Advanced Synthesis ECOs with Gate
Array Fillers
Masters of Science Thesis in Embedded Electronic System Design

CHIRAYU SHAH

Chalmers University of Technology
Departement of Conmputer Science and Engineering
Gothenburg, Sweden August 2014
Master’s Thesis 2014:1



The Author grants to Chalmers University of Technology and University of Gothenburg
the non-exclusive right to publish the Work electronically and in a non-commercial pur-
pose make it accessible on the Internet. The Author warrants that he/she is the author
to the Work, and warrants that the Work does not contain text, pictures or other mate-
rial that violates copyright law.

The Author shall, when transferring the rights of the Work to a third party (for ex-
ample a publisher or a company), acknowledge the third party about this agreement.
If the Author has signed a copyright agreement with a third party regarding the Work,
the Author warrants hereby that he/she has obtained any necessary permission from
this third party to let Chalmers University of Technology and University of Gothenburg
store the Work electronically and make it accessible on the Internet.

Advanced Synthesis ECOs with Gate Array Fillers

Chirayu Shah

c©Chirayu Shah, August 2014

Examiner:Professor Per Larsson-Edefors

VLSI Research Group
Chalmers University of Technology
Department of Computer Science and Engineering
SE-412 96 Göteborg
Sweden
Telephone:+46 (0)31-772 1700

Department of Computer Science and Engineering
Göteborg, Sweden August 2014.

i



Abstract

Engineering Change Orders (ECOs) are commonly used in the Application Specific In-
tegrated Circuits (ASIC) industry to either fix design bugs or to add new features to the
design after first tape out. Due to rapid increase in the design complexity the metal-
mask ECOs have become inevitable. Generally, redundant standard cells, known as
spare cells, are used to realize such type of ECOs. However, these cells suffer from a
major drawback of having predefined functionality and location [1]. As a result, their
use becomes limited. To overcome this inflexibility, gate array type spare cells are used.

As the gate array spare cell is configurable, it opens up new possibilities of doing big
ECOs. On the other hand, most ECO algorithms offered by the commercial tools are
not smart enough to handle such type of ECOs. In order to overcome this limitation,
designers prefer using conventional ASIC flow rather than realizing such big changes as
an ECO.

In this thesis, a methodology to implement large scale ECOs is presented. This method-
ology aims to overcome the existing limitations of using the ECO algorithms by incor-
porating conventional ASIC flow algorithms to perform an ECO. The methodology has
been implemented using gate array type cells. Simulation results show that for a medium
sized design (12k gates) the implementation consumed 60% more dynamic power and
occupied 75% more area as compared to the using regular standard cells. However, if
a proper gate array library containing all the required cells is used for mapping, this
number would be reduced to 30% and 35% respectively. On the other hand, due to
advantages like faster time to market and small manufacturing costs [2] the area and
power overhead incurred get compensated.

Keywords: ECO, metal-configurable gate array spare cells, ASIC flow, metal-masks
ECOs, spare cells.

i



Acknowledgements

I am thankful to following people for their contribution:

• My supervisor at Nordic Semiconductor ASA, Dr. Johnny Pihl (Senior R&D
Engineer). Without him this master thesis project wouldn’t be possible. He helped
me regarding the practical arrangements as well as technical matters. He is the
”producer”.

• My second supervisor at Nordic Semiconductor ASA, Dr. Oyvind Granheim (Se-
nior R&D Engineer). He is ”the” guy who has experience with ECOs. He always
showed me the right direction whenever needed and supported me through out the
project. He is the ”director”.

• My examiner at Chalmers University of Technology, Professor Per Larsson-Edefors
for his constant support throughout the thesis.

• The System Integration group at Nordic Semiconductor for their technical support
and friendly behavior towards me.

• My parents for providing the opportunity, support and help through my Master’s
Studies.

ii



Contents

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1 ECO in the context of ASIC design . . . . . . . . . . . . . . . . . . . . . . 6
2.2 IC Compiler ECO flows comparison . . . . . . . . . . . . . . . . . . . . . 7
2.3 Pre-mask ECO flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Post-mask ECO using conventional spare cells . . . . . . . . . . . . . . . . 15
2.5 Metal-configurable Gate Array spare cells . . . . . . . . . . . . . . . . . . 19
2.6 Advantages of using Gate Array cells as spare cells . . . . . . . . . . . . . 24
2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3 Gate Array spare cell methodology . . . . . . . . . . . . . . . . . . . . . 26
3.1 Adding Gate Array fillers as spare cells . . . . . . . . . . . . . . . . . . . 26
3.2 Post-mask ECO using Gate Array fillers . . . . . . . . . . . . . . . . . . . 31

3.2.2 Gate Array ECO process . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

ii



CONTENTS

4 Advanced ECO methodology . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1 Purpose of finding new methods for doing an ECO . . . . . . . . . . . . . 35
4.2 Overview of the Advanced ECO methodology . . . . . . . . . . . . . . . . 36

4.2.1 Overview of the Block Level Flow . . . . . . . . . . . . . . . . . . 36
4.2.2 Overview of the Top level flow . . . . . . . . . . . . . . . . . . . . 38

4.3 Example description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.4 Implementation- Block Level Flow . . . . . . . . . . . . . . . . . . . . . . 39

4.4.8 Results-Block Level Flow . . . . . . . . . . . . . . . . . . . . . . . 46
4.5 Implementation - Top level flow . . . . . . . . . . . . . . . . . . . . . . . . 46

4.5.1 Top level Design Assumptions . . . . . . . . . . . . . . . . . . . . . 46
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5 Structural ASICs vs. Advanced ECO methodology . . . . . . . . . . . 55
5.1 Relevant background on Structured ASICs . . . . . . . . . . . . . . . . . . 55
5.2 Structured ASICs and the Gate Arrays . . . . . . . . . . . . . . . . . . . . 57
5.3 Advanced ECO methodology and the structured ASICs . . . . . . . . . . 58
5.4 Applications Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6 Results and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1 Evaluation setup and procedure . . . . . . . . . . . . . . . . . . . . . . . . 61

6.1.1 Setup Information . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.2 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

6.2 Cell area and gate count comparison . . . . . . . . . . . . . . . . . . . . . 63
6.3 Power comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

7 Conclusion and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . 67
7.1 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
A IC Compiler Pre-mask ECO TCL script. . . . . . . . . . . . . . . . . . . . . . 72
B IC Compiler Post-mask ECO TCL script . . . . . . . . . . . . . . . . . . . . . 74
C Adding Gate Array fillers as spare cells . . . . . . . . . . . . . . . . . . . . . . 76

iii



List of Figures

1.1 Figure representing the ECO Problem . . . . . . . . . . . . . . . . . . . . 4

2.1 Purpose of an ECO in the context of the ASIC design flow . . . . . . . . 7
2.2 Comparison between ICC’s ECO flows . . . . . . . . . . . . . . . . . . . . 8
2.3 Complete pre-mask ECO flow . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 An ECO specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Before the ECO Placement [3] . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 After the ECO Placement [3] . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.7 The design state during ECO (top) and after ECO routing (bottom) [3] . 13
2.8 Placed and Routed design having gaps in the rows . . . . . . . . . . . . . 14
2.9 Placed and routed design after filling gaps with filler cells . . . . . . . . . 14
2.10 Formality[4] verification flow . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.11 Representation of a design having conventional Spare cells . . . . . . . . . 16
2.12 Feasibility analysis flow in post-mask ECO . . . . . . . . . . . . . . . . . 16
2.13 Spare cell feasibility check ”successful” . . . . . . . . . . . . . . . . . . . . 18
2.14 Spare cell feasibility check ”failure” . . . . . . . . . . . . . . . . . . . . . . 18
2.15 Original design having appropriate spare cells . . . . . . . . . . . . . . . 19
2.16 Design after spare cell mapping . . . . . . . . . . . . . . . . . . . . . . . . 19
2.17 Array of unconnected transistors spread to form gate array (GA) structure 20
2.18 Prefabricated Layers and Programmable layers of a GA cell . . . . . . . . 21
2.19 Using GA type base cell to realize metal-only ECO . . . . . . . . . . . . . 21
2.20 Use of base cell to make different types of cells . . . . . . . . . . . . . . . 22
2.21 Different sized GA filler cells available in the library . . . . . . . . . . . . 23
2.22 ICC’s ECO swapping process . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.1 A Floorplan in which rows are created with site_type=unit . . . . . . . 28
3.2 Concept of a row with respect to unit tile parameters . . . . . . . . . . . 28
3.3 Standard cell unit tile (”unit”) . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Output of report_unit_tiles command showing the mapped filler cells 29
3.5 Design showing GA fillers inserted . . . . . . . . . . . . . . . . . . . . . . 30

iv



LIST OF FIGURES

3.6 Final output design containing GA Fillers and marked as spare cells . . . 31
3.7 Design highlighting the different sized GA spare cells . . . . . . . . . . . 33
3.8 Design after carrying out spare cell mapping . . . . . . . . . . . . . . . . . 33
3.9 Transistor gate level view of the design before the ECO . . . . . . . . . . 34
3.10 Transistor gate level view of the design after the ECO . . . . . . . . . . . 34

4.1 Overview of the Advanced ECO methodology . . . . . . . . . . . . . . . . 36
4.2 Modified ASIC design Flow for using GA type cells . . . . . . . . . . . . . 37
4.3 Example used for explaining the Advanced ECO methodology . . . . . . . 38
4.4 Synthesized netlist containing only GA type cells . . . . . . . . . . . . . . 40
4.5 Transistor gate (red line) level view of the top level design . . . . . . . . . 41
4.6 GA cells placed on rows created using site_type = unit . . . . . . . . . 42
4.7 Tile used for creating rows with site_type=unit . . . . . . . . . . . . . . 43
4.8 Tile used for creating rows with site_type=gaunit . . . . . . . . . . . . 43
4.9 Rows created using site_type = gaunit . . . . . . . . . . . . . . . . . . 44
4.10 Figure showing GA cells placed as well as aligned . . . . . . . . . . . . . . 44
4.11 Final block level design filled with GA type cells everywhere. . . . . . . . 45
4.12 New functionality instantiated as an MACRO in the top level design . . . 46
4.13 ECO flow for top level design containing blocks implemented using GA

type cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.14 Original RTL of the top level design . . . . . . . . . . . . . . . . . . . . . 48
4.15 Modified RTL of the top level design . . . . . . . . . . . . . . . . . . . . . 48
4.16 Top level design after MACRO mapping . . . . . . . . . . . . . . . . . . . 48
4.17 Transistor level view of the unaligned transistor mapping . . . . . . . . . 49
4.18 Transistor level view of the aligned transistor mapping . . . . . . . . . . . 49
4.19 Figure highlights the gaps within the MACRO . . . . . . . . . . . . . . . 49
4.20 Top Level Netlist showing the MACRO (top cordic) . . . . . . . . . . . . 50
4.21 Top Level Netlist showing the I/Os of the MACRO connected (ECO-Netlist) 51
4.22 Final top level design after the MACRO integration . . . . . . . . . . . . 51
4.23 Modified RTL of the top level design (”existing chip”) . . . . . . . . . . . 52
4.24 The final layout netlist after the complete ECO process . . . . . . . . . . 52
4.25 Original top Level design with GA filler . . . . . . . . . . . . . . . . . . . 53
4.26 Final top level design after the MACRO integration . . . . . . . . . . . . 53
4.27 Compare GDS Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

5.1 Architecture of Structured ASICs [5] . . . . . . . . . . . . . . . . . . . . . 56

6.1 Graph showing the gate count of the design synthesized using GA cells
vs. regular standard cells . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6.2 Graph showing area occupied GA cells as compared to regular standard
cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.3 Graph comparing dynamic power consumption between GA cells and reg-
ular standard cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

v



LIST OF FIGURES

6.4 Leakage power consumption comparison between GA cells and regular
standard cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

vi



List of Tables

5.1 Comparison between different types of design styles . . . . . . . . . . . . . 57

vii



List of Abbreviations

ASIC Application Specific Integrated Circuits

CTS Clock Tree Synthesis

DRC Design Rule Checks

ECO Engineering Change Order

FPGA Field Programmable Gate Array

GA Gate Array

GDS Graphic Data System

HVT High Threshold Voltage

ICC Synopsys IC Compiler

I/O Input/Output

NMOS N-type Enhancement mode Metal Oxide Semiconductor

PMOS P-type Enhancement mode Metal Oxide Semiconductor

RTL Register Transfer Level

TCL Tool Command Language

viii



1
Introduction

ECO commonly referred to as Engineering Change Order is the process of making
modifications to the design without running the entire ASIC design flow again. These
modifications can either be in the form of a bug fix or adding a new feature. Furthermore,
these could be needed either before the chip/design has been taped out (Pre-Mask) or
after the chip has been taped out (Post-Mask).

In the pre-mask stage the ECOs are usually performed to save design time [6]. In
this stage the design is already placed and routed but not taped out, so new cells can
still be placed. Performing the entire flow again for a small change is time consuming.
Thus the need for an ECO. The ECOs can be further divided into two types in this
stage.

• Functional ECO: This type of ECOs deals with making changes in the logic
functionality of the design.

• Non-functional ECO (timing ECO): This type of ECOs deals with changes
that involve timing and signal integrity improvement.

In the post-mask stage the ECOs are usually performed to save design cost [6]. In this
stage, the chip has been taped out, which means that the transistor masks have already
been made. As the masks have been made, only the metal layers could be altered to
implement the changes to the design. This is due to the fact that the cost of making
metal-masks is much less compared to making all the masks from scratch [7]. Such pro-
cess in which only metal layers are used to perform the change is known as metal-only
ECO [1]. Such type of ECO is performed using Spare Cells [8][9]. These cells are
spread over the design before the chip is taped out.

The focus of this thesis is on carrying out functional ECOs in the post-mask stage

1



CHAPTER 1. INTRODUCTION

using gate array type spare cells. This chapter will discuss the motivation behind,
focusing on post-mask stage, using gate array cells as spare cells and finding new ECO
methods.In addition to this, the chapter will describe the problem, the objectives of this
thesis, and the outline of the method used to fulfill these objectives.

1.1 Motivation

With the increase in the design complexity, more bugs escape the pre-silicon validation
and are found post-silicon [7]. As a result post-mask ECOs are increasing. Also, to fix
these late failures in a short time and with a low mask cost [10], metal-only ECOs are
essential. In order to perform these ECOs, redundant cells, also known as spare cells
are required. When a new functionality is to be added or a failure is detected post tape
out, the inputs and outputs of these spare cells are rewired according to the intended
functionality in order to use them.

In the conventional flow, these spare cells are of type regular standard cells. As a
result they have predefined logic functionality such as AND,OR etc. Furthermore, once
added, the amount of spare cells as well as the location of these in the design become
fixed [1]. This inflexibility could result in an unfixed design. For example, under an
ECO specification (see section 2.2), if the desired gates needed to fix the bug are not
available in the spare cells, that desired functionality must be deleted or the process
of design must be done from scratch. In addition, this type of spare cells always draw
leakage current as their inputs must be tied to ground to avoid gate floating [1]. Due to
such disadvantages only limited amount of spare cells are inserted in the design.

In order to address this issue and overcome the disadvantages of conventional spare
cells, a new type of spare cells called metal-configurable gate array spare cells
[11][1] is used. The major advantage of using this type of cells as spare cells is that the
desired functionality can be made just by re-configuring the metal layers of these cells.
Furthermore, as a gate array type cell is composed of unconnected transistor gates [2],
these cells consume no leakage current. On the downside, their configurable nature calls
for increased area and power overhead [2]. But, as this overhead gets compensated by
faster time to market and smaller costs, this type of cells is used in this thesis to realize
an ECO.

Usually the amount of changes to be made as an ECO is not big [12]. This is due
to the fact that it is not possible to have all the required functionality spare cells in
the first place to implement very big changes. Secondly, it becomes impractical and
time consuming task to manually modify netlist having large number of changes. Due
to these reasons most incremental placement algorithms are not made smart enough to
figure out such big ECOs [12]. For example, the Synopsys IC Compiler(ICC) tool would
just crash if the ECO changes exceeded a certain threshold. Because programmable gate
array spare cells overcome the disadvantages of conventional spare cells, there arises a

2



CHAPTER 1. INTRODUCTION

need to devise a new ECO methodology that would work irrespective of size of an ECO.

1.2 Problem Description

Using metal-configurable gate array cells as spare cells opens up whole new possibilities
like adding new features to the design as well as fixing bugs after the chip has been taped
out. Moreover, using only metal layers reduces the mask cost and as a result also the de-
sign time drops. In a response to this new technology, Nordic Semiconductors proposed
this thesis topic in order to explore the possibilities of metal-configurable gate array cells.

The problem can be understood if we consider Figure 1.1. The left side of the Fig-
ure shows the original taped out chip that contains a spare area. The right side of the
Figure shows the synthesized netlist representing the new functionality to be added.
The problem can be described as follows. Given a new functionality in the form of
synthesized netlist and a taped out chip having a spare area filled with gate array type
cells, investigate new methods to add this new functionality onto the spare area as an
ECO. More importantly, the method should be such that the ICC tool should be able to
carryout a post-mask ECO using gate array type cells irrespective of the functionality
size.

It is assumed that there exists enough space on the taped out chip to accommodate
the new functionality as an ECO. This space can either be created by swapping (see
section 4.5.1) or having a dedicated spare area (as in Figure 1.1) inside the chip.

3



CHAPTER 1. INTRODUCTION

MACRO 1

RAM4 RAM2

RAM1RAM3

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

RAM5 RAM6

Pad

Processor

Pad

Pad

Pad

Pad

BUS

CACHE MMU

STANDARD CELLS

SP
A

R
E A

R
E

A

STANDARD CELLS NETLIST FOR THE NEW 
FUNCTIONALITY(>20k gates)

Figure 1.1: Figure representing the ECO Problem

1.3 Objectives

The following are the objectives of this master thesis.

1. Based on ICC’s spare cell methodology [13], develop a methodology to insert gate
array type fillers inside the chip and carry out a small ECO in the post tape out
phase.

2. Using the above methodology as base, propose and explain a new ECO methodol-
ogy that works on big ECOs and uses gate array type cells.

3. Evaluate the use of gate array type cells in the proposed methodology and compare
it against regular standard cells in terms of:

(a) Area consumption and gate count.

(b) Power consumption (Dynamic and Leakage).

(c) Timing requirements.

1.4 Thesis Outline

The rest of the thesis is organized as follows:

Chapter-2 Gives the background information of the important thesis terms. It in-
cludes explanation of conventional ICC ECO flows, basic concepts of gate array type

4



CHAPTER 1. INTRODUCTION

cells and using them as spare cells and advantages of using gate array spare cells over
the conventional spare cells.

Chapter-3 Presents methodology to insert gate array filler cells and perform a sim-
ple post-mask ECO using them.

Chapter-4 Proposes the new ECO methodology named Advanced Synthesis ECO method-
ology and explains the usage of gate array type cells to implement the new functionality
and the usage of gate array type spare cells to integrate the functionality as a post-mask
ECO.

Chapter-5 Introduces the Structured ASIC design methodology and compares it’s im-
plementation flow with the newly proposed methodology.

Chapter-6 Presents the evaluation results by using gate array type cells in the method-
ology and compares them against the regular standard cells.

Chapter-7 Provides conclusions of the work presented and gives suggestions for the
future work.

5



2
Background

This chapter will provide background information of the keywords present in the title
of this project. This includes keywords like ”ECO”,”gate array (GA)”,”fillers”. To start
with, section 2.1 will describe the need for an ECO at several ASIC design stages. In
section 2.2, the ICC’s ECO flows are compared. section 2.3 explains the pre-mask ECO
flow in detail. The subsequent section discusses the post-mask stage using conventional
spare cells. Section 2.5 introduces the GA cells and their configuration process and it
is followed by a section that points out the advantages of using GA filler cells as spare
cells. At the end of the chapter, the summary is presented.

The flows and the corresponding commands mentioned in this thesis are based on the
Synopsys IC Compiler. Only the basic information regarding the commands is men-
tioned here. For detailed descriptions of these commands, the reader is encouraged to
refer to the tool user manual [14]. In case of different tool, the concepts should remain
the same but the corresponding commands would differ.

2.1 ECO in the context of ASIC design

Figure 2.1 shows the conventional ASIC design flow on the top, and the various stages
where there could be a need for an ECO, at the bottom. After every stage except after
tape out, one can perform pre-mask ECO to avoid re-running the previous stages; thus
saving time. The post-mask ECO block is shown in green as this stage is the focus of
this thesis.

6



CHAPTER 2. BACKGROUND

LOGIC 
SYNTHESIS

FLOOR 
PLANNING
(PHYSICAL 

SYNTHESIS)

PLACEMENT CTS

ROUTING
& 

FILLER 
INSERTION

FINAL 
NETLIST

SYNTHESIZED 
NETLIST

RTL 
DESCRIPTION

TAPE OUT
(GDSII)

STILL BUGS

RTL FROZEN

MAKE CHANGES 
WITHOUT RE-

RUNNING 
SYNTHESIS

SYNTHESIS DONE

CAN STILL RUN 
ECO- 

PLACEMENT ON 
ECO CELLS

PLACEMENT DONE

CAN STILL DO 
METAL ONLY 
ECOs USING 

“SPARE CELLS”

TAPED OUT BASE
 LAYERS

PRE-MASK ECOs
POST-MASK 

ECO

Figure 2.1: Purpose of an ECO in the context of the ASIC design flow

It can be seen from the Figure 2.1 that, once the Register Transfer Level (RTL) code
is frozen, in case of a bug found in the code, the ECO could be performed to fix that
bug. Moreover, even if the synthesis has been carried out, by modifying the netlist
manually the necessity of re-running the synthesis can be avoided. Furthermore, if the
standard cells have been placed, the new cells can still be added by performing the ECO
placement. Finally, after the tape out by doing metal-only ECOs, the necessity to change
the base mask layers could be avoided. Hence, saving cost and time.

2.2 IC Compiler ECO flows comparison

Figure 2.2 compares the ICC’s ECO flows. If the ECO has to be performed in pre-mask
stage then the corresponding flow in ICC is known as Unconstrained ECO flow else,
known as Freeze Silicon ECO flow for the post-mask stage.

In the unconstrained ECO flow, the placement of the design is not fixed, so new cells
can be added and existing cells can be moved or deleted. As a result, spare cells are not
needed. After deciding the ECO changes to be made, the original netlist is manually
edited to implement those changes. Subsequently, the power and ground connections
should be updated if the new cells have been added. Furthermore, the new cells to be
added as an ECO are placed using the incremental placement known as ECO placement,
followed by incremental Clock Tree Synthesis (CTS) and the ECO route. One additional
step of fixing the routing Design Rule Checks (DRC) is performed if the ECO route was
not able to remove all the DRC. The amount of iterations the tool has to perform in
order to minimize the DRC is known as routing closure [14].

7



CHAPTER 2. BACKGROUND

ECO

Unconstrained
/Freeze Silicon

Unconstrained
(Pre-Mask)

Update Netlist 
with ECO

Freeze Silicon
(Post-Mask)

Update Netlist 
with Freeze 
silicon ECO

Update Power 
& Ground 

Connections

Spare Cell 
Mapping 

ECO
PLACEMENT

NO
Normal 

Placement
& Routing

ECO 
Placement

& 
Incr. CTS

ZRT_ECO 
ROUTE

Update Power 
& Ground 

Connections

Freeze Silicon 
Placement

FIX Routing 
DRCs

Figure 2.2: Comparison between ICC’s ECO flows

In the freeze silicon ECO flow, the placement of the design is fixed so no new cells can
be added. As a result, the spare cells that were added prior to running ECO process
have to be used to perform the desired change. The design netlist is updated according
to the freeze silicon ECO requirements first. Then the spare cell mapping is carried
out which in conventional sense would be to find appropriate functionality cells in the
design. This mapping could either be performed automatically by the tool or manually
by hand. Subsequently, the power and ground connections of the spare cells used should
be updated. The existing cells which are no longer required would not get deleted, rather
would be made as spare cells. The ECO routing step and the DRC fixing step is similar
to the unconstrained flow.

Methods to represent an ECO

Whenever a bug is encountered in the design , a new specification is defined to remove
that bug. This specification can be called as an ECO. In ICC, irrespective of the type
of ECO, the ECO changes can be represented using two methods.

1. By using Verilog Netlist. : The new netlist which is formed by editing the
original netlist is called ECO netlist. Example 2-1 shows that ECO netlist.

8



CHAPTER 2. BACKGROUND

Example 2-1: ECO changes described using Verilog Netlist

// O r i g i n a l N e t l i s t .
. . . . . .
//ECO changes to be made
MYOR2 D1BWP7THVT OR2 eco ( . Z ( c n t 0 o r z ) , . A1 ( n18 inv ) ,

. A2 ( cnt [ 0 ] ) ) ;
MYINV D1BWP7THVT INV1 eco ( . I ( n18 ) , .ZN ( n18 inv ) ) ;
MYNR3 D2BWP7THVT NOR3 eco ( . A1 ( cnt [ 2 ] ) , .ZN ( cnt 1 nor ) ,

. A3 ( cnt [ 1 ] ) , . A2 ( cnt [ 0 ] ) ) ;

2. By using Tool Command Language (TCL) commands to describe the
changes.
Example 2-2 shows the ECO change described using TCL command.

Example 2-2: The ECO changes described using TCL commands

c r e a t e n e t cnt [ 0 ] o r z

d i s connec t ne t cnt [ 0 ] [ g e t p i n s {U98/ I } ]
c r e a t e c e l l OR2 eco MYOR2 D1BWP7THVT

connect net cnt [ 0 ] [ g e t p i n s {OR2 eco/A2} ]
connect net cnt [ 0 ] o r z [ g e t p i n s {U98/ I } ]
connect net cnt [ 0 ] o r z [ g e t p i n s {OR2 eco/Z} ]

#connec t ne t n18 [ g e t p i n s {OR2 eco/A1} ]
c r e a t e c e l l INV1 eco MYINV D1BWP7THVT

###For adding an i n v e r t e r wi th d r i v e 1 on the n18 net
c r e a t e n e t n18 inv
#d i s c o n n e c t n e t n18 [ g e t p i n s {OR2 eco/A1} ]

connect net n18 [ g e t p i n s { INV1 eco/ I } ]
connect net n18 inv [ g e t p i n s {OR2 eco/A1} ]
connect net n18 inv [ g e t p i n s { INV1 eco/ZN} ]

#adding an 3−input nor ga t e wi th d r i v e 2 to c n t r e g 1 and U87
c r e a t e n e t cnt [ 1 ] nor

c r e a t e c e l l NOR3 eco MYNR3 D2BWP7THVT

di s connec t ne t cnt [ 1 ] [ g e t p i n s {U87/A1} ]
connect net cnt [ 1 ] [ g e t p i n s {NOR3 eco/A3} ]

connect net cnt [ 1 ] nor [ g e t p i n s {NOR3 eco/ZN} ]
connect net cnt [ 1 ] nor [ g e t p i n s {U87/A1} ]
connect net cnt [ 0 ] [ g e t p i n s {NOR3 eco/A2} ]
connect net cnt [ 2 ] [ g e t p i n s {NOR3 eco/A1} ]

9



CHAPTER 2. BACKGROUND

The cells marked in red in the above examples are the new cells to be added in the
design also known as ECO cells. In case of a pre-mask ECO, these cells would be
added from the cell library, whereas in the case of post-mask ECO, spare cells having
similar functionality would be required.

2.3 Pre-mask ECO flow

The aim of this section is provide detailed explanation of all the necessary steps involved
in performing a pre-mask ECO. As said earlier, in this type of ECO, as the design has
not been taped out, cell placement can be changed.

Figure 2.3 shows the detailed flow for doing an pre-mask ECO. The Figure assumes
that the design has already been placed and routed and has spare cells. The part of the
flow which is marked in red lines is the ICC’s unconstrained ECO flow (as in Figure 2.2).

LOGIC 
SYNTHESIS

FLOOR 
PLANNING
(PHYSICAL 

SYNTHESIS)

PLACEMENT CTS

ROUTING
& 

FILLER 
INSERTION

FINAL 
NETLIST

SYNTHESIZED 
NETLIST

RTL 
DESCRIPTION

ECO NETLIST
COMPARE 

NETLIST
PLACE_ECO_CELLS ROUTE_ZRT_ECO

ECO FINAL 
NETLIST

Unconstrained ECO FLOW

Formal 
Verification

Modified 
RTL

ECO Description GDSII

YES

NO

Figure 2.3: Complete pre-mask ECO flow

The basic flow can be divided into 6 main steps.

Step-1 Describe an ECO specification.

Step-2 Capture the specification.

Step-3 Compare the ECO netlist with the design netlist.

Step-4 Update the placement with ECO changes.

Step-5 Perform the ECO routing.

Step-6 Insert filler cells.

10



CHAPTER 2. BACKGROUND

Step-1: Describe an ECO specification

An ECO specification is a manually drawn diagram to describe the changes to be made.
The specification can be made by exploring the original netlist using the graphical tools
such as Design Vision. This tool helps to explore a portion of the schematic to find the
nets to be removed and to add the new nets. Figure 2.4 shows an example of manually
drawn specification. The blue marked nets and cells are the changes to be made to the
existing design. The crossed nets are the nets to be removed.

Q

Q
SET

CLR

D

Q

Q
SET

CLR

Dn53 n55

eco1_n

eco2_net

eco1_cell eco2_cell 

CFG_CNT_U1721 

n52

n
5

1

Figure 2.4: An ECO specification

Step-2: Capture the specification

As mentioned in the section 2.2, there are two ways to represent the ECO specification.
Example 2-3 shows corresponding ECO netlist of the specification shown in Figure 2.4.

Example 2-3: ECO netlist (ECO by Verilog)

// O r i g i n a l N e t l i s t .
MYBUF D1BWP7THVT CFG CNT U1721 ( . I ( n52 ) , .ZN ( n51 ) ) ;
. . . . . .
//ECO changes to be made
MYND D2BWP7THVT e c o 1 c e l l ( . Z ( eco1 n ) , . A1 ( n51 ) ,

. A2 ( n53 ) ) ;
MYINV D1BWP7THVT e c o 2 c e l l ( . I ( eco1 n ) , .ZN ( eco2 net ) ) ;

Step-3: Compare the netlists

In this step, the tool compares the design netlist (original netlist) with the ECO netlist
and displays the ECO changes. From this output, one can also verify whether the tool
identified all the intended changes.

If the changes are made by directly editing the original netlist like in example 2-3 then
the command used in the ICC tool is:
eco_netlist -by_verilog_file "verilog file"

11



CHAPTER 2. BACKGROUND

If the changes are described using the TCL file then the command used in the tool is:
eco_netlist -by_tcl_file "TCL file".

Step-4: Incremental placement (ECO placement)

If the standard cells are fully placed, the placement of the new cells is performed using
incremental placement command. The command place_eco_cells is used for this pur-
pose. This command uses the spare area in the design to place the new cells.

Figure 2.5 shows the placed and routed design having spare cells. Figure 2.6 shows
the design after incremental placement is performed. It can be seen from the latter that
the ICC tool did not use spare cells for placing new cells. Rather, it took new cells from
the library and placed them onto the spare area. The two instances that were added
according to the example 2-3 are circled.

Figure 2.5: Before the ECO Place-
ment [3] Figure 2.6: After the ECO Placement

[3]

Step-5: ECO Routing

ECO routing is performed only if design has already been fully routed. The command
used in the ICC to perform this type of routing is route_zrt_eco. The command
reroutes the modified nets and remove the dangling nets after an ECO. Moreover, it also
fixes the remaining routing DRC. Dangling nets are the old nets which are no longer
driven by any pins/ports. Such type of nets are created as a result of ECO changes.

Figure 2.7 shows the explanation of what happens during an ECO and after the ECO
routing. During the ECO, the tool deletes old wires or adds new wires according to the
ECO specification. The dangling wires created due to this can be seen in the top Figure.

12



CHAPTER 2. BACKGROUND

Figure 2.7: The design state during ECO (top) and after ECO routing (bottom) [3]

During the ECO routing, the router touches only those nets which have been modified
due to the ECO. Thus the critical nets in the design would be preserved. The bottom
Figure shows the state of the design after the ECO routing. The physical connections
made by the ECO router can easily be recognized from the Figure.

Step-6: Insert filler cells

After achieving the routing closure in the ECO routing, filler cells are used to fill the
remaining gaps in the design. Command insert_filler_cells is used in ICC to insert
fillers. There are two advantages of filling.

• Ensure the continuity of the power and ground rails in the floorplan.

• Ensure the continuity of the N+/P+ well.

Figures 2.8 and 2.9 show a routed design before and after the filler cell placement. Gaps
in the standard cell rows can be seen in the former Figure. After filling the gaps, the
resulting design can be seen in latter Figure. Different size filler cells used according to
the size of gaps are highlighted in latter.

13



CHAPTER 2. BACKGROUND

Figure 2.8: Placed and Routed de-
sign having gaps in the rows Figure 2.9: Placed and routed design

after filling gaps with filler cells

2.3.1 Formal Verification

Formal Verification is performed to detect the differences that might have been intro-
duced into the design during it’s implementation. It is used to verify the logical equiva-
lence between the two designs without the use of test vectors.

As shown in Figure 2.10, ECO final netlist (implemented output) that is generated after
an ECO has to be verified against the Modified RTL (expected output). They should be
equivalent in order for the ECO to be successful. Otherwise the ECO netlist (see Figure
2.3) has to be modified until successful verification is achieved. Tool Synopsys-Formality
was used to perform this. This step is explained using an example in section 4.5.1.

RTL Description

Final Netlist

Design Process
Formal 

Verification

Logically 

Equvivalent ?

Yes/No

Figure 2.10: Formality[4] verification flow

14



CHAPTER 2. BACKGROUND

2.3.2 Complete ECO flow script

Appendix A gives the script for doing the unconstrained ECO flow. It provides necessary
commands to perform the basic flow. The script assumes the case where the design is
already placed and routed and there is a need for a pre-mask ECO.

2.4 Post-mask ECO using conventional spare cells

In this section, the freeze Silicon ECO flow of ICC is explained in detail. The aim of the
section is to show how the conventional spare cells are used to carry out a post-mask
ECO.

As said earlier, in the post-mask ECO, the placement of the design is fixed, so new
cells cannot be added and old cells cannot be moved or deleted. As a result, the metal-
only ECO is essential. In order to perform this, spare cells which in conventional sense
are logic gates would be needed. These cells could be added before or after the place-
ment, so that they can be used to accommodate late arriving changes.

The most important design decision in using the conventional spare cells is to decide
upon the type of cells to be used as spare cells [8][9][15] and ways to spread them inside
the design [16][17][7]. In ICC, there are two ways to insert spare cells in the design.
One is by manually instantiating them in the Verilog netlist and another is by using
insert_spare_cells command. The former should be use to insert them before the
placement and the latter should be after the placement. Furthermore, once the spare
cells have been inserted, they can be spread inside the design either manually, using
spread_spare_cells command, or automatically by the tool.

Figure 2.11 shows the representation of a design after inserting the spare cells. Each
spare cell depict a predefined logic gate that can be activated by connecting it’s inputs
and outputs according to the ECO specification.

15



CHAPTER 2. BACKGROUND

INV

AND2 OR2 DFF
SPARE

1
INV1 AOI2

DFF INV FILL1 BUF4AND3

SPARE3NAND3 NOR2XOR2

XNOR2 DFF
SPARE

1

SPARE4DFFXNOR2INV1AND3

FILL1

FILL2

FILL8 FILL1

Figure 2.11: Representation of a design having conventional Spare cells

Figure 2.2 shows the basic freeze silicon ECO steps. One additional step of feasibility
analysis is performed in this flow before spare cells are mapped. The main purpose of
doing feasibility analysis is to find out whether there are enough spare cells in the design
to perform the ECO. This step is not required in the pre-mask ECO.

COMPARE 

ECO_NETLIST & 

ORIG. NETLIST

Remove_cell

Feasiblity 

Check

Pass

Map Spare cells 

Automatic ?
NO

Manual select 

spare cells  

YES

Automatic 

select spare 

cells

Activate the 

spare cells

Fail
Warnings and/or 

error

YES
Make that cell a 

Spare cell

NO STOP

Add_cell/

resize_cell etc.

Figure 2.12: Feasibility analysis flow in post-mask ECO

16



CHAPTER 2. BACKGROUND

Combining the Figures 2.2, 2.3 and 2.12, the complete flow can be divided into following
steps.

Step-1 Describe and Capture the specification.

Step-2 Compare the ECO netlist with the design netlist

Step-3 Check the feasibility of the spare cells

Step-4 Carry out Spare cell Mapping

Step-5 Perform the incremental routing and fix the routing DRC.

Step-6 Perform Formal Verification

Step-1: Describe and Capture the specification

The function of this step is same as the first two steps of sections 2.3. However, this flow
is explained using example 2-1 instead of example 2-3.

Step-2: Compare the netlists

Considering example used to explain this flow the corresponding command would be:
eco_netlist -by_verilog_file "verilog file" -freeze_silicon. In comparison
with the command used in unconstrained ECO flow, the additional argument ”freeze silicon”
is used in this flow. As a result the functionality of the command changes a bit. When-
ever tool detects new cells to be added in this stage, the tool does small feasibility checks
for spare cells like valid spacing etc (as shown in Figure 2.12). Furthermore, the removed
cells are not entirely deleted, rather made as spare cells for the future ECO.

Step-3: Feasibility checks

Apart from the checks performed by compare netlist command, additional checks like
finding matching spare cells inside the design to map to, is performed in this step. Com-
mand check_freeze_silicon is used in ICC for carrying out feasibility analysis.

Figure 2.13 shows the feasibility report that was able to find the required spare cells
and Figure 2.14 shows the report in which the missing spare cells are highlighted.

17



CHAPTER 2. BACKGROUND

MYOR2_D1BWP7THVT

MYINV_D1BWP7THVT

MYNR3_D2BWP7THVT

Figure 2.13: Spare cell feasibility
check ”successful”

MYOR2_D1BWP7THVT

MYINV_D1BWP7THVT

MYNR3_D2BWP7THVT

MYOR2_D1BWP7THVT

MYINV_D1BWP7THVT

MYNR3_D2BWP7THVT

Figure 2.14: Spare cell feasibility
check ”failure”

Step-4: Spare cell mapping

After verifying the feasibility of the spare cells, in this step, ECO cells are mapped to
spare cells. The spare cells have fixed functionality,so an ECO cell cannot be place at any
spare cell location.Thus one can either use place_freeze_silicon command to carry-
out automatic mapping or map_freeze_silicon command to perform manual mapping.

Figure 2.15 represents the original taped out design with spare cells and Figure 2.16
shows the design after the spare cell mapping. The three ECO cells highlighted in ex-
ample 2-1 can be seen in latter Figure as mapped to appropriate sized spare cells.

18



CHAPTER 2. BACKGROUND

Figure 2.15: Original design having
appropriate spare cells

Figure 2.16: Design after spare cell
mapping

Step-5: ECO Routing

Once the spare cell mapping is carried out, the connections are made using ECO routing.
This step is already explained in section 2.3.

Step-6: Formal Verification

In this step the routed design is formally verified. Refer to section 2.3.1 for more infor-
mation.

2.4.1 Complete ECO flow script

Appendix B gives the script for doing the freeze silicon ECO flow in ICC.

2.5 Metal-configurable Gate Array spare cells

This section introduces GA type cells and explains the concept of metal-configurable GA
spare cells.

2.5.1 Definition

A gate array is an integrated circuit whose internal structure is an array of gates with
interconnects initially unspecified [2]. The GA implementation can be divided into two
steps.

1. Making transistor masks which is an array of uncommitted transistors gates on
the GA chip

2. Metal interconnection between the transistors to configure a logic functionality.

19



CHAPTER 2. BACKGROUND

From the above two points it can be said that GA spare cells can be developed by
combining array of gates with structured ASICs [18].

2.5.2 Gate Array structure at chip level

Figure 2.17 shows the array of uncommitted transistor gates spread across the chip. It
can be seen that the height of every GA cell is fixed and cells are arranged to form rows
so that they can share same power and ground signals. This arrangement is similar to
the structured ASICs. The space between two rows known as channel is used for inter
cell routing.

Figure 2.17 also highlights the width and the height of the smallest GA cell (unit tile).
The area occupied by this cell is known as one gate array tile. Similarly, the area oc-
cupied by smallest size standard cell is known as standard cell tile. It can be noticed
that, as it is a GA type cell, there is no metal connection between pMOS and nMOS
transistors. Thus, it can be said that, each row consists of arrays of GA tiles.

UNIT TILE

Figure 2.17: Array of unconnected transistors spread to form gate array (GA) structure

In terms of area, the size of a GA tile is more as compared to standard cell tile because
the number of transistors and routing space inside a regular standard cell is optimized
for fixed functionality. As a result, a significant reduction in width is achieved. One the
other hand, in a GA structure, transistors gates are prefabricated, and only metal layers
are available for configuring the GA cells. Therefore, there needed to be enough room
for intra-cell routing.

Figure 2.18 shows the diagram of prefabricated and programmable layers for a GA
cell. It becomes clear from the Figure that the transistor masks are prefabricated and
only the metal layers are used to program the GA cells. Furthermore, the number of
metal layers used to program the GA cells depends on the library used. By reducing

20



CHAPTER 2. BACKGROUND

this number, the metal-masks costs can be reduced.

N+ N+ P+ P+ N+P+

PO PO

N-well

P-substrate

C
O

C
O

C
O

C
O

METAL 1

V
1

METAL 2

V
2

METAL 3

V
3

METAL 4

V
4

METAL 5

VIA-1

VIA-2

VIA-3

VIA-4

Programmable 
layers

Prefabricated 
layers

Figure 2.18: Prefabricated Layers and Programmable layers of a GA cell

2.5.3 Programming gate array cells

A GA tile which, when configured as spare cell, can be used to perform metal-only ECO
is known as a base cell. The base cell should be such that any desired functionality can
be configured upon it using metal layers.

Configuring the new functionality using base cell

Figure 2.19 describes the process carrying out metal-only ECO using base cell. The new
functionality to be added is in the form of netlist. As shown in the Figure, this netlist
consists of only GA type cells. Every ECO cell in the netlist is created by configuring
the metal layers of the base cell.

GAND

Q

Q
S ET

CLR

S

R

GDFF

GBUFF

GOR GAND

ECO NETLIST USING GA CELLS

BASE CELL= ”SPARE CELL”

EACH GA CELL 
FUNCTIONALITY 

CONFIGURED  USING METAL 
LAYERS ONLY

Figure 2.19: Using GA type base cell to realize metal-only ECO

21



CHAPTER 2. BACKGROUND

Figure 2.20 shows some examples of configuring the base cell. Combinational logic as
well as sequential logic can be realized using a base cell. The combinational logic requires
two base cells placed subsequent to each other inside the chip, whereas the sequential
logic requires nine base cells to do a metal-ECO. Thus, it can be said that any logic cell
can be configured using a base cell if sufficient number of base cells is available.

BASE CELL Combinatinal Logic
(2*BASE CELL)

Sequential Logic
(9*BASE CELL)

Figure 2.20: Use of base cell to make different types of cells

Base cell in the context of this project

In this project, the smallest sized GA filler cell is used as a base cell (as the title suggest).
Figure 2.21 shows different sizes of GA filler cells that are defined in the library. It can
be seen that that the GA fillers in this library have gates, source and drains shorted
without connections to supply or ground rails. Hence they do not leak.

22



CHAPTER 2. BACKGROUND

Figure 2.21: Different sized GA filler cells available in the library

2.5.4 Configuring process used in ICC to carryout an ECO

After placing GA fillers inside the design, in case of an ECO, the tool will first search
for GA ECO cell in the library and then swap that GA cell with the GA filler. As the
transistor mask structure of all the GA type cells is similar, only metal layers would
change. Hence, the metal-only ECO.

Figure 2.22 shows example of the ECO swapping process. Lets say the ECO netlist
consists of two ECO cells MYGA NAND and MYGA MUX4. For an ECO, the ICC
would first look for corresponding cells in the GA library. After finding the cells in the
library, the tool would look for equal or big sized GA fillers (MYGA FILL) in the design.
Finally it will perform the swapping as shown in the Figure.

23



CHAPTER 2. BACKGROUND

MYGA_FILL2 MYGA_NAND

MYGA_FILL4 MYGA_MUX4

Figure 2.22: ICC’s ECO swapping process

2.6 Advantages of using Gate Array cells as spare cells

As compared to conventional spare cells, the following are the advantages of using GA
fillers as spare cells

1. As GA fillers are placed in the filler cell area, that area can also be now utilized
for spare cell mapping. This is not the case with conventional spare cell. Figure
2.11 shows that spare cells occupy area other than filler cell area. Thus it can be
said that 100% spare area of the chip can be utilized using the former one.

2. Unlike conventional spare cells, the inputs of GA fillers are shorted (see Figure
2.21) as a result they do not consume any additional leakage current.

3. GA fillers can be programmed to any functionality. Thus they do not suffer from
inflexibility. This is certainly not the case with fixed functionality spare cells.

4. Finding efficient location for spare cell mapping is also a problem in case of con-
ventional spare cells. Failing to find one would affect the routing closure, which
ultimately could lead to timing closure failure.

24



CHAPTER 2. BACKGROUND

2.7 Summary

This chapter gave the essential background of the important key terms used in the
project title, which should assist the reader to understand the detailed flows mentioned
in the subsequent chapters. In the beginning, the need for an ECO was explained with
respect to various ASIC design stages. This was followed by a detailed explanation of
ICC’s ECO flows. Finally the chapter concluded by providing the necessary background
regarding GA type cells and their usage to realize an ECO. The next chapter will modify
the ICC’s freeze silicon ECO flow and will present the GA spare cell methodology.

25



3
Gate Array spare cell

methodology

In this chapter, the freeze silicon ECO flow mentioned in the previous chapter will be
modified to use gate array (GA) filler cells as spare cells. The steps mentioned are
designed to work with ICC tool and are based upon the steps described by Synopsys
[13]. The spare cell methodology will be explained in two parts. Section 3.1 will explain
necessary steps to to insert GA filler cells as spare cells. Subsequent section will make
the use of this spare cells and will present the modified ECO flow. In the last section,
the chapter will be summarized.

3.1 Adding Gate Array fillers as spare cells

This section explains the method to add GA fillers inside the design and configure them
to use them as spare cells. Appendix C shows the complete script for the same. The
script can be explained in five steps.

Step-1 Preparation for adding GA fillers

Step-2 Map the filler cells

Step-3 Report the mapped cells

Step-4 Insert them inside the design

Step-5 Setting them as spare cells

Step-1: Preparation for adding Gate Array fillers

In the most common case, the GA filler cells would not be configured as normal filler
cells. This is because standard cell type fillers are usually used for filling the gaps. In

26



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

such a case, the GA fillers need to be set as filler cells otherwise, the tool will complain.
The script shown below (example 3-1) serves this purpose. example 3-1 highlights the
name of the GA filler cells defined in the library.

Example 3-1: Script to set GA filler cells as normal filler cells

cmMarkCellType
setFormField mark ce l l type l ibrary name MY LIB
setFormField mark ce l l type ce l l name MYGA FILL.ˆ∗
setFormField mark ce l l type pattern match 1
setFormField mark ce l l type c e l l t y p e ”std f i l l e r ”
formApply mark ce l l type
formOK mark ce l l type

Step-2: Map filler cells

The second command in the script map_unit_tiles, is used when there are multiple so
called unit tiles defined in the library. As the dimensions of the GA cells differs from
regular standard cells, there exists a need for differently sized unit tiles, one for GA, and
one for regular standard cells.

The width and height of a unit tile should be such that, it equals to the corresponding
tile (GA/standard cell) as defined in the section 2.5.2. Before moving into the details of
each unit tile it is important to know the purpose of a unit tile irrespective of cell type
in the ASIC design.

Concept of a Unit Tile

Unit tile is used when rows are created in the floorplan [19]. Without this information
the floorplan will not be created. A unit tile contains following information.

• It defines valid spacing for the standard cells. This information is used by place-
ment command.

• It contains the basic wire pitches according to the technology file. Routing tracks
are defined based on this information.

– Wire pitch= Wire spacing + wire width

• It defines power and ground rails.

Figure 3.1 shows the floorplan in which rows are created with unit tile named (Tile
Name) ”unit”. It also highlights three other parameters site_count, site_space,

site_type. These parameters are explained in the subsequent paragraph.

27



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

Figure 3.1: A Floorplan in which rows are created with site_type=unit

Figure 3.2 explains the two parameters site_space and site_count. It can be seen
from the Figure that a row is divided into sites. The width of each site that is the
parameter site_space, would be equal to the width of the unit tile and site_count

would be the number of sites in a row. The third parameter, site_type determines
which unit tile type to be used to define the rows. Thus, depending on which site_type

is used, the other two parameters will change accordingly.

Site

Site_space=
GA/STD. tile width 

Site Site Site Site Site Site

Site_count=No. of Sites in a row

Figure 3.2: Concept of a row with respect to unit tile parameters

Figure 3.3 shows a standard cell unit tile named ”unit” that is defined in the library by
default. Whenever the floorplan is created using site_type= unit, the tool uses this
unit tile to create the rows. This is the default unit tile used by ICC to create rows.
The parameter site_space which is marked in the Figure with spacing of ’0.2’ is same
as the number shown in Figure 3.1. The routing tracks and the power and ground rails
can easily be seen in the Figure.

28



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

Figure 3.3: Standard cell unit tile (”unit”)

Function of this step

The command is used to map multiple GA filler cells to unit tiles in the design. Without
this mapping information, the tool will not be able to place the GA filler cells inside the
design as GA filler cells have different tile size. The additional argument unit_tile_name
defines the name of the unit tile to map to.

Step-3: Report Unit tiles

The command report_unit_tiles reports the tile mapping created by map_unit_tiles

command. Figure 3.4 shows the mapping report. It shows the mapping of GA filler cells
to unit tile named ”unit”.

MYGA_FILL10BWP7THVT

MYGA_FILL4BWP7THVT

MYGA_FILL3BWP7THVT

MYGA_FILLBWP7THVT

MYGA_FILL2BWP7THVT

Figure 3.4: Output of report_unit_tiles command showing the mapped filler cells

29



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

Step-4: Insert Gate Array fillers

After mapping the GA fillers, the next step is to insert them. The fourth command in
the script performs this. The tool will search for rows defined with site_type=unit and
having gaps to place filler cells.

Figure 3.5 shows the design after GA fillers have been added inside the design. It
also highlights the is_spare_cell attribute which tells that the cell currently cannot
be used as spare cell. This points to the next step.

Figure 3.5: Design showing GA fillers inserted

Step-5: Setting the spare cell attribute

The last command in the script is used to set the added GA fillers as spare cells. This
is done by setting the is_spare_cell attribute. Figure 3.6 shows the final design after
GA fillers have been added and marked as spare cells.

30



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

Figure 3.6: Final output design containing GA Fillers and marked as spare cells

3.2 Post-mask ECO using Gate Array fillers

In the previous section, the process of inserting GA fillers and marking them as spare
cells was shown. In this section, the method to perform an ECO using this spare cells is
presented. Before jumping into the ECO flow, some additional preparation for the spare
cell mapping must be performed.

3.2.1 Preparation for the ECO

In order for tool to be able swap every GA ECO cell with appropriately sized GA filler
(see section 2.5.4) this step must be performed. Example 3-2 shows the script for the
same. The scripts works only on the cells which has a FRAM view [14] defined in
the library. By giving all the GA cells the gate_array_master_type attribute, any
number of GA ECO cells could be swapped with a spare cell (as long as the swapping
requirements are matched) [13].

Example 3-2: Script for being able to swap GA ECO cell with GA spare cell

set type 1 [ l i s t $ a l l G A c e l l s i n l i b r a r y ]
open mw lib −wr i t e r e f MY LIB
foreach c e l l $type 1 {
echo $ c e l l
open mw cel $ cell.FRAM
s e t a t t r i b u t e −c lass mw cel $ cell.FRAM \
gate a r ray maste r type 1

31



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

c lo se mw ce l −save
}

3.2.2 Gate Array ECO process

The GA ECO flow can be divided into five steps. Note that, only the necessary changes
concerning GA flow (if any) would be mentioned here. Refer to section 2.4 for detailed
explanation about the steps.

The following are the steps to perform a freeze silicon ECO using GA cells in ICC
.

Step-1 Describe and Capture the specification.

Step-2 Compare the ECO netlist with the design netlist.

Step-3 Check the feasibility of the spare cells.

Step-4 Carry out Spare cell Mapping.

Step-5 Perform the incremental routing and fix the routing DRC.

Step-6 Perform Formal Verification.

Step-7 Verify the metal-only ECO.

Step-1: Describe and Capture the specification

The only difference in this step is that, the ECO specification should use GA type cells
in the ECO netlist instead of regular standard cells (see example 2-1 or example 2-2).
This means that, OR gate, INV gate and NOR gate should be replaced by their GA
equivalents. Example 3-3 shows modified ECO netlist that uses GA type cells.

Example 3-3: ECO netlist of Example 2-1 modified to use in GA ECO flow

// O r i g i n a l N e t l i s t .
. . . . . .
//ECO changes to be made
MYGA OR2D1BWP7THVT OR2 eco ( . Z ( c n t 0 o r z ) , . A1 ( n18 inv ) ,

. A2 ( cnt [ 0 ] ) ) ;
MYGA INVD1BWP7THVT INV1 eco ( . I ( n18 ) , .ZN ( n18 inv ) ) ;
MYGA NR3D2BWP7THVT NOR3 eco ( . A1 ( cnt [ 2 ] ) , .ZN ( cnt 1 nor ) ,

. A3 ( cnt [ 1 ] ) , . A2 ( cnt [ 0 ] ) ) ;

Step-2: Compare the ECO netlist with the design netlist

This step is exactly same as the explained in section 2.4.

32



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

Step-3: Check the feasibility of the spare cells

The feasibility analysis process is simpler in this case as compared to using conventional
spare cells. Instead of searching for matching spare cell, the tool would just have to check
for the appropriately sized GA filler. The remaining flow (see Figure 2.12) however, will
remains the same.

Step-4: Carry out Spare cell mapping

The mapping process is rather a replacement process. As described earlier, the tool picks
the GA ECO cells from the library, finds the spare cells (GA fillers) in the design and
swaps the spare cells with GA ECO cells.

Figures 3.7 and 3.8 show the design before and after the mapping. It can be noticed from
the Figures that, the ”MYGA OR2”(OR2_eco) ECO cell was swapped with two spare
cells having a drive strength ’1’. Moreover, the ”MYGA NR3”(NOR3_eco) was swapped
with a spare cell having drive strength, ’3’ and so on.

Figure 3.7: Design highlighting the
different sized GA spare cells

Figure 3.8: Design after carrying out
spare cell mapping

Step 5: ECO Routing

The routing process is no different from the one explained in section 2.4. Figure 3.9
shows the transistor gate level view of the design before the ECO. Figure 3.10 shows the
resultant design after the complete ECO flow has been performed. It can be seen from
the latter that the location of underlying transistor masks after the ECO has not been
changed, rather only the internal metal routing has been changed. This validates the
metal-only ECO concept.

33



CHAPTER 3. GATE ARRAY SPARE CELL METHODOLOGY

Figure 3.9: Transistor gate level view
of the design before the ECO

Figure 3.10: Transistor gate level
view of the design after the ECO

Step-6:Formal Verification

This step is similar to the one explained in section 2.3.1.

Step-7: Verifying the metal-only ECO

Apart from verifying the ECO functionally, it is also important to verify whether only
metal layers were used to carry out an ECO. In this case, the size of the ECO was small
and more importantly, the ICC’s ECO flow was used. As a result, it is not necessary to
verify metal-only ECO, as tool makes sure this thing. However, it will be shown later
that, in case of big ECOs, where tool commands are not used, this step has to be verified
manually.

3.3 Summary

This chapter discussed the ICC’s freeze silicon ECO flow using GA fillers as spare cells.
The chapter started by providing steps for adding GA fillers and marking them as spare
cells. The subsequent section explained necessary steps to perform a simple ECO using
such spare cells with the help of an example. During this process, it was shown that once
the ECO is performed one needs to verify the ECO both functionally as well as in terms
of layers. The next chapter will make the use of this flow and discuss more advanced
methods to carry out an ECO whenever the amount of changes to be made are big.

34



4
Advanced ECO methodology

Previous chapter explained the ICC’s gate array (GA) ECO flow to perform an small
ECO like bug fixing. This chapter focuses on using GA ECO flow to perform larger
ECOs like adding a new functionality. The limitations of the conventional ICC ECO
flow is mentioned in section 4.1. In order to overcome this, section 4.2 will provide
overview of the new ECO methodology known as Advanced ECO methodology. This
flow is explained using an example mentioned in section 4.3. The flow is divided in two
parts and so the subsequent sections serves the purpose of explaining each of them in
detail . The chapter will conclude by summarizing this new method.

4.1 Purpose of finding new methods for doing an ECO

The conventional GA ECO flow suffers from some major limitations when it comes to
realizing big ECOs. These limitations are as follows:

• Impractical to manually modify netlists having large number of changes.

• Placement algorithms used by the incremental placement command fail to provide
optimal results.

• There is no option to perform full CTS. Only incremental CTS can be performed.

• Difficult to solve issues related to routing, timing optimization etc.

Thus, the purpose of finding a new methodology is to overcome the above mentioned
limitations by incorporating the full placement and routing algorithms into the ECO
methodology.

35



CHAPTER 4. ADVANCED ECO METHODOLOGY

4.2 Overview of the Advanced ECO methodology

Figure 4.1 gives the overview of the Advanced ECO methodology. As shown in Figure,
this flow can be divided into two parts top level flow and block level flow. The block
level flow takes RTL of the new functionality as an input and performs conventional
ASIC design flow on it (as in Figure 2.1). This flow is modified in order to make it work
on GA cells. After performing the ASIC design steps, the FRAM [14] view of the new
functionality is created so that, the design can be used as a MACRO in the original chip,

Advanced Synthesis ECO Flow

Top_Level Flow Block_Level Flow

Intergration 
of the Block

Freeze 
Silicon ECO 

flow

RTL to GDSII 
Flow

Creating 
FRAM view

Figure 4.1: Overview of the Advanced ECO methodology

The top level flow takes the MACRO as an input and integrates it into the original taped
out chip (top level design). This integration process involves using freeze silicon ECO
flow mentioned in the previous chapter.

4.2.1 Overview of the Block Level Flow

Figure 4.2 shows detailed flow chart of the steps to be performed in the block level flow.
Starting with the RTL (Block RTL), followed by synthesis using only GA cells. Sub-
sequently, the gate level netlist is verified in the formal verification step. The physical
design steps are then performed.

The physical design steps are also modified to be able to place and route GA type
cells using conventional commands. The steps involved in the physical design flow are
given in the subsequent subsection. After routing is completed, the final layout netlist
is verified against the block RTL using formal verification. Finally, the FRAM view is
created as mentioned earlier. The stream out of a Graphic Data System (GDS) file is
optional.

36



CHAPTER 4. ADVANCED ECO METHODOLOGY

Block_RTL

Synthesis using
 only GA cells

Synthesized Netlist Reference Design
Formal 

Verification

No

Block_Final Netlist 

Formal 
Verification

Reference Design

Make FRAM view

YES

Modified
Physical Design for 

GA cells

No

Yes

Block_GDS

Figure 4.2: Modified ASIC design Flow for using GA type cells

Physical design steps

The following physical design steps are performed, once the synthesized netlist is ob-
tained that used only GA type cells

• Create Floorplan

• Full Placement using conventional placement command place_opt instead of in-
cremental placement (place_freeze_silicon).

• Full CTS.

• Routing using conventional routing command route_opt instead of route_zrt_eco

• After achieving the routing closure, GA fillers are inserted in the all the remaining
area.

• Create FRAM view using create_macro_fram command.

37



CHAPTER 4. ADVANCED ECO METHODOLOGY

4.2.2 Overview of the Top level flow

As said earlier, the top level flow is used to integrate the MACRO created in the block
level flow . The top level flow can be divided into six main steps.

Step-1 Modify the RTL of the top level design

Step-2 Macro Mapping.

Step-3 Write out Verilog Netlist and modify it

Step-4 Perform Gate Array freeze silicon ECO steps

Step-5 Formal Verification.

Step-6 Metal-only ECO verification.

4.3 Example description

The methodology is explained using the example shown in Figure 4.3 . The blocks that
are surrounded by the red dashed line is the new functionality to be added (ECO). The
blocks that are surrounded by light red line is the top level design (existing chip). The
spare area, which will used to integrate the new functionality, is not shown in the top
level design.

MUX

R
A

M

Control

M
u

ltip
ly A

ccu
m

u
late

 
U

n
it(M

A
C

)

Select

CORDIC IP

MUX Select

ECOEXISTING 
CHIP

Figure 4.3: Example used for explaining the Advanced ECO methodology

38



CHAPTER 4. ADVANCED ECO METHODOLOGY

The green arrows, connecting ”ECO” block with the ”existing chip” block, are the inter-
face arrows. After mapping the MACRO onto the spare area, the inputs and outputs
of the MACRO are connected according to these interface arrows. This connection
procedure is carried out using the freeze silicon ECO flow.

4.4 Implementation- Block Level Flow

In this section, the block level flow given in Figure 4.2 is explained in detail.

4.4.1 Block RTL

Example 4-1 shows the TOP level RTL code of the new functionality to be added. The
top module name ”top cordic” would be used to represent this functionality in the top
level design.

Example 4-1: RTL description of the new functionality

‘ include ”mac updated . v”
‘ include ”con t r o l . v”
‘ include ” s e l e c t g e n e r a t e . v”
‘ include ”pa Cordic . sv ”
‘ include ”Cordic . sv ”

module t op co rd i c (
input clk , r e s e t ,
input integer word ,
input pa Cordic : : Mode mode1 ,
input signed [ 2 3 : 0 ] out1 ram , out2 ram ,
output signed [ 2 3 : 0 ] out co rd i c1 , mac out rea l , mac out imag ,
output signed [ 2 8 : 0 ] out co rd i c2 ,

output sel mux ram
) ;
wire ou t c t r l , out done , sel mux ;
wire signed [ 2 3 : 0 ] c o rd i c ou t x , c o r d i c ou t y ;
wire signed [ 2 8 : 0 ] c o r d i c o u t z ;

reg signed [ 2 3 : 0 ] mux out real , mux out imag ;

mac v upd m1 ( . c l k ( c l k ) , . r e s e t ( r e s e t ) , . cnt word (word ) ,
. dataa ( out1 ram ) , . datab ( out2 ram ) , . c t r l ( o u t c t r l ) ,

. a dd e r ou t r e a l ( mac out rea l ) , . adder out imag ( mac out imag ) ,
. done ( out done ) ) ;

c on t r o l c1 ( . c l k ( c l k ) , . done ( out done ) , . c t r l ( o u t c t r l ) ) ;
s e l e c t o u t s e l ( . c l k ( c l k ) , . c t r l ( o u t c t r l ) , . s e l ( se l mux ) ) ;

assign mux out rea l = ( sel mux ) ? mac out rea l : out1 ram ;
assign mux out imag = ( sel mux ) ? mac out imag : out2 ram ;
assign sel mux ram = sel mux ;
endmodule

4.4.2 Synthesis using only Gate Array cells

In this step, the RTL code is synthesized. Figure 4.4 shows the synthesized gate level
netlist that contains only GA type cells. Note that, the library should contain all the
required GA cells for mapping. Missing some of the cells, could result in a netlist having
regular standard cells, which would ultimately lead to GA ECO failure.

39



CHAPTER 4. ADVANCED ECO METHODOLOGY

Figure 4.4: Synthesized netlist containing only GA type cells

4.4.3 Floorplanning

After successful formal verification, the physical design steps are followed. The first step
in this flow is to create a floorplan.

In the floorplan step, rows are created among other things. [14]. These rows would
be used by the placement tool to place the cells. In the context of this methodology, as
the netlist contains GA type cells, the rows should be created such that,

• The conventional placement command should be able to place the GA type cells.

• The placed GA cells should follow GA structure (spare area) of the top level design.

Reason for emphasizing on Gate Array cell placement

In ICC, the placement command is designed to work with regular standard cells. This
means that, by default the tool looks for regular standard cells in the design netlist. As
a result, in some cases the tool would generate error message when trying to place GA
type cells (This is shown in section 4.4.5).

40



CHAPTER 4. ADVANCED ECO METHODOLOGY

Reason for following the Gate Array structure

Figure 4.5 shows the transistor gate level view of the top level design . The area within
white dotted lines is the spare area filled with GA filler cells. This spare area is used to
map the MACRO.

Figure 4.5: Transistor gate (red line) level view of the top level design

It can be seen that, the underlying transistor gates of GA filler cells are aligned below
each other and follow a regular GA structure (as in Figure 2.17). Due to such type of
arrangement in the top level design, it should be made sure that the placed GA cells
follow the same structure. The benefit of doing this is that when the resultant MACRO
is to be mapped upon the spare area, the underlying transistor gates of the cells inside
the MACRO, would exactly map upon the underlying transistor gates of the GA filler
cells (as in Figure 4.18).

After the successful mapping, the GA filler cells are removed from beneath the MACRO.
This means that the uncommitted transistors of GA fillers at the top level design would
be replaced by the committed GA cells from the MACRO. Furthermore, as the MACRO’s
underlying transistor gates lie at the same location, the transistor masks would be reused
and only the metal-masks would have to be made. Thus, the reason for following the
GA structure, is to achieve metal-only ECO irrespective of the ECO size.

41



CHAPTER 4. ADVANCED ECO METHODOLOGY

4.4.4 Placement of the Gate Array cells

Figure 4.6 shows the transistor gate level view of the placed GA cells whose floorplan
was created using site_type = unit (by default floorplan). Comparing this Figure with
Figure 4.5, it can be said that, if the MACRO was made from this design and mapped
onto the spare area, the MACRO mapping will not be successful. Figure 4.17 gives an
idea of the unaligned transistor mapping. As a result, the second floorplan objective
would have failed.

Figure 4.6: GA cells placed on rows created using site_type = unit

Thus, it can be concluded that, rows should not be created using standard cell unit tile
(”unit”), if GA cells are to be placed.

Need for creating new unit tile for Gate Array cells

The reason for the misaligned placement of the GA cells is that the GA cells are bigger
than standard cells. The tool will be able to place the cells wherever it finds the sufficient
place in a row, but, the GA structure will not be followed. In order to overcome this,
a new unit tile has to be defined in the library such that the width of this unit tile
(site_space) would be equal to the width of the base cell (GA filler). As a result of
creating such unit tile, the distance between the sites in a row, would be equal to the
base cell width, which would aid to follow the GA structure.

Gate Array unit tile

In order to follow the GA structure, a new unit tile is defined with site_type = gau-

nit. Generally, this unit tile would be available as predefined in the library, if the library

42



CHAPTER 4. ADVANCED ECO METHODOLOGY

contained the GA type cells. The reason for not discussing and/or mentioning this unit
tile earlier, was to give the reader the purpose for creating one .

Figures 4.7 and 4.8 show the comparison between standard cell unit tile (”unit”) and
the GA unit tile (”gaunit”). It can be noticed that, the height is same for both but the
width is equal to the corresponding base cell width.

Figure 4.7: Tile used for creating
rows with site_type=unit

Figure 4.8: Tile used for creating
rows with site_type=gaunit

4.4.5 New methods for placing Gate Array cells

As the default floorplan did not work for GA type cells, several other methods have been
tried out. These methods assume that, both ”gaunit” and ”unit” tiles are available in the
library.

Using site_type= gaunit to create rows in a floorplan

In this scenario, all the rows were created using ”gaunit” tile. Figure 4.9 shows the
corresponding floorplan. The number ’1.0’ besides the site_space parameter indicates
the width (in µm) of the ”gaunit” tile. Note that, this number is the same as the one
shown in Figure 4.8.

43



CHAPTER 4. ADVANCED ECO METHODOLOGY

Figure 4.9: Rows created using site_type = gaunit

When tried to place the cells, the tool terminated with an error showing that it was not
able to find the regular standard cells in the design. The reason for this kind of error
was, the tool being not able to find rows created with site_type= unit. Thus, it can
be concluded that, during placement, the ICC tool by default assumes that the design
contains regular standard cells.

Using site_type= gaunit and site_type= unit to create rows in a floorplan

In order to resolve the errors of previous scenario, in this scenario, some rows were
created using standard cell unit tile (”unit”) and some were created using the GA unit
tile (”gaunit”). Consequently, the tool was able to place the GA type cells. Figure 4.10
shows that output. Furthermore, it can also be seen that, the second floorplan objective
is met i.e. the cells are aligned. Thus, this floorplan choice was chosen.

Figure 4.10: Figure showing GA cells placed as well as aligned

44



CHAPTER 4. ADVANCED ECO METHODOLOGY

On the down side, it was found that, the tool only placed the cells on the rows which were
created using ”gaunit” tile. This happened because GA cells are attached to ”gaunit”
tile in the library. As a consequence, number of rows that were created using ”unit” tile
remained unused. In order to minimize this number and find out the exact number of
rows needed with site_type=unit, several iterations of floor planing were performed.
In the end, it was found that only one row should exist in the design with site_type

= unit.

4.4.6 Insert Gate Array fillers in all the remaining area

After placing GA cells according to the requirements and the design is routed, GA type
filler cells are inserted to fill the remaining gaps. Figure 4.10 shows the gaps to be filled
and Figure 4.11 shows the final design after filling all such gaps. In the latter, every
part of the design is filled with some kind of GA type cell. It is important to make sure
this is the case, otherwise the overall methodology would fail. The reason for not having
gaps is explained in the MACRO mapping step of the top level flow (see section 4.5.1).

Figure 4.11: Final block level design filled with GA type cells everywhere.

4.4.7 Create MACRO of the final placed and routed design

In this step the MACRO is created. It is used to represent the complete design as a
block in the top level design. The FRAM view which is required for this representation
is created using create_macro_fram command in ICC.

45



CHAPTER 4. ADVANCED ECO METHODOLOGY

4.4.8 Results-Block Level Flow

The output of this flow is shown in Figure 4.12. It shows the MACRO instantiated in
the top level design but not mapped. The command create_cell is used to instantiate
the design as a MACRO from the FRAM view.

Figure 4.12: New functionality instantiated as an MACRO in the top level design

4.5 Implementation - Top level flow

In this section, the steps of the top level flow as mentioned in the section 4.2.2 would
be explained in detail. As mentioned earlier, the purpose of this flow is to integrate the
MACRO representing the new functionality into the original taped out chip.

4.5.1 Top level Design Assumptions

In order to integrate the MACRO as an ECO in the top level design, several assumptions
have been made regarding the top level design.

1. The top level design contains a spare area filled with GA type fillers.

2. Some functionality exists in the top level design that is implemented using GA
type cells.

The first assumption is used to explain this flow. It says that, in order to integrate the
MACRO into the top level design, there should exist a dedicated spare area inside the
chip. Moreover, this spare area should be filled with GA filler cells.

46



CHAPTER 4. ADVANCED ECO METHODOLOGY

The second assumption is optional but it is more feasible as compared to the first one.
It can be explained using Figure 4.13. The left side of the Figure shows the original
taped out chip having some blocks (in green) implemented using GA cells. The new
functionality (”New func GA”) to be added as an ECO is shown in orange.

MACRO 1

RAM4 RAM2

RAM1RAM3

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

RAM5 RAM6

Pad

Processor

Pad

Pad

Pad

Pad

BUS

CACHE MMU

STANDARD CELLS

STANDARD CELLS

Func1
_GA

Func2
_GA

New_Fun
c_GA

.FRAM

New_Func
_comb
.FRAM

MACRO 1

RAM4 RAM2

RAM1RAM3

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

Pad

RAM5 RAM6

Pad

Processor

Pad

Pad

Pad

Pad

BUS

CACHE MMU

STANDARD CELLS

STANDARD CELLS

New_Fun
c_comb
.FRAM

Figure 4.13: ECO flow for top level design containing blocks implemented using GA type
cells

Assuming that the new functionality is implemented using GA type cells, it is then
possible to swap the new functionality with one of the blocks in green. The process is
explained as follows. The RTL of the new functionality is combined with the RTL of
the remaining green block. This combined RTL becomes the input to the block level
flow. Subsequently, the block level flow is performed on this input. Lets say the output
is ”New func comb.FRAM”. In order to insert this combined MACRO as an ECO, both
the green blocks are deleted from the original taped out chip and in that same location,
the combined MACRO is mapped using top level flow. Thus ensuring the metal-only
ECO.

Step-1: Modify the RTL of the top level design

The RTL of the top level design is modified to include the RTL of the MACRO (the
Block RTL). This combined RTL is called the modified top level RTL. This RTL will be
used as a reference for formal verification, later in the flow.

Figure 4.14 shows the top level RTL of the original taped out chip and Figure 4.15

47



CHAPTER 4. ADVANCED ECO METHODOLOGY

shows the modified top level RTL. The modifications highlighted in the later, were
made according to the connections defined in the example (green lines in Figure 4.3).

Figure 4.14: Original RTL of the top
level design

Figure 4.15: Modified RTL of the top
level design

Step-2: MACRO mapping

The next step is to map the instantiated MACRO onto the spare area. Figure 4.16 shows
the MACRO, mapped upon the filler cell area. During the mapping, one has to make
sure that the transistors gates of the cells inside the MACRO, exactly map upon the
transistors gates of the filler cells, as explained in section 4.4.3.

Figure 4.16: Top level design after MACRO mapping

48



CHAPTER 4. ADVANCED ECO METHODOLOGY

Figures 4.17 and 4.18 show the difference between unaligned transistor mapping and
aligned transistor mapping. Comparing the width of the transistor gates in both the
Figures, it can be seen that, the transistor gates are thicker in the former. This shows
that the mapping is not perfectly achieved.

Figure 4.17: Transistor level view of
the unaligned transistor mapping

Figure 4.18: Transistor level view of
the aligned transistor mapping

Reason for not having gaps in the MACRO

After successful mapping is achieved, the GA fillers are removed from beneath the
MACRO. If the MACRO contained some gaps, those areas would not have any transistor
gates (as in Figure 4.19). On the other hand, there were no gaps in the spare area of
top level design. This means that in the original chip, the transistor masks layers would
change. As a result, the design will fail the metal-only ECO verification.

Figure 4.19: Figure highlights the gaps within the MACRO

49



CHAPTER 4. ADVANCED ECO METHODOLOGY

Step-3: Write out a verilog netlist and modify it

Before making connections between the mapped MACRO and the top level design, a
verilog netlist is written out. Figure 4.20 shows that netlist. The top level module name
of the block RTL can be seen in the netlist. This means that the tool has found the
instantiated MACRO. The I/Os of the MACRO are not connected to any of the pins at
the top level design, hence the netlist shows it as ”UNCONNECTED”.

Figure 4.20: Top Level Netlist showing the MACRO (top cordic)

The written out netlist is then modified according to the connections defined in the
example. This netlist now becomes the ECO netlist (as in section 3.2.2). Figure 4.21
shows that modified netlist. The changes made in the netlist are highlighted in red.

50



CHAPTER 4. ADVANCED ECO METHODOLOGY

Figure 4.21: Top Level Netlist showing the I/Os of the MACRO connected (ECO-Netlist)

Step-4: Perform Gate Array ECO process

After creating the ECO netlist, the remaining freeze silicon ECO flow (only till step-5) is
then performed. Figure 4.22 shows the ECO routed design. It highlights the connections
made between the MACRO and the top level design.

Figure 4.22: Final top level design after the MACRO integration

51



CHAPTER 4. ADVANCED ECO METHODOLOGY

Step-5: Formal Verification

The final layout netlist which is written out after the ECO routing is then formally ver-
ified. That is checking whether the layout is functionally equivalent with the expected
outcome (modified top level RTL). This is the most important and time consuming step
in the flow.

The two designs to be compared are shown in Figures 4.23 and 4.24. The former is
the reference design and the latter is the implemented design. If the formal verification
fails, the following steps are looped until successful verification is achieved.

• Modify the final netlist to make it a new ECO netlist.

• Repeat the step-4 of this flow.

• Perform formal verification on the new netlist.

Figure 4.23: Modified RTL of the top
level design (”existing chip”)

Figure 4.24: The final layout netlist
after the complete ECO process

Step-6: Compare GDS using Calibre

Apart from verifying the final layout in terms of functionality, one also has to verify that
only metal layers have been used to perform this change. In order to verify that, the
GDS has to be streamed out and compared. The two designs whose GDS were compared
are shown in Figures 4.25 and 4.26. The original top level design is shown on the left
and the final top level design is shown on the right.

52



CHAPTER 4. ADVANCED ECO METHODOLOGY

Figure 4.25: Original top Level de-
sign with GA filler

Figure 4.26: Final top level design
after the MACRO integration

The stream out of the GDS is done using ICC. But the comparison is done using the
Mentor Graphics Calibre tool. The corresponding command in the latter is compare_gds.
Figure 4.27 shows the output of that command.

Figure 4.27: Compare GDS Results

The keywords mentioned in the compare GDS output can be explained as follows.

53



CHAPTER 4. ADVANCED ECO METHODOLOGY

Layer Shows the different layer numbers. For example,

17 Polysilicon layer

30 Contact layer

31-35 Metal layers

51-54 Via layers

117 Text layer

remaining different transistor masks layers

Geometry Count Shows the amount of changes the tool found in the final top level
design (implementation design), when compared to original top level design (ref-
erence design).

The command compares both the designs and points out the layer differences. It gives a
number showing the amount of differences found. For example, Layer 17 shows Geom-
etry Count of ’0’. This means that no differences have been found with respect to the
polysilicon layer. Thus it can be said that the transistor poly-masks have been reused.
Layer 31 shows Geometry Count of ’536685’. This means that tool has found this many
changes with respect to the that metal layer. This tells that the metal layers in the
implementation design differs from the referenced design. Such type of output conforms
the metal-only ECO.

4.6 Summary

In this chapter, the advanced ECO methodology was proposed and explained in detail.
The major advantage of this methodology is that, it works irrespective of the size of an
ECO. The chapter began by giving the block diagram which divided the methodology
into two parts. Subsequent sections explained individual parts in detail with the help
of an example. The chapter concluded by verifying the output functionally as well as
verifying that only metal layers were used to implement the change. As this methodology
uses GA type cells to realize an ECO, their implementation style is quite similar to the
structured ASICs Therefore, the next chapter will be dedicated to discuss and compare
the advanced ECO methodology with the structured ASIC methodology.

54



5
Structural ASICs vs. Advanced

ECO methodology

The concept used by advanced ECO methodology to perform metal-only ECO is very
similar to the structured ASIC design methodology that prevailed few years back. Metal-
configurable gate array (GA) spare cells are used in the former. Because these cells are
developed by combining GA with the structured ASICs (see section 2.5.1), it is interest-
ing to find out differences and similarities between structured ASICs and the advanced
ECO methodology. This chapter will start by providing the relevant background on
structured ASICs. Section 5.2 will compare the GA design style and the structured
ASICs design style. In the subsequent section the advanced ECO methodology which
uses the GA concept is compared with the structured ASIC methodology. The last sec-
tion is dedicated to discuss both the methodologies from application point of view. At
the end of this chapter the summary is presented. Note that, it is assumed that the
reader knows the basic concepts of Field Programmable Gate Array (FPGA) and the
standard cell based ASIC.

5.1 Relevant background on Structured ASICs

As the FPGA volumes and transistor densities continue to increase,FPGA sales started
to reduce the standard cell based ASIC sales. Thus ASIC vendors came up with a
structure similar to FPGA to lure the FPGA customers back to ASIC. This so called
structure is known as Structured ASIC [20].

5.1.1 Architecture of Structured ASIC

The architecture of structured ASIC falls between standard cell based ASIC and FPGA
[5]. It is made up of two parts.

55



CHAPTER 5. STRUCTURAL ASICS VS. ADVANCED ECO METHODOLOGY

Structured Element Can be combinational and/or sequential logic blocks

Array of Structured Elements The array can be uniformly spread or non-uniformly
spread.

The architecture of the structured ASIC is shown in Figure 5.1 . Structured elements
such as Embedded RAM, Prefabricated I/Os are partially formed elements, which means
that these elements are pre-configured. Whereas sea of tiles is an array of a structured
element called tile, where each tile can be either be partially formed or uncommitted
(like gate arrays). The new functionality is configured using such elements.

Figure 5.1: Architecture of Structured ASICs [5]

5.1.2 Implementation flow

In order to implement a new functionality using the structured ASICs architecture,
following design flow is followed.

Logic synthesis : Mapping the RTL of the new functionality into the structured ele-
ments. For example, if the structured element on the chip is of type combinational
logic then only such kind of cells are used for mapping the RTL.

Placement : Mapping of each structured element in the synthesized netlist onto array
of structured elements on the chip [5]. Once the basic mapping is done, in order
to improve the timing, placement optimization’s are performed.

56



CHAPTER 5. STRUCTURAL ASICS VS. ADVANCED ECO METHODOLOGY

Clock tree Synthesis : Distributing the clock network among the structured elements
with a goal to minimize clock skew and delay.

Routing : Full routing is performed, that is connecting different structured elements
via wires.

5.1.3 Comparison

Table 5.1: Comparison between different types of design styles

cell based ASIC Structured ASICs FPGA

Design Complexity Hard to Design Easy to design Easy to design

Development time Long Short Short

NRE costs Big Small None

Design Size Supports Large Supports Medium Small or Limited

Complexity support Large Medium Limited

Performance High Medium Limited

Man power > 5 > 2 < 5 1 or 2

Price per chip Cheapest Medium Expensive

Unit Cost(High Volume) Low($11) Medium($21) High($40)

Power Consumption Low Medium High

S

It is clear from the table above that structured ASICs combine the best of cell based
ASIC and FPGA.

5.2 Structured ASICs and the Gate Arrays

Comparing the structured ASIC architecture with the GA structure (as in Figure 2.17),
it can be noticed that, the basic concept is similar. While the latter is made up of array
of unconnected MOS transistor unit tile, the former is made up of array of structured
element that could be in the form of tiles. Had the internal structure of a tile been
like unconnected MOS transistor gate, the structured ASIC architecture would become
similar to GA.

For implementation, both the technologies make the use of the prefabricated elements
to implement the new functionality. Thus, the base layers remain the same and only
metal layers are used to configure the functionality. As a result, the time to market and
costs are reduced in both the design styles. From the above similarities it can be said
that the structured ASIC enjoy all the advantages of GA design style. Hence, GA can

57



CHAPTER 5. STRUCTURAL ASICS VS. ADVANCED ECO METHODOLOGY

be considered as forefathers of the structured ASIC technology.

A major difference between both the technologies is that, the structured ASIC also offers
an array of partially formed elements like the ones shown in the architecture. Whereas
in a GA structure, it’s always the array of uncommitted MOS transistors. This advan-
tage makes a structured element also like a FPGA element but without the additional
overhead for field programmability [18].

5.3 Advanced ECO methodology and the structured ASICs

The advanced ECO methodology uses GA type cells to carryout an ECO and thus it
makes the methodology very similar to the structure ASICs. In this section, both the
concepts would be compared to find out similarities and differences. The comparison
will be made with respect to the assumptions of the top level design. These assumptions
are mentioned in section 4.5.1.

5.3.1 Comparison based on Assumption 1

In this assumption, it is assumed that the top level design contains the spare area filled
with GA fillers. Comparing both the methodologies, it can be said that, in case of the
advanced ECO methodology, the GA fillers are used as base cells, whereas, on the other
hand, these would be used as a structured element. Furthermore, in the former case, the
new functionality is configured by swapping the corresponding MACRO with the base
cells. Whereas in the latter case, the structured ASIC implementation flow would be
performed upon the structured element. Because the MACRO is implemented using GA
type cells, the result would still be similar in both the cases i.e. reuse of prefabricated
layers.

Thus from the above comparison it can be concluded that, advanced ECO method-
ology uses the concept of structured ASICs for doing metal-only ECO, but the way of
configuring the new functionality is different. The former uses the MACRO mapping,
whereas the latter uses structured ASIC implementation flow.

5.3.2 Comparison based on Assumption 2

In this assumption it is assumed that the top level design already contains some func-
tionality implemented using GA cells. Adding a new functionality in case of structure
ASICs, would be re-configuring the structured elements by following the structured ASIC
implementation flow. However, in the advanced ECO process, a new MACRO is created
and swapped with the old one.

Note that the size of the new functionality cannot exceed the size of the configurable
area. This is true for both the methodologies. In the structured ASICs, as the complete

58



CHAPTER 5. STRUCTURAL ASICS VS. ADVANCED ECO METHODOLOGY

chip is configurable, the size cannot exceed the chip area, while in the advanced ECO
methodology the size cannot exceed the spare area.

5.3.3 Summary of the comparison

In summary, regardless of top level assumption, if the architecture of structure ASIC is
made up of GA fillers , the whole chip would be filled with these type of cells. But in
the case of advanced ECO methodology, only a small portion of the chip is dedicated
to such type of cells. As the remaining portion is implemented using regular standard
cells, the advanced ECO methodology can be said to be using cell based ASIC flow for
designing major portion of the chip and structured ASIC flow for performing the ECO
in the prefabricated area (spare area).

5.4 Applications Domain

This section will describe the manner in which the semiconductor industry has ap-
proached the concept of structured ASICs in the past. In addition, several ways of
using advanced ECO methodology will also be presented here.

5.4.1 Structured ASIC applications

Few years back, companies like ALTERA, eASIC introduced the structured ASIC tech-
nology to the commercial market with products like Altera’s HardCopyII and eASIC’s
Nextreme. These products offer configuration of several VIA and metal layers to make
the functionality. The following paragraph demonstrates the use of Altera’s Hardcopy
structured ASIC.

The design to be produced was first tested on the FPGA platform to conform the design
features. After getting the approval from the customers, the design was migrated to
structured ASIC platform for production. The customers were able to change the design
on the fly by customizing the chip. Altera provided complete platform for Prototyp-
ing using an FPGA and then production using structured ASIC. For example, Infineon
technologies, prototyped and tested their designs using the FPGAs and then used the
Altera’s Hardcopy structured ASIC to make several versions of the customized designs
according to the customer requirements [21].

5.4.2 Advanced ECO methodology applications

The advanced ECO methodology could be used in following ways.

• Whenever the size of an ECO is quite big, typically 12k gates, then this flow comes
very handy as it makes the use of conventional ASIC flow commands.

59



CHAPTER 5. STRUCTURAL ASICS VS. ADVANCED ECO METHODOLOGY

• For feature swapping. As this flow uses GA type cells, it allows to swap a portion
of the chip implemented using GA type cells with a new feature also implemented
using GA type cells.

Thus one can think of designing a chip containing some features already implemented
using GA cells. So later on, in case of adding a new feature, that portion of the chip can
be swapped out within minimum time and with minimum cost.

5.5 Summary

This chapter presented the structured ASICs technology and discussed the similarities
and differences with the advanced ECO methodology. First section provided the basic
concept of the structured ASICs; Section 5.2 showed that the structured element in the
structured ASICs is similar to a GA tile as well as a FPGA element. Subsequently, the
advanced ECO methodology was compared with the structured ASIC implementation
flow. In the comparison summary, it was concluded that the advanced ECO methodology
is a mixture of cell based ASIC and structured ASIC. The next chapter will present area
and power comparison results for the advanced ECO methodology followed by discussion
of those results.

60



6
Results and Evaluation

Chapter 4 explained the block level flow and the top level flow of the advanced ECO
methodology in detail. The block level flow uses gate array (GA) type cells with con-
ventional ASIC flow in order to create a MACRO of a new functionality. As a result,
this chapter performs the evaluation of using such type of cells in block level flow and
compares it against using regular standard cells