Transcript src_oct99_int

Interconnect Planning, Synthesis, and
Layout for Performance, Signal
Reliability and Cost Optimization
SRC Task ID: 605.001
PI: Prof. Jason Cong (UCLA)
Students: Lei He, David Pan, Xin Yuan
Mentors: Dr. Prakash Arunachalam (Intel)
Dr. Norman Chang (HP)
Dr. Wilm Donath (IBM)
Dr. Stefan Rusu (Intel)
Project Overview

Objective: investigate an
interconnect-centric design flow and
methodology, consisting of:
Interconnect
Planning
Interconnect
Synthesis
Interconnect
Layout
Key Issues in Interconnect Planning

Three levels of planning:
 Interconnect
architecture planning (pre-design)
 Interconnect
planning with RTL-floorplan
 Interconnect
planning with physical-level
floorplan

Enabling tools: interconnect estimation
models for interconnect synthesis/layout
Review: Accomplishments in Year 1

Efficient (constant time) and accurate (90%) interconnect
delay estimation models for 2-pin nets under different
interconnect optimization algorithms [Cong-Pan, IWLS’98,
SRC/TECHCON’98, ASPDAC’99]


Optimal wire sizing (OWS)

Simultaneous driver and wire sizing (SDWS)

Simultaneous buffer insertion/sizing and wire sizing (BISWS)
Interconnect architecture planning [Cong-Pan,DAC’99]

Propose a unified wire-width planning framework

Obtain a surprising result that our pre-determined two-width can
achieve close to optimal solution for a large wire length range !

Can handle different objective functions
Accomplishments in Year 2



Efficient and accurate interconnect estimation
models for multiple-pin nets [Cong-Pan,
TAU’99]
Buffer block planning for interconnect-driven
floorplanning [Cong-Kong-Pan, ICCAD’99]
Further study on interconnect architecture
planning
Interconnect Estimation for Multiple-Pin Nets
G
G0
S1
Cs3
Cs1
S3
Sn
Csn
Input
Cs2


S2
Objective: estimate delay/area under different
interconnect optimizations (e.g, OWS, BISWS) quickly
(100K - 1M nets per second)
Different targets:
1. Minimize the delay to a single critical sink (SCS)
2. Minimize the maximum delay (defined as the tree delay) for
multiple critical sinks (MCS)
Estimation for Multiple-Pin Nets

Very difficult:
 No
closed-form wire shaping or buffer insertion
 All available optimization algorithms are
iterative based
 Multiple critical sinks may exist at the same
time !

Our approach
 Reduce
the multiple-pin net estimation problem
into one or several 2-pin net estimation
problems, then use our previous (Year 1) results
Reduction for OWS of SCS
G
S1
Cs1
S3
G0
Sk
Csk
Input
Cs2
G0
G
Sk
C1
Input
S2
C2
Csk
Single-Line-Multiple-Load (SLML)
OWS for SCS

Transform SLML to SLSL (i.e., 2-pin net)
Rd
l1
lk
l2
Sk
W
C1
C2
Ck-1
Ck
OWS for SCS

Transform SLML to SLSL (i.e., 2-pin net)
Rd
lk
l2
l1
Sk
W
C0
C1
C2
k
i1li
j 1
l
CL  
CL
Ck-1
j
 Cj
k
C 0   Cj  CL
j 1
Delay/Area Estimation for OWS/SCS

Closed-form delay estimation for the critical sink
  1l
2 1l
Tows ( Rd , l , CL)  RdC 0   2

 Rdcf
W ( 2l ) W ( 2l )
where

1
1
4
rca , 
1
2 
2


Rdrcacfl   l

rca
RdCL
W(x) is Lambert’s W function defined as wew  x

Closed-form area estimation for the critical path
Aows ( Rd , l , CL) 
r (cf l  2CL)
l
2 Rd ca
Summary for Interconnect Estimation


Develop delay and area estimation models for
multiple-pin nets with consideration of various
interconnect optimizations
Consider different optimization objectives
 Single
critical sink (SCS)
 Multiple critical sinks (MCS)

Apply various optimization alternatives:
 Optimal
wire sizing (OWS)
 Buffer insertion/sizing and wire sizing (BISWS)
Model 5mm
TRIO 5mm
Model 10mm
TRIO 10mm
Model 20mm
TRIO 20mm
Model 5mm
TRIO 10mm
2
2.5
1.5
2
W_avg (um))
Delay (ns)
Delay/Area Comparison with TRIO
1
0.5
0
0
5000
10000
15000
20000
L_1 (um)
TRIO 5mm
Model 20mm
Model 10mm
TRIO 20mm
1.5
1
0.5
0
0
5000
10000
L_1 (um)
15000

Rd = 180ohm, C1 = 100 fF, C2 = 10 fF

One internal load, l1 = 0.1 to 0.9 x l (l = 5, 10 or 20 mm)
Max. allowable wire width is 20x min. width; wire is segmented in
every 10um.

20000
Accomplishments in Year 2



Efficient and accurate interconnect estimation
models for multiple-pin nets (Cong-Pan,
TAU’99)
Buffer block planning for interconnect-driven
floorplanning (Cong-Kong-Pan, ICCAD’99)
Further study on interconnect architecture
planning
Motivation for BBP



For high-performance DSM designs, many buffers
may be inserted to optimize/meet interconnect delay
(e.g., up to 800,000 for 50nm tech., [Cong’97, SRC
Work Paper])
The introduction of so many buffers will
significantly change a floorplan; thus shall be
planned to ensure timing/design convergence.
Need proper buffer block planning (BBP) to address



buffer location constraints (e.g., hard IP blocks)
“dead area” utilization
regularity for ease of layout and power/ground network
sharing
Buffer Block Planning Problem
buffer blocks
white space
grey (soft) block
(limited buffers)
black (hard) block
(no buffer allowed)


Given: initial floorplan, buffer capacity for each soft
block, and performance constraint for each net;
Output: “optimal” location/dimension of buffer
blocks such that the overall chip area and the number
of buffer blocks are minimized.
Feasible Regions for BI

Feasible region is the maximal region that a buffer
can be placed to meet given delay constraint.
1 buffer
driver
CL
driver
k buffers
CL
Feasible Regions for BI


We obtain the closed-form formula for FR’s
Important observation: even under tight delay
constraint, FR for BI can still be pretty large!
=> FR provides a lot flexibility to plan buffer location
6000um
7000um
8000um
9000um
10000
8000
um

6000
4000
2000
0
0
0.2 delta 0.4
0.6
• FR distance
under different
delay budgets
• Delay budget is
(1+delta) Topt (the
best delay by
optimal buffer
insertion)
Feasible Regions for BI

FR extended to 2-dimension with obstacles
sink
source
2-D FR
Restricted (RES) line (delay minimal BI positions)
Overview of BBP Algorithm
1. Build polar graphs for given floorplan;
2. Build tile data structure;
3. For each tile, compute its area slacks;
4. Compute FR(s) for each net;
4. While (there exists some buffer to be inserted) {
Pick_A_Tile  that can insert most buffers w/o area
penalty; if no such tile exists, pick the one with most BI
demand ;
Insert_Buffers into : insert all those buffers whose FR’s
intersect with  to create BB w/o area penalty; or insert
one buffer into  to expand its channel;
Update chip dimension, FR, and area slacks, etc.
}
Experimental Setting

Two Scenarios (for buffer insertion flexibility):
 RES:
restricted buffer insertion position(s) as to
minimize delay
 FR: feasible buffer region as to meet delay
constraint

Two Algorithms (for buffer clustering):
 RDM:

a buffer is randomly assigned to any feasible
location
 BBP: buffers are assigned with appropriate
clustering
6 MCNC + 5 randomly generated circuits (0.18um tech)
#nets that meet delay constraints
1200
1000
800
600
400
200
0
RDM/RES
RDM/FR
BBP/RES
BBP/FR
FR provides a lot more flexibility than RES (e.g.,
to avoid obstacles) during BI, thus can better
meet delay constraints
Area Increase (%) due to BI
5.00
4.00
3.00
2.00
1.00
0.00
RDM/RES
RDM/FR
BBP/RES
BBP/FR
BBP/FR can effectively cluster individual buffers
together with marginal area increase (less than 2% in all
above test cases), by high utilization of “dead areas”.
Comparison of #BB
600
500
400
300
200
100
0
RDM/RES
RDM/FR
BBP/RES
BBP/FR
BBP reduces #BB from RDM by a factor of up to 3x;
BBP/FR further reduces #BB from BBP/RES by up to 34%
Accomplishments in Year 2



Efficient and accurate interconnect estimation
models for multiple-pin nets (Cong-Pan,
TAU’99)
Buffer block planning for interconnect-driven
floorplanning (Cong-Kong-Pan, ICCAD’99)
Further study on interconnect architecture
planning
 Our
two width-planning is still valid for certain
range (2x) of driver size variation
 Currently investigating wider range of variations
Deliverables






Development of efficient and accurate interconnect performance
estimation models for interconnect-driven synthesis and planning
(Completed - 30-Jun-1999)
Development of interconnect architecture planning framework
(Completed - 30-Jun-1999)
Development of efficient algorithms for integrated interconnect
planning & floorplanning capabilities at the physical level
(Completed - 30-Sep-1999)
Development & validation of accurate noise models to guide the
interconnect synthesis algorithm for signal reliability (Planned 31-Dec-1999)
Development of optimal or near-optimal interconnect synthesis
algorithm for multiple spatially or temporally related signal nets
for performance & signal reliability optimization (Planned - 31Dec-1999)
Development of efficient algorithms for integrated interconnect
planning & floorplanning capabilities at the RTL-level; Software
(Planned - 31-Dec-2000)
Technology Transfer

TRIO (TRIO-Repeater-Interconnect-Optimization)
package


Integrated into Intel design technology
Available on the web:


IDEM (Interconnect Delay Estimation Model) package


Prototype provided to Intel
Package will be available this week to all SRC member
companies:


http://cadlab.cs.ucla.edu/~trio
http://cadlab.cs.ucla.edu/~trio
BBP (Buffer Block Planning) for physical level
floorplanning

Interest from Intel and HP
Summary and Future Work




Efficient and accurate interconnect estimation
models
Interconnect architecture planning
Buffer Block Planning
Future Work:
 Noise
estimation and planning
 RTL interconnect planning
Milestones








Development of a computational model for interconnect architecture
planning based on a given design characterization (specified in terms of
target clock rate, interconnect distribution, depths of logic,network, etc.)
(31-Dec-1998)
Development of estimation models for interconnect layout optimizations
suitable for pre-layout synthesis and planning (31-Dec-1998)
Development of efficient algorithms for integrated interconnect planning
and floorplanning capabilities at the RTL-level (31-Dec-1999)
Completion of the ongoing effort on the development on a multi-layer
general-area gridless routing system (31-Dec-1999)
Development of optimal or near-optimal interconnect synthesis algorithm
for multiple spatially or temporally related signal nets for performance
and signal reliability optimization (31-Dec-1999)
Development and validation of very efficient but accurate noise models to
relate the noise with the physical parameters to guide the interconnect
synthesis algorithm for signal reliability optimization (31-Dec-1999)
Development of efficient algorithms for integrated interconnect planning
and floorplanning capabilities at the physical level (31-Dec-1999)
Development of efficient algorithms for integrated interconnect planning
and floorplanning capabilities at the RT-level (31-Dec-2000)