Low Power Solutions: A System Design Perspective

56
LOGO Low Power Solutions: A System Design Perspective Nik Sumikawa

description

Low Power Solutions: A System Design Perspective. Nik Sumikawa. Low Power: Why?. 1. Standard Embedded Solutions. 2. 3. 3. Innovative Solutions. 4. 4. Solutions for Mobile Platforms. Contents. Low Power: Why?. Power vs. Performance Technology Scaling VLSI Embedded Technology Trend - PowerPoint PPT Presentation

Transcript of Low Power Solutions: A System Design Perspective

Page 1: Low Power Solutions: A System Design Perspective

LOGO

Low Power Solutions:A System Design Perspective

Nik Sumikawa

Page 2: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Contents

Low Power: Why?1

Standard Embedded Solutions2

Innovative Solutions33

Solutions for Mobile Platforms44

Page 3: Low Power Solutions: A System Design Perspective

Company Logo

Low Power: Why?Power vs. Performance

Technology Scaling VLSI Embedded

Technology Trend Green Stimulus Scaling Size

www.themegallery.com

Page 4: Low Power Solutions: A System Design Perspective

Company Logo

What You Should Think AboutLow power design strategies

Components: Microcontrollers, peripherals, ect.

Low power design with hardware Low power design with software Low power design in mobile device

www.themegallery.com

Page 5: Low Power Solutions: A System Design Perspective

Company Logo

Low Power Embedded Systems

TELOS: Low power wireless

embedded system Low duty cycle

principle Minimizes dynamic

power consumption

www.themegallery.com

Page 6: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Low Duty Cycle Principle

WakeUp

Process

SleepPrepDeep

Sleep

Sleep Mode

Timer or Interrupt event

Page 7: Low Power Solutions: A System Design Perspective

Company Logo

Low Duty CycleLow processing to sleep ratio

Extended sleep period

Responsively: fast wake-up and sleep times

Minimize Interrupts: Context switching overhead

Nik Sumikawa

Page 8: Low Power Solutions: A System Design Perspective

Company Logo

Low Duty Cycle: DMADirect Memory Access (DMA):

Controls bus and transfers data with minimal processor overhead

Significance Transfer data while sleeping Minimize processor overhead

Nik Sumikawa

Page 9: Low Power Solutions: A System Design Perspective

Low Duty CycleFails with significant

processing

Alternatives: Dynamic Voltage and

Frequency Scaling (DVFS)

Dynamic Power Management (DPM)

Nik Sumikawa

Image: http://www.domainmagnate.com/wp-content/uploads/2009/03/failure-success.jpg

Page 10: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Dynamic Power

DynamicPower

P = CVdd2f

Capacitance

Frequency

Voltage

Energy Source

Battery

Design Variables

Page 11: Low Power Solutions: A System Design Perspective

Company Logo

Reducing Dynamic Power

Dynamic Voltage and Frequency Scaling Scale voltage when sleeping/Idle

Voltage term quad. proportional to power Reduce frequency Minimize line capacitance

Long traces have large capacitance

www.themegallery.com

Page 12: Low Power Solutions: A System Design Perspective

Dynamic Power Management

Generalize power management

Multiple policies Single-policy Multiple-policy Task-scaling

www.themegallery.com

Rajami and Brock [2]

Page 13: Low Power Solutions: A System Design Perspective

Single-policy StrategyIdle Scaling (IS)

Operate at full speed when processing workload

Reduce the frequency and voltage when idleGoal:

Reduce the CPU and bus frequencies Meet continuous DMA requirements Provide acceptable latency when resuming

from idle

Nik Sumikawa

Rajami and Brock [2]

Page 14: Low Power Solutions: A System Design Perspective

Multi-policy StrategiesLoad scaling (LS):

Balance system operating point with current or predicted processing demands

Run system with minimal idle timeOther:

Manage systems state based on status of the systems energy source

Nik Sumikawa

Rajami and Brock [2]

Page 15: Low Power Solutions: A System Design Perspective

Task-scaling StrategiesApplication scaling (AS):

Used for workloads that are difficult to power manage• Audio and video processing• Begin processing next sample immediately

Operate a lower operating point Increases to higher operating point when it

begins to fall behind.

Nik Sumikawa

Rajami and Brock [2]

Page 16: Low Power Solutions: A System Design Perspective

Results of DPM

IS: Idle Scaling LS: Load Scaling AS: Application Scaling

Frame-Scaling (FS): perfect knowledge of processing requirements of video frame

Nik Sumikawa

Rajami and Brock [2]

Page 17: Low Power Solutions: A System Design Perspective

Too Many Low Power States

Disadvantages: Confusion Wrong low power

state

Solution: Minimize the number

of state Decrease complexity

Nik Sumikawa

Image: http://kunaljanu.files.wordpress.com/2009/02/ ist2_1457667confusion-1.jpg

Page 18: Low Power Solutions: A System Design Perspective

Sources of Power ConsumptionMicrocontrollerBus architecture

On chip communication External communication

Memory hierarchyPeripherals

Nik Sumikawa

Rajami and Brock [2]

Page 19: Low Power Solutions: A System Design Perspective

Communication ArchitecturesAdvanced Microcontroller Bus Architecture

ARM bus protocol for system-on-a-chip (SOC) Advanced High Performance Bus (AHB)

• Pipelined• Memory mapped• Up to 16 masters, 16 slaves

Advanced Peripheral Bus (APB)• Non pipelined• Single master, up to 16 peripherals

Nik Sumikawa

Rajami and Brock [2]

Page 20: Low Power Solutions: A System Design Perspective

AMBA On-chip BusNik Sumikawa

Rajami and Brock [2]

Page 21: Low Power Solutions: A System Design Perspective

Power ProfilingNik Sumikawa

Rajami and Brock [2]

86% power consumed by logic14% power consumed by bus lines

Page 22: Low Power Solutions: A System Design Perspective

Power Reduction TechniquesPower Management

Shut down bus interfaces to idle slavesBus Encoding

Reduces # of line transitions, but not bus transactions

Traffic Sequencing Reduce multiple masters interleaving bus

access

Nik Sumikawa

Rajami and Brock [2]

Page 23: Low Power Solutions: A System Design Perspective

Power Reduction TechniquesNik Sumikawa

Rajami and Brock [2]

No technique achieves large saving alone

Page 24: Low Power Solutions: A System Design Perspective

Power vs EnergyPower is amount of energy over an

amount of time (Watts = Joules / second)Battery provides finite amount of energy

Goal: minimize energy use, not just powerIn mobile systems we care about energy

Budget energy to prolong battery life

Nik Sumikawa

Rajami and Brock [2]

Page 25: Low Power Solutions: A System Design Perspective

Static System OptimizationCompiler techniques

Instruction energy consumption profiling• Done empirically

Instruction reordering• Without affecting correctness• Improve register utilization• Reduce memory accesses• Reduce pipeline stalls

Nik Sumikawa

Page 26: Low Power Solutions: A System Design Perspective

Static System OptimizationCode Compression

Post compilation static optimization Reduces storage size of instructions Can have a large impact Requires complex design space exploration Goal for mobile system: reduce power

consumption while preserving performance

Nik Sumikawa

Page 27: Low Power Solutions: A System Design Perspective

Code Compression ChallengesRandom access decompression

Defining decodable block beginnings Jump to new locations in program without

decoding all blocks betweenSolutions

Begin compressed blocks on byte boundaries Store translation table

• More efficient the compression, larger the table Recalculate branch offsets to compressed

addresses

Nik Sumikawa

Page 28: Low Power Solutions: A System Design Perspective

Code Compression RequirementsAdditional hardware

Additional memory to store table Decompression unit

Design decisions Table generation/lookup Compression technique

Nik Sumikawa

Page 29: Low Power Solutions: A System Design Perspective

Code Compression ImplementationSPARC ISAOptimize consumption of complete SOCMultiple iterations on binaryInstructions split into 4 categories

Group 1: immediate instructions (code = 0) Group 2: branch instructions (code = 11) Group 3: dictionary instructions (code = 100) Group 4: uncompressed instr (code = 101)

Nik Sumikawa

Page 30: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

Optimized Binary

CompiledBinary

Update branch offsets

Branch compression

Markov model

Phase 4

Phase 3Immediate compression

Phase 2

Phase 1

Page 31: Low Power Solutions: A System Design Perspective

As a Result…Bus Compaction

Instructions transmitted no longer require entire bus

Use the extra lines to transmit the next compressed instruction

Nik Sumikawa

Page 32: Low Power Solutions: A System Design Perspective

Decompression ArchitecturePre-Cache

Decompression engine between memory/cache

Post-Cache Decompression engine between cache/cpu

Nik Sumikawa

Page 33: Low Power Solutions: A System Design Perspective

SimulationFull SOC simulation7 sample apps run

Nik Sumikawa

Page 34: Low Power Solutions: A System Design Perspective

ResultsNik Sumikawa

Page 35: Low Power Solutions: A System Design Perspective

INCLUDE?Nik Sumikawa

Page 36: Low Power Solutions: A System Design Perspective

ResultsNet energy saving observed

22-82% power savings from code compression

What about additional hardware?Bonus

Increased performance Reduced area

Nik Sumikawa

Page 37: Low Power Solutions: A System Design Perspective

VerdictStatic power optimization

Potentially large payoff for little preprocessingStill more sources of consumption

We’ve observed SOC savings What about peripherals?

Nik Sumikawa

Page 38: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Energy Budget

Voice Call

SMS

Emails

Pictures

localization

Energy Budget

Page 39: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Energy Budget: Localization

How much of the energy budget should be given to localization? Depends on the user

Grant increase allotment when localization is a higher priority

Page 40: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Localizations Methods

1

GPS• Very accurate• Power

Hungry

2

GSM• Lower accuracy• Lower power

requirement

3

WiFi• Mod. Accurate• Mod. Power

requirement

Page 41: Low Power Solutions: A System Design Perspective

Nik Sumikawa

Constandache, Gaonkar, Sayler, Choudhury, Cox [3]

Power vs. Precision

Power: amount of energy required by peripheral in order to determine location

Localization

Precision:Accuracy of the device used for localization

Page 42: Low Power Solutions: A System Design Perspective

Power Consumptionwww.themegallery.com

Constandache, Gaonkar, Sayler, Choudhury, Cox [3]

30 Second sampling intervals

Power Consumption: GPS: High baseline WiFi: Low baseline with high spikes GSM: Low baseline with varying spikes

Page 43: Low Power Solutions: A System Design Perspective

Company Logo

Power Consumptionwww.themegallery.com

30 Second sampling intervals

Results: GPS: increased

baseline

Page 44: Low Power Solutions: A System Design Perspective

Company Logo

Localization Accuracy

Accuracy varied based on location

ALE: Average Location Error

Wifi and GSM oversampled

www.themegallery.com

Page 45: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

Add Your Text

Add Your Text

Add Your Text

Add Your Text

Add Your Text

Add Your Text

Add Your Text

Add Your Text

Page 46: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

Add Your Text Add Your Text

Add Your Text

Page 47: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

Add Your Title

Add Your Title

Add Your Title

ThemeGallery is a Design Digital Content & Contents mall developed by Guild Design Inc.

ThemeGallery is a Design Digital Content & Contents mall developed by Guild Design Inc.

ThemeGallery is a Design Digital Content & Contents mall developed by Guild Design Inc.

Page 48: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

Add Your Title

TextText

Text

Text

Page 49: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Cycle Diagram

Sources

B

E

C

D

AMicrocontroller

Add Your Text

Add Your Text

Add Your Text Add Your Text

Page 50: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

Your TextYour Text

Your TextYour Text

Your TextYour Text

Your TextYour Text

2001 2002 2003 2004

Page 51: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Progress Diagram

Phase 1 Phase 2 Phase 3

Page 52: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Diagram

2003.10 Add Your Text2003.10 Add Your Text2003.10 Add Your Text

2001

2002

2003

2004Company History

2001.10 Add Your Text2001.10 Add Your Text2001.10 Add Your Text

2002.10 Add Your Text2002.10 Add Your Text2002.10 Add Your Text

2004.01 Add Your Text2004.03 Add Your Text2004.05 Add Your Text

Page 53: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Map

Page 54: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Text1

Text2Text3

Text5

Text4

3-D Pie Chart

Page 55: Low Power Solutions: A System Design Perspective

Company Logo

www.themegallery.com

Block Diagram

Add Your Text

Add Your Text

concept Concept Concept Concept

Page 56: Low Power Solutions: A System Design Perspective

LOGO

www.themegallery.com