Graph Models for Global Routing: Grid Graph

Each cell is represented by a vertex.

Two vertices are joined by an edge if the corresponding cells are adjacent to each other. The occupied cells are represented as lled circles, whereas the others are as clear circles.

w w

. w
c

u t jn
d

r o w
a

. ld
b

m o c
d c

Graph Model: Channel Intersection Graph

Channels are represented as edges. Channel intersections are represented as vertices. Edge weight represents channel capacity.

Extended channel intersection graph: terminals are also represented as vertices.

channel intersection graph

w w

extended channel intersection graph

. w

u t jn

r o w

. ld

m o c

Global-Routing Problem
Given a netlist N={N1 , N2 , . . . , Nn}, a routing graph G = (V, E ), nd a Steiner tree Ti for each net Ni, 1 i n, such that U (ej ) c(ej ), ej E n and i=1 L(Ti ) is minimized, where c(ej ): capacity of edge ej ; xij = 1 if ej is in Ti; xij = 0 otherwise;

n U (ej ) = i=1 xij : # of wires that pass through the channel corresponding to edge ej ;

L(Ti): total wirelength of Steiner tree Ti. For high-performance, the maximum wirelength (maxn i=1 L(Ti )) is minimized (or the longest path between two points in Ti is minimized).

w w

. w

u t jn

r o w

. ld

m o c

Global Routing in dierent Design Styles

global routing

full custom flexible channels most general problem

standard cell flexible channels

fixed feedthroughs

w w

. w

u t jn

r o w

gate array

. ld

m o c

FPGA

fixed channels

fixed routing tracks switchbox constraints

Global Routing in Standard Cell

Objective

Minimize total channel height. Assignment of feedthrough: Placement? Global routing? For high performance, Minimize the maximum wire length. Minimize the maximum path length.

w w

. w

u t jn

r o w
failed net

. ld

m o c

feedthroughs

Global Routing in Gate Array

Objective Guarantee 100% routability. For high performance, Minimize the maximum wire length. Minimize the maximum path length.

w w

. w

u t jn

r o w

. ld

m o c

2 tracks

failed connection Each channel has a capacity of 2 tracks.

Global Routing in FPGA

Objective Guarantee 100% routability. Consider switch-module architectural constraints. For performance-driven routing, Minimize # of switches used. Minimize the maximum wire length. Minimize the maximum path length.

w w

. w

u t jn

r o w

. ld

m o c

s s

?
s

failed connection Each channel has a capacity of 2 tracks.

w w

. w

u t jn

r o w

switch module

. ld

m o c

Classication of Global-Routing Algorithm

Sequential approach: Assigns priority to nets; routes one net at a time based on its priority (net ordering?). Concurrent approach: All nets are considered at the same time (complexity?)
globalrouting algorithm

sequential approach

twoterminal

linesearch

w w

. w

u t jn
Soukup

r o w

. ld

m o c

concurrent approach

multiterminal

hierarchical integer programming

maze

Steinertree based

Lee

Hadlock

Global-Routing: Maze Routing

Routing channels may be modelled by a weighted undirected graph called channel connectivity graph. Node channel; edge two adjacent channels; capacity: (width, length)
1 3,2 1 4 8 2 5 A B 3 A 6 9 7 10 B 4 1,2 5 1,2 2 3,2 3

2 2,1

0,0 5

w w
2,1 6 7 0,1 0,4 2,1 9 2,5 10

n j . w
1 3,2 4 1,2 3,2 8

3,2 8

w tu
2 2,1 5 0,0 6 0,1 3 2,1 2,1 9

or
3,2 1,2 3,2 9 7 0,4 2,5 10

1,6

. ld
7 10 1 4 8

m o c
0,1 5 0,2 6 1,0 5 1,2 6

7 0,5

route AA via 567

7 1,4

3,6

route BB via 567

2 5 A B

3 A 6 9 7 10 B

route BB via 52369107

updated channel graph

maze routing for nets A and B

9

Global Routing by Integer Programming

i i Suppose that for each net i, there are ni possible trees ti 1 , t2 , . . . , tni to route the net. Constraint I: For each net i, only one tree ti j will be selected. Constraint II: The capacity of each cell boundary ci is not exceeded. Minimize the total tree cost. Question: Feasible for practical problem sizes?

Key: Hierarchical approach! an routing instance

2 1 1 1, 2 3

grid graph
C2 = 2

C2

C4

w w
C1 C4 C3

. w
2
C2 C1 C3

u t jn
C1 = 2

r o w
C3 = 2
C2

. ld
C2

m o c
2 1 2 3
C2 C1

a feasible routing
1

1, 3 3

C4 = 2

2, 3

1, 2

C1 C3 C3

C4 C1

C4

C4 C3

trees of net 1

trees of net 2

trees of net 3
10

An Integer-Programming Example
Boundary B1 B2 B3 B4 t1 1 0 1 0 1 t1 2 1 0 1 1 t1 3 1 1 1 0 t2 1 1 0 1 0 t2 2 0 1 1 1 t2 3 1 1 0 1 t3 1 1 1 0 0

gi,j : cost of tree ti j g1,1 = 2, g1,2 = 3, g1,3 = 3, g2,1 = 2, g2,2 = 3, g2,3 = 3, g3,1 = 2, g3,2 = 2. Minimize 2x1,1 + 3x1,2 + 3x1,3 + 2x2,1 + 3x2,2 + 3x2,3 + 2x3,1 + 2x3,2 subject to x1,1 + x1,2 + x1,3 x2,1 + x2,2 + x2,3 x3,1 + x3,2 x1,2 + x1,3 + x2,1 + x2,3 + x3,1 x1,1 + x1,3 + x2,2 + x2,3 + x3,1 x1,2 + x1,3 + x2,1 + x2,2 + x3,2 x1,1 + x1,2 + x2,2 + x2,3 + x3,2 xi,j

w w

. w

u t jn

r o w
= = = =

. ld

m o c
I : t1 ) I : t2 ) I : t3 ) II : B 1) II : B 2) II : B 3) II : B 4)

t3 2 0 0 1 1

1 (Constraint 1 (Constraint 1 (Constraint 2 (Constraint 2 (Constraint 2 (Constraint 2 (Constraint 0, 1, 1 i, j 3

11

Hierarchical Global Routing

Marek-Sadowska, Router planner for custom chip design, ICCAD, 1986. At each level of the hierarchy, an attempt is made to minimize the cost of nets crossing cut lines. At the lowest level of the hierarchy, the layout surface is divided into R R grid regions with boundary capacity equal to C tracks. Let Rl be the # of grid regions of a given cut line l; a cut line can be divided into l M =R sections. C Global routing can be formulated as a linear assignment problem: xi,j = 1 if net i is assigned to section j ; xi,j = 0 otherwise. Each net crosses the cut line exactly once: Capacity constraint of each section:

wij : cost of assigning net i to section j . Minimize

w w
Cut line

j . w
Section

u t n

r o w
N x i=1 ij

. ld

m o c

M x j =1 ij

= 1, 1 i N .

C, 1 j M .
M w x . j =1 ij ij

N i=1 Upper leftmost pin of net

Bounding box

Ideal sections to route this net

Lower rightmost pin of net

12

The Routing-Tree Problem

Problem: Given a set of pins of a net, interconnect the pins by a routing tree.

gate array

Minimum Rectilinear Steiner Tree (MRST) Problem: Given n points in the plane, nd a minimum-length tree of rectilinear edges which connects the points. M RST (P ) = M ST (P S ), where P and S are the sets of original points and Steiner points, respectively.

w w

. w

u t jn

standard cell

r o w

. ld

m o c
building block

13

Steiner points

minimum spanning tree MST

w w

. w

u t jn

r o w

. ld

MRST

m o c

Theoretic Results for the MRST Problem

Hanans Thm: There exists an MRST with all Steiner points (set S ) chosen from the intersection points of horizontal and vertical lines drawn points of P . Hanan, On Steiners problem with rectilinear distance, SIAM J. Applied Math., 1966. Hwangs Theorem: For any point set P ,
Cost(M ST (P )) Cost(M RST (P ))

Hwang, On Steiner minimal tree with rectilinear distance, SIAM J. Applied Math., 1976. Best existing approximation algorithm: Performance bound

Foessmeier et al, Fast approximation algorithm for the rectilinear Steiner problem, Wilhelm Schickard-Institut f ur Informatik, TR WSI-93-14, 93. Zelikovsky, An 11 approximation algorithm for the network Steiner problem, Al6 gorithmica., 1993.

w w

n j . w

w tu
MST

or

. ld

3 . 2

m o c
61 48

by Foessmeier et al.

MRST

Hanan grid

Cost(MST)/Cost(MRST) > 3/2

14

A Simple Performance Bound

Easy to show that
Cost(M ST (P )) Cost(M RST (P ))

2.

Given any MRST T on point set P with Steiner point set S , construct a spanning tree T on P as follows: 1. Select any point in T as a root. 2. Perform a depth-rst traversal on the rooted tree T . 3. Construct T based on the traversal.
2 7 1 6 3 4 T 8 5 2

w w

. w
6 1 3

u t jn
7 8 4 5

r o w
2 1 6 3

. ld
8 4

m o c
5 1 3

T 5 4

depthfirst traversal every edge is visited twice

Cost(T) <= 2 Cost(T)

15