Embryonics: A New Methodology for Embryonics: A New Methodology for Designing Field-Programmable Gate Arrays Designing Field-Programmable Gate Arrays

with Self-Repair and Self-Replicating with Self-Repair and Self-Replicating PropertiesProperties

Laboratory for Reliable Computing (LaRC)

Electrical Engineering Department

National Tsing Hua University

Daniel Mange, Member, IEEE, Eduardo Sanchez, Member, IEEE, Andre Stauffer,Member, IEEE, Gianluca Tempsti, Member, IEEE, Pierre Marchal,

Member, IEEE, and Christian Piguet

IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEM, VOL. 6, NO. 3, SEPTEMBER 1998

Chan-Chuan Lee

2

OutlineOutline Introduction

The Foundations of EMBRYONICS

System of Ordered Binary Decision Diagrams(OBBD)

A New Field-Programmable Gate Array Based on a Multiplexer Cell

Cellular differentiation: Genome Interpretation

Cellular Division: Duplication of The Genome

Self-Replication and Self-Repair Properties

Conclusions

3

Introduction(1/2)Introduction(1/2) The growth and the operation of all living beings

are directed through the interpretation, in each of their cells, of a chemical program, the DNA string or genome.

Any logic system can be represented by an order binary decision diagram(OBBD), and then embedded into a fine-grained field-programmable gate array(FPGA) .

The cellular array thus obtained is perfectly homogeneous: the function of each cell is defined by a configuration(or gene) and all the genes in the, each associated with a pair of coordinates, make up the genome of the articial organism.

4

Introduction(2/2)Introduction(2/2) The interpreter extracts from the genome, the

gene of a particular cell as a function of its position in the array.

Self-Replication(the automatic production of one or more copies of the original organism)

Self-Repair (the automatic repair of one or more faulty cells)

5

Objectives and StrategyObjectives and Strategy Develop very large scale integrated (VLSI)

circuits capable of self-repair and self-replication.

Self-replication allows complete reconstruction of the original device in case of a major fault while self-repair is opposite.

Order binary decision diagram(OBBD) greatly simplify the realization of a new family of FPGA’s, based a fine-grain cell. This cell is called MUXTREE.

6

The Foundations of Embryonics(1/2)The Foundations of Embryonics(1/2) The general hypothesis about the environment

First feature: Multicellular Organization

7

The Foundations of Embryonics(2/2)The Foundations of Embryonics(2/2) Second feature: Cellular Differentiation.

Third feature:Cellular Division.

8

Synthesis of Ordered BinarySynthesis of Ordered Binary OBBD is a graphical representation which

exploits well the 2-D space and immediately suggests a physical realization on silicon.

OBBD lead us to a natural decomposition into cells realizing a logic test, easily implemented by a multiplexer.

9

Modulo-4 up-down CounterModulo-4 up-down Counter An example for this

10

Modulo-4 up-down Counter(1/2)Modulo-4 up-down Counter(1/2)

11

Modulo-4 up-down Counter(2/2)Modulo-4 up-down Counter(2/2)

12

Hardware ImplementationHardware Implementation The goal is to implement directly the ordered

binary decision diagram on silicon.

Replace each test elements with 2-to-1 multiplexer.

The two state functions Q1+ and Q0+ are available at the outputs of the top multiplexers.

13

A New Field-Programmable Gate A New Field-Programmable Gate Arrays Based on A Multiplexer CellArrays Based on A Multiplexer Cell

Each of the two inputs of the multiplexer(labeled “0” and “1”) will be programmable.

The output of the mux will be, therefore, connected to the inputs of the muxs in the neighboring cells to the north, northeast, and northwest.

Sequential systems require the presence, in each cell, of a synchronous memory element, a D-type flip-flop.

14

MUXTREE CellMUXTREE Cell

Fig-(a) : Detailed architecture

Fig-(b) : The 20-bit data GENE 19:0 with P=PRESET, R=REG, and EB = EBUS.

15

The Switch Block SBThe Switch Block SB Fig-(a) :

Interconnection possibilities.

Fig-(b) : Detailed Architecture.

16

Nine-Cell Implementation of The Up-Nine-Cell Implementation of The Up-Down CounterDown Counter

Fig-(a) : logic Level

Fig-(b) :bus level

Use two D-type flip-flops,generates the variables Q1 and Q0 in place of Q1+ and Q0+.

17

Hierarchical Overview of the Three Hierarchical Overview of the Three LayersLayers

For the sake of simplicity, decompose it in three components.

Memory stores a single gene per address.

18

Microprogrammed RealizationMicroprogrammed Realization Use a microprogram to compute the local

coordinates X and Y and to extract from our artificial genome.

Up-down counter can be considered as a truth table whose input are coordinates or addresses X and Y and whose output are genes GENE 19:0

Express coordinates X and Y in pure binary code, using the logic variables X1, X0, Y1, Y0.

19

Computation and Genome Computation and Genome RepresentationRepresentation

Fig-(a) : Gene computation

20

Computation and Genome Computation and Genome RepresentationRepresentation

Fig-(b) X coordinates computation.

Fig-(c) Up-down counter genome.

21

NANOPASCAL : A High-Level NANOPASCAL : A High-Level Language Language

Define a programming language well suited for the description, the interpretation and the duplication of the genome.

22

NANOPASCAL language NANOPASCAL language Fig-(a) : Syntactic diagram

23

NANOPASCAL languageNANOPASCAL language Fig-(b) : Microprogram GENOME

24

NANOPASCALINE : An Interpreter for the NANOPASCAL NANOPASCALINE : An Interpreter for the NANOPASCAL LanguageLanguage

Detailed architecture with format and operation code (OPC) for the six instruction of the language

25

Cellular Division : Duplication of the Cellular Division : Duplication of the GENOMEGENOME

The duplication of the GENOME microprogram is accomplished automatically, in parallel with its interpretation.

The GENOME microprogram is thus duplicated in permanence, resulting in a great simplicity of excellent wiring and reliability.

Since an eventual transient fault(copy error)during a cycle will be corrected in the next cycle.

26

Properties of the Up-Down CounterProperties of the Up-Down Counter

Fig-(a) Self-replication

Fig-(b) Self-repair

27

BIODULE: Demonstration Artificial BIODULE: Demonstration Artificial Digital CellDigital Cell

28

BIODULE : Demonstration Artificial BIODULE : Demonstration Artificial Digital Cell Digital Cell

Detailed architecture

29

ConclusionsConclusions The result of the paper is the development of a

new family of FPGA’s called MUXTREE.

Self-repair and self-replication are easy to realize.

Future perspectives. The main drawback of the BIODULE cell is the lack

of balance between MUXTREE. To develop a new coarse-grained FPGA.