“Self-Resetting Stage Logic”

33
“Self-Resetting Stage Logic” Presented by : Vishal Jain, ID: 200411029 Guided By: Prof. Anutosh Maitra

description

“Self-Resetting Stage Logic” Presented by : Vishal Jain, ID: 200411029 Guided By: Prof. Anutosh Maitra. What is SRSL? Synchronization approach Data flow clock less designs Uses single rail encoding. Why new thing? - PowerPoint PPT Presentation

Transcript of “Self-Resetting Stage Logic”

Page 1: “Self-Resetting Stage Logic”

“Self-Resetting Stage Logic”

Presented by : Vishal Jain, ID: 200411029 Guided By: Prof. Anutosh Maitra

Page 2: “Self-Resetting Stage Logic”

What is SRSL?1. Synchronization approach 2. Data flow 3. clock less designs 4. Uses single rail encoding.

Why new thing?SRSL is a simple, efficient and easy to

implement therefore it is acceptable.

Page 3: “Self-Resetting Stage Logic”

What has been implemented and what is going on in SRSL? >16 stage four bit linear pipe (April ,2004 by Abdel Ejnioui, University of Central Florida) > RAM cell design had been tested successfully > Trying to implement for larger network.

Page 4: “Self-Resetting Stage Logic”

2. Basic Circuit – 1 Bit1. Introduction

3. Basic Circuit – Multi Bits4. Pipeline : Linear 4-Stages

6. Analysis5. Nonlinear : Fork and Join

Contents:

7. SRSL in Multiagent Model

Page 5: “Self-Resetting Stage Logic”

Introduction

Self Resetting stage logic:A pipeline stage resets itself before starting a new execution cycle.

Two Phases1. Evaluation Phase : when it executes the instruction and output

is result of evaluation of its input2. Reset Phase : When output is Null

Data flow

Evaluate phase

Reset phase

Page 6: “Self-Resetting Stage Logic”

2. SRSL for ONE BIT

Every Long journey starts with Small Step

Page 7: “Self-Resetting Stage Logic”
Page 8: “Self-Resetting Stage Logic”

1

Page 9: “Self-Resetting Stage Logic”

1

Page 10: “Self-Resetting Stage Logic”

1

Page 11: “Self-Resetting Stage Logic”

1

Page 12: “Self-Resetting Stage Logic”

1

Page 13: “Self-Resetting Stage Logic”

1

0

Page 14: “Self-Resetting Stage Logic”

0 0

Page 15: “Self-Resetting Stage Logic”

0 0

Page 16: “Self-Resetting Stage Logic”

1

Page 17: “Self-Resetting Stage Logic”

3. Extending SRSL for multi bits

One and one makes eleven

Page 18: “Self-Resetting Stage Logic”
Page 19: “Self-Resetting Stage Logic”

4. Linear Pipelining with SRSL

Let us do some more interesting thing

Page 20: “Self-Resetting Stage Logic”

Principle of Linear PipeliningLet T be a task which can be partitioned Into k subtasks (aka stages)according to linear precedenceRelation: T={T1,T2,…,Tk};A subtask Tj can not start until {Ti i<j} areFinished.This can be modeled with the linear precedence graph:

T2T1 Tk

Page 21: “Self-Resetting Stage Logic”
Page 22: “Self-Resetting Stage Logic”

Can you help me if I am a pipeline with Fork and Join ?

5. Fork and Join operations in pipeline

Page 23: “Self-Resetting Stage Logic”

Fork and Join of DATA streams:1. Fork :

2. Join :

A

B

F

E

D

C

Page 24: “Self-Resetting Stage Logic”
Page 25: “Self-Resetting Stage Logic”
Page 26: “Self-Resetting Stage Logic”

6. Analysis of Linear Pipeline

Page 27: “Self-Resetting Stage Logic”

Notations:d( En) = time of Execution phase of nth staged( Rn) = time of Reset phase of nth stage Pn= time of Stage n

Observations because of Invert-And delay 1. d(En)=d(Rn)=Pn/22. d(Rn-1) < d(Rn)3. d(En-1) > d(En)

Page 28: “Self-Resetting Stage Logic”
Page 29: “Self-Resetting Stage Logic”

What we observe from graph??We can’t go beyond a certain limit of cascading because evaluation reset phase time decreases and here comes limitation of SRSL. Therefore it is not Scalable in present scenario.

Page 30: “Self-Resetting Stage Logic”

7. How SRSL may be benificial for Multiagent Model?

Page 31: “Self-Resetting Stage Logic”

Noval Idea:I suggest use this mechanism to implement Data Join in “Hierarchical Multiagent Model” and present possible advantages by using this model in hierarchical Multiagent Model:

a b

c d

e

The key ideas behind the SRMM are1. “Whenever Data is ready it is delivered to the higher level Agent. Without any signaling protocol which saves time.”2. “The Higher level agent always get latest data.”

Page 32: “Self-Resetting Stage Logic”

References:

1. G.Jung , V.Sudharajan and G.E. Sobelman “ A self resetting

32-bit parallel adder”.

2. “Parallelism Scalability and Programmability” Kai Hwang

3. “Multistage Modeling” search results of Google.com

4. Lecture Notes , Prof. Anutosh Maitra

Page 33: “Self-Resetting Stage Logic”

Thanking You