1 of 16April 25, 2006

Minimizing System Modification in anIncremental Design Approach

Paul Pop, Petru Eles, Traian Pop, Zebo PengDepartment of Computer and Information Science, Linköpings universitet, Sweden

Vers

ion

N+

1V

ers

ion

N+

1

NN

Modify(re-map)so that thecurrent applicationswill fit.

Map and scheduleso that thefuture applicationswill have a chanceto fit.

Do not exist yetat Version N!

Modify(re-map)so that thecurrent applicationswill fit.

Map and scheduleso that thefuture applicationswill have a chanceto fit.

Do not exist yetat Version N!

N-1

N-1

Incremental Design

Start from an already existing system with applications

Implement new functionality on this systemMapping and Scheduling

To reduce design and testing time:As few as possible modifications of the existing applications

After the new functionality has been implemented:It should be easy to add functionality in the future

Mapping and scheduling of distributed embedded systems for hard-real time applications Static cyclic scheduling of processes and messages,

Bus access scheme: time-division multiple-access.

Incremental design process Already existing system,

Implement new functionality,

a) Existing system modified as little as possible, b) new functionality can be easily added to the system.

Mapping strategy a) Subset selection to minimize modification time,

b) Two design criteria, objective function.

Summary

Input A set of existing applications. A current application to be mapped. The system architecture.

Output

A mapping and scheduling of the current application, so that the incremental design requirements are satisfied.

Requirements

a) constraints of the current application are satisfiedand minimal modifications are performed to the existing applications.

b) new future applications can be mapped on the resulted system.

Problem Formulation

Initial mapping and scheduling

Requirement a) Subset selection problemSelect that subset of existing applications so that the current application fits and the modification cost R() is minimized:

Three approaches to the subset selection problem

Exhaustive Search (ES)

Ad-Hoc Solution (AH)

Subset Selection Heuristic (SH)

Requirement b) Objective function minimization:

Mapping Strategy

Characterizing future applications: Typical WCETs Typical message sizes Smallest expected period Tmin

Expected necessary processor time tneed Expected necessary bandwidth bneed

Characterizing existing applications:

1 2 4 5

6

7

3

8 9 10

150 70 70

50

50

20

),0max(),0max()()( 22221111m

needmP

needPmmPP CbwCtwCwCwC

0

200

400

600

800

1000

1200

320 400 480 560 640

AHSHES

0

20

40

60

80

100

120

140

320 400 480 560 640

AHSHES

Avera

ge M

odifi

cati

on C

ost

R()

Number of processes

Avera

ge E

xecu

tion T

ime [

min

]

Number of processes

Perc

enta

ge o

f fu

ture

applic

ati

on

s [%

]

Number of processes in the current application

0

20

40

60

80

100

120

40 80 160 240

MS*

AD

I/O Interface

Comm. Controller

CPU

RAM

ROM

ASIC

Node

S0 S1 S2 S3 S0 S1 S2 S3

TDMA Round

Cycle of two rounds

Slot

Design Criteria

Paul Pop, Petru Eles, Traian Pop, Zebo Peng:An approach to Incremental Design ofDistributed Embedded Systems,Design Automation Conference, 2001

First criterion: C1P, C1

m How well the resulted slack sizesaccommodate a future application

min(40, 80, 0) = 0msmin(40, 0, 80) = 0msmin(80, 80, 40)

CP2= 40ms

Periodic slack

N1N2N3

Bus

min(40, 40, 40) = 40msmin(40, 40, 40) = 40msmin(80, 80, 40) = 40ms

CP2 = 120ms

Period 0Round 0 Round 1Round 2 Round 3

Round 0 Round 1Round 2 Round 3

Tmin

Period 1 Period 2

360 ms

S1 S2 S3 S1 S2 S3 S1 S2 S3 S1 S2 S3

S1 S2 S3 S1 S2 S3 S1 S2 S3 S1 S2 S3

N1N2N3

Bus a)

b)

P1

= 40ms

Second criterion: C2P, C2

m

How well the slack is distributed in time toaccommodate Tmin, tneed and bneed

a)

Existing Current

Future:

Slack

P1 P2

b)c)

i

iRR )(

Futureapplications

Futureapplications

Currentapplications

Currentapplications

Existingapplications

Existingapplications