware

greedy-local

greedy-global

19

37

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

(Floorplan) Design for Change

l Accurate data is not available in early phases:l Missing cells: Some cells have not been designed yet. Use

estimates from previous design.l Formulate designer’s experience in estimation of the cell area.

l Evolving (cell / IP) library: Final area of modules can be estimatedfrom current area.

l Geometric synthesis, timing improvement, ….

l Nostradamus: A Floorplanner of Uncertain Designs [Bazargan-Kim-Sarrafzadeh]

38

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Problem Formulation

l Distribution lists for width/height:

l Wi = { (wi1,p(wi1)),…,(wi,mi,p(wi,mi)) } Σj p(wij) = 1

l Hi = { (hi1,p(hi1)), … , (hi,ni,p(hi,ni)) } Σj p(hij) = 1

l Example:l W1 = { (5,0.3) , (7,0.5), (8,0.2) }l W2 = { (2,0.9), (3,0.1) }

l H1 = { (1,.1),(2,.2),(7,.7) }l H2 = { (4,.4),(6,.6) }

l We could also use:l Li = { (wi1, h i1, pi1), … , (wim, him, pim) }

20

39

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Output Distributions

40

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Input/Output CorrelationWidth: Input and output std dev of diff data sets

020406080

100120140

a1 a3

b1

b3 c1 c3 c5 d1 d3

e1 e3 f1 f3 g1

h1 h3

c6f1

data sets

std

de

v

Ouput deviation Scaled input deviation

Height: Input and output std dev of diff data sets

01020304050607080

a1

a3 b1

b3 c1 c3 c5 d1 d3

e1

e3 f1 f3 g1 h1

h3

c6f1

data sets

std

de

v

Output deviation Scaled input deviation

21

41

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Experimental Results

Ratio of actual area of traditional methods to area of Nostradamus : 30% uncertainty

0%

20%

40%

60%

80%

100%

120%

140%

a1 a2 a3 a4 b1 b2 b3 b4 f1 f2 f3 f4 g1 g2

Conservative Optimistic Exp. Val.

42

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Experimental Results

Ratio of actual area of traditional methods to area of Nostradamus : 10% uncertainty

0%

20%

40%

60%

80%

100%

120%

140%

a1 a2 a3 a4 b1 b2 b3 b4 f1 f2 f3 f4 g1 g2

Conservative Optimistic Exp. Val.

22

43

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Placement

44

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Placement Problem

Input:• A set of cells and their complete information (a cell library).• Connectivity information between cells (netlist information).

Output:• A set of locations on the chip: one location for each cell.

Goal:• Total chip area and wirelength, under timing constraints.

Challenge:• Rapid growing circuit size (> 1 million).• Multi objectives (timing constraint, routability)

23

45

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Placement Problem

Input:Cells and Interconnections

Output:Locations on a 2-D plane

46

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Predictors based on the given algorithm or based onabsolute truth

A bad placementA good placement

24

47

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Placement Algorithms (both incremental, so do

we understand incremental?)

l Constructive Placemente.g. Analytical optimizationBuild up placement from scratch.Fast, but most of them lack of solution quality.

l Iterative Improvemente.g. Simulated annealingGradually improve existing solutions.Slow, but with good solution quality, easy toconsider multi-objectives.

48

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Placement

Caused by:

l Adding cells/nets

l Meeting timing constraintswhile minimizing total chip area.

l Reducing congestionA post processing algorithm is more effective

25

49

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Proximity In Placement

Topological Proximity Geometrical Proximity

A

B

C

BA

Cell A and B are topological proximate Cell A and B are geometrical proximateIn a good placement, topological proximity and geometrical proximity generally correlate well

C D

50

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Preliminary studies

Reducing length of long net

Helps meeting timing constraints

26

51

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Shrinking long nets

Pick up cells which are located on the edge of theBounding box, switch them with other cells.

52

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Experimental setup

Given a placement produced by a fresh placement run,

find the longest nets,Let its length be L, then find all the nets which arelonger than (1-x%)*L, reduce all these long nets to (1-x%)*L or less.Data reported are number of long nets and x%

27

53

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Experiment results

It is not difficult to reduce the longest net by up to 20% without considerably increasing the total wirelength

N/A131.85%100.14%20.06%125114avql

1.67%60.78%50.16%30.03%121854avqs

N/A50.06%10.04%10.01%115059industry3

N/A39N/A170.48%60.06%212142industry2

N/A132.48%70.37%30.06%12907Primary2

�WL# nets�WL# nets�WL# nets�WL# nets

30%20%10%5 %

Length of long nets are reduced to (1-x)% of longest net

#cellsCircuit

54

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Experiment results

Experiments of Shrinking Long Nets

0.00%

0.50%

1.00%

1.50%

2.00%

2.50%

3.00%

3.50%

0% 5% 10% 15% 20%

Long nets length reduction

To

tal

WL

in

cre

as

e

primary2 industry2 industry3 avqs avql

It’s a greedy, simple approach.More effective algorithms are to be studied.

28

55

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

PART III:Incremental Routing

56

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Routing Problem

l Basic operations:l Net removal (easy)

l Net addition (focus of this talk)

l Objectivel Preserve as much previous results as possible

l Applicationsl Changes of the netlist due to functionality modification

l Modification of placement due to performance consideration

l Change of a route (width, spacing, possible buffer insertion)after post-layout extraction and simulation

l ...

29

57

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Sub-problems in IncrementalRouting

l Single Net Routing

l No change of existing routes

l Rip-up and Reroute

l Change some existing routes but no change of

placement/floorplan

l Incremental Floorplan and Placement Update

l Invoke when Rip-up & re-route fails

Increasing flexibility

Increasing runtime

58

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Single Net Routing Problem

l Problem FormulationGiven

- a set of existing routes/obstacles

- the width and spacing of a given net

Find a valid route for this net

l Need for gridless (GLS) routingl Variable wire width for delay optimization

l Variable wire spacing for noise constraints

l Challengesl Construction of GLS routing graph

l Representation of GLS routing graph

l Search on GLS routing graph

30

59

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Existing Approaches to GLS SNR

l Obstacle expansionl Reduced to zero-width routing

l Construction of routing graphl Tile-based approach

l Point-based approachl Uniform grid graph

l Non-uniform grid graph

l Search on the routing graph

60

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Obstacle Expansion for Gridless Routing

l Expansion of Obstacles According to Width Rule andSpacing Rule to Reduce the Routing Problem intoZero Width Routing

Ww=4 Sw=2Ew=Ww/2+Sw

31

61

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Tile-based Approach[Margarino CAD’87; Sechen ISPD’98]

1. Tile Partition

2. Corner-Stitching Data Structure [Ousterhout, TCAD84]

3. A Tile-to-Tile Searching

s

t

T1

T2 T4

T3

T7

T5

T6

62

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Uniform Grid Approach [Lee’61]

Representing the region using uniform grids

Graph can be very large!

32

63

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Minimal Routing Graph Approach[Wu T-Comp’87; Zheng TCAD’96]

Extend the boundries of expanded rectangles

Sparser than uniform- grid but irregular

64

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Limitations of Existing Approachesfor Incremental Routing

l Pre-expansion of obstacles - very costlyl Need to re-expand if routing a net of different width and/or spacing

l Pre-construction of routing graph - very costlyl Incremental route may use only a small portion of the graph (but

cannot be determined a priori)

33

65

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

S

T

Non-Uniform Grid Graph Approach[Cong-Fang-Khoo, ICCAD’99]

l Given a set of obstacles and a source s and sink tl GS is an orthogonal grid graph

66

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

S

T

Implicit Representation of GS

y1

y2

y3

y4

y5

y6

y7

y8

x1 x3 x4x2 x5 x6 x7 x8 x9 x10

l Store x, y locations in arraysl Space Pre-construction

l Can be maintained incrementally (O(logN) time)

)(NO )log( NNO

34

67

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

S

T

Searching on GS

y1

y2

y3

y4

y5

y6

y7

y8

x1 x3 x4x2 x5 x6 x7 x8 x9 x10

l Determine the Location of Neighboring Node: Easy

Free?

l Determine If the Location Is Valid: Difficult

68

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

No Pre-Expansion of Obstacles

l Point enclosure query -> rectangle intersection query

a

b

dc

q

a

b

dc

q

35

69

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Our 2-D Query Data Structure

a

b

dc

q

Data Structure

Is q in free space

Query

c b,d

70

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Experimental Results

70117225668232741.5 x 3192.8Eco-7

37224013959232926.1 x 1676.8Eco-6

1969477542231556.6 x 1676.8Eco-5

2326336417231372.5 x 1608.6Eco-4

2320046417531372.5 x 1593.3Eco-3

2324536417231372.5 x 1593.3Eco-2

2323096417331372.5 x 1593.3Eco-1

RectsCellsPinsLayersDimensionEx.

l Examples

l Comparisonl Explicit Uniform Grid Graph

l Iroute from Magic Layout Editor [Arnold 88]

36

71

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Comparison of Memory UsageUnit: MB

3.01.0014.3Ratio

84.743.6641.1Eco-7

52.615.9359.4Eco-6

35.212.7191.0Eco-5

32.610.9161.7Eco-4

32.67.2160.2Eco-3

32.710.9160.2Eco-2

32.710.9160.2Eco-1

IrouteNon-Uni. GridUni. GridsEx.

72

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Comparison ofRuntime and ECO Routing Quality

47591 / 2079.7935690 / 7438.2Eco-7

56423 / 2074.2937858 / 7024.7Eco-6

34543 / 1243.1427061 / 5412.3Eco-5

24385 / 15357.3921760 / 12224.0Eco-4

36593 / 10368.7034736 / 11034.5Eco-3

13162 / 6226.5811332 / 446.3Eco-2

22629 / 8942.1517374 / 6619.1Eco-1

Iroute

Result(wl/vc)Time(s)

Non-Uniform

Result(wl/vc)Time(s)

Ex.

37

73

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Sub-problems in IncrementalRouting

l Single Net Routing

l No change of existing routes

l Rip-up and Reroute

l Change some existing routes but no change of

placement/floorplan

l Incremental Floorplan and Placement Update

l Invoke when Rip-up & re-route fails

Increasing flexibility

Increasing runtime

74

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Rip-up and Reroute

l Invoke when net-by-net routing fails

l Objectivesl Complete all the nets

l Minimize number of rip-ups

l Challengesl Gridless nature of routing problem makes routing resource

difficult to model and manage

38

75

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Overview of Existing R&R Approaches

l Maintain Design-Rule Correctnessl Pros: always has a valid partial solution

l Cons: too restrictive; net order dependent

l Allow Temporary Design-Rule Violationl Pros: allow cross and touch [Kawamura ICCAD’90], more flexibility

l Cons:l Difficult to model routing resources in gridless routing

l Lack of global picture

l Planning Based Ripup and Re-routel Preferred approach

l Can model routing resources with flexible resolution

l Allow a more global view

l Can be applied iteratively

76

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Planning Graph Construction

39

77

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Planning Based Rip-up & Re-route

l Advantagesl More Global Picture

l Reasonably Accurate Routing Region Information

l Very efficient on the planning graph

l Similar to global routing, but only serve as a guide

l Re-planning Approachesl Local Refinement

l Ripup and Re-plan

78

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Local Refinement

S1

S2

S3

T1

T2T3

40

79

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Local Refinement

S1

S2

S3

T1

T2T3

80

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Local Refinement

S1

S2

S3

T1

T2T3

41

81

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Ripup & Re-planning

S1

S2

S3

T1

T2T3

82

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Sub-problems in IncrementalRouting

l Single Net Routing

l No change of existing routes

l Rip-up and Reroute

l Change some existing routes but no change of

placement/floorplan

l Incremental Floorplan and Placement Update

l Invoke when Rip-up & re-route fails

Increasing flexibility

Increasing runtime

42

83

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Incremental Floorplan andPlacement Updatel Problem Formulation

l When rip-up and re-route fails, determine how to complete thedesign with minimum changes of the floorplan and/or placement

l Need efficient heuristics to decide what routing regionsto enlarge

l May apply compaction/de-compaction techniques

l Earlier discussions on incremental floorplan andplacement also apply

84

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Conclusions

l Design iteration is unavoidable

l Incremental PD is an important enabling technology to

allow design convergence

l Trade-off of efficiency, preservability, and design quality

is the key issue in incremental designs

l Much work is needed in this area

l Need to start thinking about “Design for Incremental

Changes” (DFI)

43

85

Jason CongMajid SarrafzadehISPD 2000: April 10-12 C

Acknowledgements

l NSF, DARPA, and Fujitsu Laboratories

l Jie Fang and Kei-Yong Khoo from UCLA

l Kiarash Barzagan, Maogang Wang, and Xiaojian Yang

from Northwestern Univ.