© imec 2001 arrm’01, oct.17 managing dynamic concurrent tasks in real-time multi-media systems...

© imec 2001ARRM’01, Oct.17

Managing dynamic concurrent tasksin real-time multi-media systems

Francky Catthoor, IMEC, Belgium


Goal of our research

• A methodology to map dynamic and concurrent real-time applications on an embedded multi-processor platform


MPEG4JPEG

Why are Applications becoming more dynamic and concurrent?

The workload decreases but the tasks are dynamically created and their size is data dependent

T1T1’ T1

T2T3

T4


MPEG-4 : multimedia spec= too huge to handle as 1 task=> break up in many interacting tasks

With this code my boss has to give me a raise!!!

TASK1

h

Cost(P)

Exec.Time

Divide-and-conquer approach of today leads to local solutions


This didn’t look so good after all???

TASK1

TASK2

TASK3

Processor 1

Global picture: ad hoc

h

h

h

t1

t2

t3

t1+t2+t3<T


TASK1

TASK2

TASK3

Processor 1

Global trade-offs with cost-performance curves

h

h

h

Luckily we have the Pareto approach!

x

x

x

t1n

t2n

t3n

t1n+t2n+t3n<T


Outline

• Motivation: new challenges in system-level design

• Overview of TCM methodology

• Cost-efficient design-time scheduling


• Maximize Quality• Minimize Power

Limited Resources

Terminal QoS requires new approach

CPU

Bus access

Mem. size

Terminal

?

Solution:Paretocurves

3D object

Videoobject

Tasks

See session Multi-media 1


0

100

200

300

400

500

600

700

800

900

1000

0 50 100 150 200 250 300 350 400 450 500 550

Frame number

Fra

me

per

iod

(m

s)

& P

roje

cted

Su

rfac

e (#

pix

els/

200)

proj. surface (a)

measured period (b)

predicted period (c)

600

Very dynamic behaviour soworst-case realisation too costly


System design issues in IT-Application domain

IN622 Mb/s

OUT

OUT

622 Mb/s

622 Mb/s

53 cycles Stringent real-time constraints

Embedded system => cost

Complex data sets• Large and irregular dynamically allocated data• Huge memory accesses

RoutingRecord

PacketRecord

FIFO

200 accesses

data in

ISR

time

data outrouting reply

out

Processes• Dynamic and concurrent processes• Global/local control• Non-deterministic events

Network layer protocols (ATM, IP)Dynamic multi-media algorithms (MPEG4/7/21)Wireless/wired terminals (Internet, WLAN)


Executive

Decoders

Presenter

Application

DataChannel

ALManager

FlexDemux

Service

DataChannel

DataChannel

DataChannel

DecodingBuffer

CompositionMemory

BIFSCritical

path

OD

...

30 msec

RenderingPre/post

Decoding

Pre/post writing

Real-time constraints, tasks and data in the IM1-MPEG4 protocol


Why aren’t multi-processor platforms used now in embedded

domains?

Live from FZ-TV

• efficient mapping requires a very high design effort when done manually

• => need for a cost-sensitive real-time system compiler


Outline





Requirements of system level design approach

Algorithms Data Structures+

ARM

IP1 IP2

RAM ROM

Architecture

Processor architecture

RAMRAM

ROM

MMU

custom logic

DSP

ROM microprocessor

TCM approach:ICPP’00, Kluwer book’99


Static task scheduling

cost

time

task1cost

time

task2cost

time

task3

platform

Real-timeconstraints

Run-time scheduler

processor1

processor2

processor3

time

platformReal-timeconstraints

C/C++ specification of dynamic concurrent system

Extraction of the gray-box model

Task-level DTSE

Concurrency improving transformations

Memory architecture

Real-timeconstraints


Outline





Reduce global system energy by task scheduling + assignment (e.g. 2-processor approach )

ARM

Processor

1Vdd=1V Vdd=3.3V

ARM

Processor

2

Taskn

Task2

Task1

Codes’01, J.of Sys.Arch, summer 2001


Trade-off between time budget (period/latency) and cost (e.g.energy) leads to Pareto curves

Time

Cost

CB1CB2CB3CB4CB5CB6

Processor alloc/assign and scheduling alternatives

for code version 1

xx

x

xNon-optimal points


TCM transformations on applicationmodel shift Pareto curve to origin

Bottleneckof case 1

Schedulingalternativesfor case 1

• Transformations shift the Pareto Curve– Decrease cost for same Time-Budget– Increase performance for same Cost-Budget

New curvefor case 2

TCM trafo

Energy

Time Budget


Comparison for original and transformed IM1 graphs on 2 processors with different Vdd

original

Transformed


Comparison between different processor platforms

Combination 2

Combination 3

(Global Pareto curve)

Global Pareto curve

2 13

45

6

6

54

32

1


Comparison of genetic algorithm (GA)

and heuristic task scheduling results

4400

4600

4800

5000

5200

5400

5600

5800

Energy-cost

37

9

38

1

38

3

38

5

38

8

39

1

39

6

40

1

42

7

43

4

Time-budget

Energy-cost vs time-budget curve

GA solutions

Heuristic


Overall solution: combination of design- and run-time schedulers

12

3

th read fram e 1

A B

th read fram e 2

Cases’00, Design&Test- Sep.’01

1 3 2

Design-time Scheduling

Design-time Scheduling

A B

• Design-time scheduling: at compile time, exploring all the optimization possibility

Run-time

Scheduling

1 A B 3 2

• Run-time scheduling: at run time, providing flexibility and dynamic control at low cost as part of synthesized RTOS


Main messages

• Embedded multi-media applications are becoming very dynamic and concurrent in nature => RTOS essential

• Task Concurrency Management approach provides the flexibility and optimization possibility while limiting the run time computation complexity

• A multiprocessor platform with different working voltages potentially provides an energy saving solution

• Application-specific run-time scheduling technique combined with design-time scheduling to provide cost-performance Pareto-curve essential for effective solution

© imec 2001 arrm’01, oct.17 managing dynamic concurrent tasks in real-time multi-media systems...

Documents

wlan slide

belgium slide

isr time data

accesses data

kluwer book99 slide

dynamic concurrent tasks

time divide

task1 task2 task3 processor