EE249 Embedded System Design: Models, Validation and...
Transcript of EE249 Embedded System Design: Models, Validation and...
-
EECS249Fall08
EE249 Embedded System Design:
Models, Validation and Synthesis
Alberto Sangiovanni Vincentelli
1
-
EE249Fall082
Carly Fiorina, CEO, Hewlett Packard Corporation
I believe we are now entering the Renaissance phase of the Information Age, where creativity and ideas are the new currency, and invention is a primary virtue, where technology truly has the power to transform
lives, not just businesses, where technology can help us solve fundamental problems.
-
EE249Fall083
Client
Server
eMerging Societal-Scale Systems
New System ArchitecturesNew Enabled ApplicationsDiverse, Connected, Physical,
Virtual, Fluid
Embedded Systems
Clusters
Massive Cluster
Gigabit Ethernet
MEMSBioMonitoring
Scalable, Reliable,Secure Services
InformationAppliances
PresenterPresentation NotesWhat is common to all of these efforts is the notion of a Societal-Scale Information System. These are systems with architectures that must span a very wide range of needs, from thousands of sensors in a home, in a business, or on a freeway, to global-scale information utilities, moving pedabytes of data per second throughout the world. These new architectures must have common notions of security, reliability, privacy and robustness across this entire spectrum. They must connect diverse, unreliable, and dynamic sources and consumers of information, in a virtual, almost fluid way.
-
EE249Fall084
Embedded Systems
Computational but not first-and-foremost a computer
Integral with physical processes sensors, actuators
Reactive at the speed of the environment
Heterogeneous hardware/software, mixed
architectures Networked
shared, adaptive
Source: Edward A. Lee
-
EE249Fall085
Observations
We are on the middle of a revolution in the way electronics products are designed
System design is the key (also for IC design!)
Start with the highest possible level of abstraction (e.g. control algorithms)
Establish properties at the right level
Use formal models
Leverage multiple scientific disciplines
-
EE249Fall086
Course overview
Managing Complexity
Orthogonalizing Concerns
Behavior Vs. Architecture
Computation Vs. Communication
-
EE249Fall087
Behavior Vs. Architecture
SystemBehavior
SystemArchitecture
Mapping
Flow To Implementation
Refinement
1
3
4
2Models of
Computation
Comm. and comp. resources
Assign functionality to arch elements
HW/SW partitioning,Scheduling
Synthesis: HW and
SW
Quantity estimation
Polis (1990-1996)VCC (1996-2003)Metropolis (2003-present)
-
EE249Fall08
ETROPOLIS
8
Behavior Vs. Communication
Clear separation between functionality and interaction model
Maximize reuse in different environments, change only interaction model
-
EE249Fall089
Administration
Office hours: Alberto : Tu-Th 12:30pm-2pm or (better) by appointment (2-4882)
Teaching Assistant:
Kelvin Lwin, [email protected]
-
EE249Fall0810
Grading
Grading will be assigned on:
Homework (~30%)
Project (~50%)
Reading assignments (~10%)
Labs (10%)
Bi-weekly homework.
HW #n is due the same day HW #n+1 is handed out
-
EE249Fall0811
Schedule
Labs (Th. 4-6):
Presentation of tools followed by hands-on tutorial and assignments (to turn in after 2 weeks, we might have to skip some labs.)
Discussion Session (Tu. 5-6)
Each student (possibly in groups of 2 people) will have to make one or more oral presentations during the class
Last two weeks of class dedicated only to projects (usually due the 1st or 2nd week of Dec.)
Auditors are OK but please register as P-NP (resources are assigned according to students)
HW #1handout
HW #1due HW #3
handout
Lab #1
Lab #1due
1 week
Lab #2
Lab #2due
Lab #3 Lab #4HW #2
due
Schedule is tightDont fall behind!!!
-
EE249Fall0812
Links
Class website
http://inst.eecs.berkeley.edu/~ee249
http://inst.eecs.berkeley.edu/~ee249
-
EE249Fall0813
Outline of the course
Part 1: Introduction Design complexity, Example of embedded systems, traditional design flow, Platform-Based Design
Part 2: Functional modeling,
analysis and simulation
Introduction to models of computation. Finite State Machines and Co-Design Finite State Machines, Kahn Process Networks, Data Flow, Petri Nets, Hybrid Systems. Unified frameworks: the Tagged Signal Model, Agent Algebra
Part 3: Architecture and
performance abstraction
Definition of architecture, examples. Distributed architecture, coordination, communication. Real time operating systems, scheduling of computation and communication.
Part 4: Mapping Definition of mapping and synthesis. Software synthesis, quasi static scheduling. Behavioral synthesis. Communication Synthesis and communication-based design
Part 5: Verification Validation vs Simulation. Verification of hybrid system. Interface automata and assume guarantee reasoning.
Part 6: Applications Automotive: CAN,Flexray, Auotosar Architecture, GM car architecture, scheduling and timing analysis Building automation: BanNet, LonWorks, ZigBee with applications to monitoring and security
-
EE249Fall0814
Outline for the Introduction
Examples of Embedded Systems
Their Impact on Society
Design Challenges
Embedded Software and Control
-
EE249Fall0815
Electronics and the Car
More than 30% of the cost of a car is now in Electronics90% of all innovations will be based on electronic systems
-
EE249Fall0816
Automotive IndustryThree Levels of Players
Source: Public financials, Gartner 2005
2005 revenue $17.4B CAGR 10% (2004-2010)
IC Vendors~15% of revenue from automotive
2004 Revenue ~$200B CAGR 5.4% (2004-2010)
Tier 1 Suppliers90%+ of revenue from automotive
Automakers 2005 Revenue: $1.1T CAGR 2.8% (2004-2010)
PresenterPresentation NotesThe global auto industry is a $1T USD marketplace, led by companies like GM, Toyota and DaimlerChrysler
They produced 64M cars last year, but only have 2.8% CAGR through 2010. And the growth would be even lower if it werent for emerging markets like China and Eastern Europe
All the major automakers derive the majority of their revenue from car vehicle sales. Some get revenue by financing but the vast majority come from car sales
http://www.st.com/stonline/index.htmhttp://www.delphi.com/http://www.toyota.co.jp/en/index.htmlhttp://www.vw.com/index.htmlhttp://www.gm.com/http://www.ford.com/en/company/about/brands/ford.htm
-
EE249Fall08
ABS: Antilock Brake SystemACC: Adaptive Cruise ControlBCM: Body Control Module
DoD: Displacement On DemandECS: Electronics, Controls, and Software
EGR: Exhaust Gas Recirculation.GDI: Gas Direct Injection
OBD: Onboard DiagnosticsTCC: Torque Converter Clutch
PT: PowertrainForefront of Innovation
Vehicle Integration
System Connection
Subsystem Controls & Features
Valu
e fr
om E
lect
roni
cs &
Sof
twar
e
-More functions & features-Less hardware
-Faster Potential inflection point. Now!
1970s 1980s 2000s 2010s 2020s
Source: Matt Tsien, GM
1990s
BCM
ABS
TCC
EGR
Electric Fan
Electric Ignition
Fuel Cell
Wheel Motor
Hybrid PT
Electric Brake
DoD
GDI
ACC
Rear Vision
Passive Entry
Side Airbags
Head Airbags
OnStar
OBD II
HI Spd Data
Rear aud/vid
CDs
Challenge: Electronics, Controls and Software Shifting the Basis of Competition in Vehicles
$118
2 (+
196%
)
50 E
CU
s (+
150%
)
100M
Lin
es o
f Cod
e (+
9900
%)
$400
20 E
CU
s
1M L
ines
AVG. AVG.
Mechanical $55%
Software $13%
Other $8%
Electronics $24%
Software $2%Other $
9% Electronics $13%
Mechanical $76%
PresenterPresentation NotesPP comments: how was the ECS % established? (base or fully loaded or avg?) LDW, RW might ask how this compares to the numbers given by Toyota which talk about 50%+ on average across their fleet (Toyota inside article which was translated for LDW and RW)- what is the difference between / definition of system connection and vehicle integration? Japanese and European OEMs label the period of 201x as vehicle cooperation , I.e. v2v, v2i, etc. -----------------------------------------Sequence of presentations: Introduction to the technology trend. Features. The customers perspective of ECS growth. (One click) The potential inflection point and the forefront of innovation. (One click) The percentage cost perspective of ECS growth. Two average cars. One in 2000, the other in 2010. (One click) The electronics and software content perspective of ECS growth. Per average car.
-
EE249Fall0818
GM SAC Vehicular Electronics, Controls and Software Study
Software content in automobiles could increase by 100 X over the next 5-6 years. Challenges will include:
Software system architecture
Partitioning for modularity & system reliability
Reuse
Standardization of interfaces
-
EE249Fall0819
-
EE249Fall0820
-
EE249Fall0821
OTIS Elevators
1. EN: GeN2-Cx 2. ANSI:Gen2/GEM
3. JIS: GeN2-JIS
-
EE249Fall0822
Segments
Attribute Type 1 Type 2 Type 3
Stops/Rise < 20 stopsOpportunity: < 6 stops (20m)
< 64 stops < 128 stops
Group Size Simplex 1 8 cars 1 8 cars
Speed< 4m/s
-
EE249Fall0823
-
EE249Fall0824
Consumer segmentsCommon technology elements
Gaming
Auto electronics
Internet
e-commerce
Locality
Broadcasting
Systems within systems
PresenterPresentation NotesIf you look at the growing consumer markets, you can see a lot of common characteristics.Most of these applications combine audio, video, and voice along with wireless or wireline connectivity.
In order to deliver these capabilities, you see a growing trend towards multimedia processors, connectivity, mixed-signal, analog, and RF content in design. Essentially, companies are delivering what used to be disparate technologies in single devices making what you design a problem of developing systems within systems.
-
EE249Fall0825
Common Situation in Industry
Different hardware devices and architectures
Increased complexity
Non-standard tools and design processes
Redundant development efforts
Increased R&D and sustaining costs
Lack of standardization results in greater quality risks
Customer confusion
-
EE249Fall0826
Outline for the Introduction
Examples of Embedded Systems
The Future of Embedded Systems and Their Impact on Society
Design Challenges
Embedded Software and Control
-
EE249Fall08
27
Concurrency and Heterogeneity
IntelMontecito
Source: Bosch
InformationSystems
Tele
mat
ics
Fau
lt
Tole
ran
t
Body Electronics B
ody
Fun
ctio
ns
Fail
Safe
Fau
ltFu
nct
ion
al
Body Electronics
Dri
vin
g an
d V
ehic
leD
ynam
ic F
un
ctio
ns
Mobile Communications Navigation
FireWall
Access to WWWDAB
GateWay
GateWay
Theft warning
Door Module Light Module
AirConditioning
Shift by
Wire
EngineManage-
ment
ABS
Steer by
Wire
Brake by
Wire
MOSTFirewire
CANLin
CANTTCAN
FlexRay
Today, more than 80Microprocessors and
millions of lines of code
-
EE249Fall08
28
Ubiquitous Sensor Networks
Challenge: The Physical Internet
Year
Log
(peo
ple
per c
ompu
ter)
Number CrunchingData Storage
ProductivityInteractive
Mainframe
Minicomputer
Workstation
PC
Laptop
PDA
Cellular phone Streaming information to and from physical world
28
PresenterPresentation NotesSo if we look at the computing spectrum today, each of the classes exist simultaneously. In a room we can put 100s of teraflops and many petabytes of computing.
Our productivity, document preparation, and personal information management fits in our pocket and a new class is emerging that will provide a means of streaming information to and from the physical world like we have never seen before.
-
EE249Fall08
Exponentials Bound to Continue
5 Billion people to be connected by 2015 (Source: NSN)
The emergence of Web2.0
The always connected community network
7 trillion wireless devices serving 7 billion people in 2017(Source: WirelessWorldResearchForum (WWRF))
1000 wireless devices per person?
[Courtesy: Niko Kiukkonen, Nokia]
EE Times,January 07, 2008
-
EE249Fall08
Infrastructuralcore
Sensory swarm
Mobileaccess
The Emerging IT Scene
-
EE249Fall08
The Technology Gradient: Computation
Driven by Moores Law
Driven by More Than Moore andBeyond Moore
-
EE249Fall08
The Technology Gradient: Communication
Mostly wired
Almost uniquely wireless
-
EE249Fall08
Challenge: Power
33
-
EE249Fall08
Scaling enabled integration of complex systems with hundreds of millions of devices on a single die
Intel KEROM dual coreISSCC 07, 290M trans.
SUN Niagara-2ISSCC 07, 500M trans.
IBM/Sony Cell ISSCC 05, 235M trans.
Challenge: Parallel Architectures
34
-
EE249Fall0835
Smart DustLaser diodeIII-V process
Passive CCR comm.MEMS/polysilicon
Active beam steering laser comm.MEMS/optical quality polysilicon
SensorMEMS/bulk, surface, ...
Analog I/O, DSP, ControlCOTS CMOS
Solar cellCMOS or III-V
Thick film batterySol/gel V2O5
Power capacitorMulti-layer ceramic
1-2 mm
Source: K. Pister, Berkeley
PresenterPresentation NotesThe majority of the volume of the mote will consist of power supply elements. The rechargeable thick film battery will store 1 J/mm^3, with low output resistance for sub-milliAmp current. The multi-layer ceramic power capacitor (COTS) will store only 1 microJoule, but will provide high current when needed (e.g. for laser pulses). In full sun, solar cells provide roughly 1 J/day per square millimeter. Different versions of dust may have only battery, only solar cell, or some combination.The sensor will be research off the shelf. The only design changes will be to optimize performance at low power. Sensors options include temperature, pressure, humidity, vibration, sound (requires several cubic millimeters), and magnetic anomaly detection.Communication will be optical, line of sight. The corner cube is used to modulate light from an external interogating laser. We have demonstrated 150 meter communication using corner cubes. 1km range in full sun should be possible. Using the on-board laser and beam steering, active optical communication should be possible over distances of tens of kilometers at least.
-
EE249Fall0836
Wireless Sensor Networks
Berkeley Dust Mote1
1From Pister et al., Berkeley Smart Dust Project
The use of wireless networks of embedded computers could well dwarf previous milestones in the information revolution - National Research Council Report: Embedded, Everywhere, 2001.
Berkeley Mote1
-
EE249Fall0837
Creating a Whole New World of Applications
From Monitoring To Automation
PresenterPresentation NotesLight, temp, humidity, pressure, presense of birds, some sensors in burrows, some outside(SHOW DEVICE)First deployment 6 months ending Nov 2002, Second deployment June 2003
-
EE249Fall0838
Energy Management and ConservationCal ISO Daily Peak Loads
January 1, 2000 - December 31, 2000
20
25
30
35
40
45
50
GW
Peak Day August 16 -43.5 GW
Commercial AC
Residential AC
Advanced thermostats operate on required level of comfort, energy cost, weather forecast and distributed measurements to offload peak times Appliances are energy and cost aware
Demand response:Make energy pricesdependent upon time-of-use
-
EE249Fall0839
-
EE249Fall0840
Tire to Vehicle
Body computer
Stability ControlSystem
Smart antenna
Smart antenna
Energy Scavenging
Power Management
SensingDevice RF Link Computing
Rx/Tx Antenna
SW&
-
EE249Fall0841
Industrial Plants
Monitoring:Vibrations, Temperature, Humidity, Position, Logistics
Control and Actuation:Robot movements, Temperature, Storage
Current solution:Wired Infrastructure
Future solution:WIRELESS
Wireless advantages:Reduce cablingEnhance flexibilityEasy to deploy Higher safety Decreased maintenance costs
Alvise Bonivento
-
EE249Fall0842
Temperature Tracking
No or little real-time data on assets, environment, or activity
Inventory/supply management
Pharmaceutical
Foods
Automated meter reading?
Source: Xbow
-
EE249Fall0843
Weyerhaeuser 20 Million Seed Management Task:
Manage 20 million fast growing seeds annually
Issue: Seed dormancy depends on a complex
combination of water, light, temperature, gasses, mechanical restrictions, seed coats, and hormone structures
Southeastern US growing regions
Source: Xbow
-
EE249Fall0844
Tree Growth Rate Variability
Old Method
Trust nature
Monitor local atmospheric conditions
Sensor Network Way:
Monitor soil temperature and moisture at various locations
Adjust irrigation schedule accordingly
-
EE249Fall0845
Preventative Maintenance Program on Oil Tankers The task:
Engine monitoring is critical for both keeping the ship operational and complying with insurance policy.
Old Methods Manually record
vibration profile with data loggers.
Post process data for engine health and diagnostics.
Source: Xbow
-
EE249Fall0846
Personal Themes
Data driven, remote feedback control
Government or industry mandates
Personal computing themes
Ubiquitous computing
Safety
Convenience
Health and performance
Entertainment
Source: Xbow
-
EE249Fall0847
Silicon-Processed Micro-needles
Lin and Pisano, IEEE/ASME J. of MEMS, Vol. 8, pp 78-84, 1999
Neural probe with fluidchannel for bio-medical appl.
Two micro-needles penetratingporterhouse (New-York) steak
-
EE249Fall0848
Applications
Disaster Mitigation (natural and otherwise) Monitor buildings, bridges, lifeline systems to assess damage after
disaster
Provide efficient, personalized responses
Must function at maximum performance under very difficult circumstances
-
EE249Fall0849
What is Disaster Response?
Sensors installed near critical structural points
Sensor measure motion, distinguish normal deterioration and serious damage
Sensors report location, kinematics of damage during and after an extreme event Guide emergency personnel
Assess structural safety without deconstructing building
-
EE249Fall0850
Discussion
What are the most challenging aspects of these applications (and how does a company make money) ?
Interaction mechanisms: sensors, actuators, wireless networks
Reliability and survivability
Infrastructure
Services
Legislation
-
EE249Fall0851
PresenterPresentation NotesOur critical infrastructures are illustrated here. As you can see, these infrastructures play a crucial role in our society and daily lives. As such, the destruction or degradation of one or more of these infrastructures could cause serious harm to our economic and national security. The President has recognized this potential threat and has ordered that steps be taken to protect our infrastructures from an attack.
In the past, threats to our nations infrastructures were mainly physical in nature. We used to be concerned primarily about threats from terrorist groups and hostile nations. Now, criminal groups, terrorists, and hostile nations can interrupt critical infrastructures through cyber attacks on crucial automation systems.
As our society becomes more global and utilizes technology to increase the efficiency of our enterprises, our nations critical infrastructures are becoming increasingly interdependent within an enterprise, across several enterprises, even across industries. For example, the financial services industry depends on the availability and reliability of the telecommunications infrastructure, which in turn relies on electric power.
Hence, future attacks against one infrastructure could have cascading effects in the operations of others. within one enterprise, across several enterprises, or industries, and potentially all over the world.
-
EE249Fall0852
Secure Network Embedded SystEms (SENSE)
Networked embedded systems and distributed control creates a new generation of future applications: new infrastructures
We need to think about how to prevent the introduction of vulnerabilities via this exciting technology
Security, Networking, Embedded Systems
-
EE249Fall0853
Outline for the Introduction
Examples of Embedded Systems
Their Impact on Society
Design Challenges
Embedded Software and Control
-
EE249Fall0854
Opportunity: Electronic Systems Design Chain
Interfaces
Fabrics
Manufacturing
Implementation
System Design
IP
Design Science
-
EE249Fall0855
Disaggregation:Complex Design Chain Management
Supply Chain Movement of tangible goods from
sources to end market Supply Chain Management is $3.8B
market projected to be $20B in 2005
Design Chain Movement of technology
(IP and knowledge) from sources to end market
Design Chain Management is an untapped market
SemiconductorCompanies
SystemCompanies
Foundries System Test Equipment
Contract Manufacturing
Design Services
Embedded Software
Software Development Tools
EDA
Processor & Hardware IP
Process & Yield Services
Fabrication Equipment
IC Packaging & Test
Mechanical CAD
SubsystemCompanies
-
EE249Fall0856
Supply Chain: Design Roles-> Methodology->Tools
Methodology
Design Roles
Tools
-
EE249Fall0857
Product Specification & Architecture Definition(e.g., determination of Protocols and Communication standards)System Partitioning and Subsystem SpecificationCritical Software DevelopmentSystem Integration
Automotive Supply Chain:Car Manufacturers
-
EE249Fall0858
1 Transmission ECU2 Actuation group3 Engine ECU4 DBW5 Active shift display6/7 Up/Down buttons 8 City mode button9 Up/Down lever10 Accelerator pedal
position sensor11 Brake switch
Subsystem Partitioning Subsystem IntegrationSoftware Design: Control Algorithms, Data ProcessingPhysical Implementation and Production
Automotive Supply Chain:Tier 1 Subsystem Providers
-
EE249Fall0859
HW layer
SW Platform layer(> 60% of total SW)
Application Platform layer( 10% of total SW)
Controllers Library
OSEKRTOS
OSEKCOMI/O drivers & handlers
(> 20 configurable modules)
Application Programming Interface
Boot LoaderSys. Config.
TransportKWP 2000
CCP
ApplicationSpecificSoftware
Speedometer
Tachometer
Water tem
p.
Speedometer
Tachometer
Odom
eter---------------
ApplicationLibraries
Nec78k HC12HC08 H8S26 MB90
CustomerLibraries
Automotive Supply Chain:Subsystem Providers
Platform Integration: firmware and glue softwareSoftware Design: Application
-
EE249Fall0860
HW layer
SW Platform layer(> 60% of total SW)
Application Platform layer( 10% of total SW)
Controllers Library
OSEKRTOS
OSEKCOMI/O drivers & handlers
(> 20 configurable modules)
Application Programming Interface
Boot LoaderSys. Config.
TransportKWP 2000
CCP
ApplicationSpecificSoftware
Speedometer
Tachometer
Water tem
p.
Speedometer
Tachometer
Odom
eter---------------
ApplicationLibraries
Nec78k HC12HC08 H8S26 MB90
CustomerLibraries
Automotive Supply Chain:Platform & IP Providers
Software platform: RTOS and communication layerHardware platform: Hardware and IO drivers
-
EE249Fall0861
Outline for the Introduction
Examples of Embedded Systems
Their Impact on Society
Design Challenges
Embedded Software and Control
-
EE249Fall0862
How Safe is Our Real-Time Software?
-
EE249Fall0863
Computing for Embedded Systems
Imag
e b
orro
wed
fro
m a
n Io
meg
a ad
vert
isem
ent
for
Y2K
soft
ware
and
dis
k dr
ives
, Sci
enti
fic
Amer
ican
, Sep
tem
ber
1999
.
-
EE249Fall0864
Mars, December 3, 1999Crashed due to un-initialized variable
-
EE249Fall0865
$4 billion development effort40-50% system integration & validation cost
-
EE249Fall0866
Complexity, Quality, & Time To Market today
* C++ CODE FABIO ROMEO, Magneti-MarelliDAC, Las Vegas, June 20th, 2001
Memory
Lines Of Code
Changing Rate
Dev. Effort
Validation Time
Time To Market
INSTRUMENTCLUSTER
Productivity
Residual DefectRate @ End Of Dev
256 Kb
50.000
3 Years
40 Man-yr
5 Months
24 Months
PWT UNIT
6 Lines/Day
3000 Ppm
128 Kb
30.000
2 Years
12 Man-yr
1 Month
18 Months
BODY GATEWAY
10 Lines/Day
2500 ppm
184 Kb
45.000
1 Year
30 Man-yr
2 Months
12 Months
6 Lines/Day
2000ppm
8 Mb
300.000
< 1 Year
200 Man-yr
2 Months
< 12 Months
TELEMATICUNIT
10 Lines/Day*
1000 ppm
-
EE249Fall0867
Software Bugs Cost $59.5 Billion a Year
Software bugs cost $59.5 billion a year, study says
INFOWORLD, JUNE 28, 2002 BY PAUL KRILL
Software bugs cost the U.S. economy an estimated $59.5 billion per year, or 0.6 percent of the gross domestic product,
according to a newly released study by the U.S. Department of Commerce National
Institute of Standards and Technology (NIST). In a statement released on Friday, NIST said
more than half the costs are borne by software users and the remainder by
software developers and vendors.
Additionally, the study found that although errors cannot be removed, more than a third of the costs, or an estimated $22.2 billion,
could be eliminated by improved testing that enables earlier and more effective
identification and removal of defects. Currently, more than half of errors are not
found until downstream in the development process or during post-sale use of software,
according to NIST.
-
EE249Fall0868
Embedded Software Architecture Today
-
EE249Fall0869
We Live in an Imperfect World!
PresenterPresentation NotesAdd one more circle with a car related headline
-
EE249Fall0870
How is Embedded Software Different from Ordinary Software?
It has to work
One or more (very) limited resources
Registers
RAM
Bandwidth
Time
Source: Alex Aiken
-
EE249Fall0871
Devils Advocate
So whats different?
All software works with limited resources
We have compiler technology to deal with it
Various forms of program analysis
Source: Alex Aiken
-
EE249Fall0872
Example: Registers
All machines have only a few registers
Compiler uses the registers as best as it can
Spills the remaining values to main memory
Manages transfers to and from registers
The programmer feels she has 1 registers
Source: Alex Aiken
-
EE249Fall0873
The Standard Trick
This idea generalizes
For scarce resource X Manage X as best as we can
If we need more, fall back to secondary strategy
Give the programmer a nice abstractionSource: Alex Aiken
-
EE249Fall0874
The Standard Trick
This idea generalizes
For scarce resource X Manage X as best we can
Any correct heuristic is OK, no matter how complex
If we need more, fall back to secondary strategy
Focus on average case behavior
Give the programmer a nice abstractionSource: Alex Aiken
-
EE249Fall0875
Examples of the Standard Trick
Compilers
Register allocation
Dynamic memory management
OS
Virtual memory
Caches
Summary: abstract and hide complexity of resources
Source: Alex Aiken
-
EE249Fall0876
Whats Wrong with This?
Embedded systems have limited resources
Meaning hard limits Cannot use more time
Cannot use more registers
The compiler must either Produce code within these limits
Report failure
The standard trick is anathema to embedded systems Cant hide resources
Source: Alex Aiken
-
EE249Fall0877
Revisiting the Assumptions
Any correct heuristic is OK, no matter how complex
Embedded programmer must understand reasons for failure
Feedback must be relatively straightforward
Focus on average case behavior
Embedded compiler must reason about the worst case
Cannot improve average case at expense of worst case
Give the programmer a nice abstraction
Still need abstractions, but likely different ones
Source: Alex Aiken
-
EE249Fall08Source Ed Lee
Another Traditional Systems Science -Computation, Languages, and Semantics
States = Bits*
results + state out
sequence f : States States
Everything computable can be given by a terminating sequential program.
Functions on bit patterns Time is irrelevant Non-terminating programs are defectiveAlan Turing
-
EE249Fall0879
Processes and Process Calculi
incoming message
outgoing message
Infinite sequences of state transformations are called processes or threads
In prevailing software practice, processes are sequences of external interactions (total orders).
And messaging protocols are combined in ad hoc ways.
Various messaging protocols lead to various formalisms.
Source Ed Lee
-
EE249Fall0880
stalled for rendezvous
stalled by precedence
timing dependence
Interacting Processes Concurrency as Afterthought
Software realizing these interactions is written at a very low level (e.g., semaphores). Very hard to get it right.
Source Ed Lee
-
EE249Fall0881
Interacting Processes Not Compositional
An aggregation of processes is not a process (a total order of external interactions). What is it?
Many software failures are due to this ill-defined composition.
Source Ed Lee
-
EE249Fall0882
Compositionality
Non-compositional formalisms lead to very awkward architectures.
-
EE249Fall0883
What About Real Time?
Make it faster!
-
EE249Fall08
First Challenge on the Cyber Side:Real-Time and Power-aware Software
Correct execution of a program in C, C#, Java, Haskell, etc. has nothing to do with how long it takes to do anything. All our computation and networking abstractions are built on this premise.
Timing of programs is not repeatable, except at very coarse granularity.
Programmers have to step outside the programming abstractions to specify timing and power behavior.
-
EE249Fall08
Second Challenge on the Cyber Side:Concurrency
Threads dominate concurrent software. Threads: Sequential computation with shared memory. Interrupts: Threads started by the hardware.
Incomprehensible interactions between threads are the sources of many problems:
Deadlock Priority inversion Scheduling anomalies Nondeterminism Buffer overruns System crashes
-
EE249Fall08
86
Common Features
Transport Layer
Network Layer
MAC Layer
Link Layer
Dis
cret
e E
vent
Physical Layer
Application
Pre-Post
Process Networks
x Low pass
Manager Tables andParameters
User CSP
ContinuousTimem
+
c
s
Systems are assembled out of heterogeneous components
Systems are distributed
Interactions difficult to define
PresenterPresentation NotesThe motivations for this work are very simple. Embedded systems are designs, like this node in a sensor network, whose parts are often expressed efficiently only in very different formalism. Now, it is essential to understand how these formalism, which we call models of computation, interact with each other. To do so we build certain relationships between the models, relationships that can be classified as abstractions and refinements.In this talk I will present conservative approximations and compare them with other
-
EE249Fall08
87
The Intellectual Agenda
To create a modern computational systems science and systems design practice with
Concurrency
Composability
Time
Hierarchy
Heterogeneity
Resource constraints
Verifiability
Understandability
-
Lee, Berkeley 88
Chess: Center for Hybrid and Embedded Software Systems
Principal InvestigatorsThomas Henzinger (EPFL)Edward A. Lee (Berkeley)Alberto Sangiovanni-Vincentelli (Berkeley)Shankar Sastry (Berkeley)Janos Sztipanovits (Vanderbilt)Claire Tomlin (Berkeley)
Executive DirectorChristopher Brooks
Associated FacultyDavid Auslander (Berkeley, ME)Ahmad Bahai (Berkeley)Ruzena Bajcsy (Berkeley)Gautam Biswas (Vanderbilt)Ras Bodik (Berkeley, CS)Bella Bollobas (Memphis)Karl Hedrick (Berkeley, ME)Gabor Karsai (Vanderbilt)Kurt Keutzer (Berkeley)George Necula (Berkeley, CS)Koushik Sen (Berkeley, CS)Sanjit Seshia (Berkeley)Jonathan Sprinkle (Arizona)Masayoshi Tomizuka (Berkeley, ME)Pravin Varaiya (Berkeley)
This center, founded in 2002, blends systems theorists and application domain experts with software technologists and computer scientists.
Some Research ProjectsPrecision-timed (PRET) machinesDistributed real-time computingSystems of systemsTheoretical foundations of CPSHybrid systemsDesign technologiesVerificationIntelligent controlModeling and simulation
ApplicationsBuilding systemsAutomotiveSynthetic biologyMedical systemsInstrumentationFactory automationAvionics
the Berkeley directors of Chess
-
EE249Fall08
Uptime: 125 years
Why cant we make Software Reliable?
Source: T. Henzinger
-
EE249Fall08
Engineering
Theories of estimation. Theories of robustness.
Computer Science
Theories of correctness.
Why cant we make Software reliable?
R BSource: T. Henzinger
-
EE249Fall08
Engineering
Theories of estimation. Theories of robustness.
Goal: build reliable systems.
Computer Science
Theories of correctness.
Temptation: programs are mathematical objects; hence we want to prove them correct.
Why cant we make Software reliable?
Source: T. Henzinger
-
EE249Fall08
B
Engineering Computer Science
The CHESS Premise: The pendulum has swung too far
RSource: T. Henzinger
-
EE249Fall08
B
Engineering Computer Science
Embedded Systems are a perfect playground to readjust the pendulum.
RPhysicality Computation
The CHESS Premise: The pendulum has swung too far
Source: T. Henzinger
-
EE249Fall08
EmbeddedSystems
Execution constraintsCPU speed power failure rates
Reaction constraintsdeadlines throughput jitter
Computation algorithms protocols reuse
Source: T. Henzinger
-
EE249Fall08
EmbeddedSystems
Execution constraintsCPU speed power failure rates
Reaction constraintsdeadlines throughput jitter
Computation algorithms protocols reuse
Embedded System Design is generalized hardware design (e.g. System C).
Source: T. Henzinger
-
EE249Fall08
EmbeddedSystems
Execution constraintsCPU speed power failure rates
Reaction constraintsdeadlines throughput jitter
Computation algorithms protocols reuse
Embedded System Design is generalized control design (e.g. Matlab Simulink).
-
EE249Fall08
EmbeddedSystems
Execution constraintsCPU speed power failure rates
Reaction constraintsdeadlines throughput jitter
Computation algorithms protocols reuse
Embedded System Design is generalized software design (e.g. RT Java).
Source: T. Henzinger
-
EE249Fall08
EmbeddedSystems
Execution constraintsCPU speed
power
failure rates
Reaction constraintsDeadlines
throughput
jitter
Computation Algorithms
protocols
reuse
Source: T. Henzinger
-
EE249Fall08
We need a new formal foundation for embedded systems, which systematically
and even-handedly re-marries
computation and physicality.
The CHESS Challenge
Source: T. Henzinger
-
EE249Fall08
Integration of the Two Cultures
EngineeringComponent model: transfer function Composition: parallel Connection: data flow
Computer ScienceComponent model: subroutine Composition: sequential Connection: control flow
[Hybrid Systems; Ptolemy; Metropolis; Metamodels]
-
EE249Fall08
Equational ModelsStrengths:
Concurrency Quantitative constraints (time, power, QoS)
Tool support:
Best-effort design Optimization
Abstract-Machine Models
Dynamic change Complexity theory
Worst-case analysis Constraint satisfaction
Integration of the Two Cultures
Engineers must understand both complexities and trade-offs .
Source: T. Henzinger
-
EE249Fall08
102
The Embedded Software SCIENCE Dilemma
Raffaello Sanzio, The Athens School
-
EE249Fall08103
Software Architecture Today
Poor common infrastructure. Weak specialization of functions. Poor resource management. Poor planning.
-
EE249Fall08104
Software Architecture Tomorrow?
EE249 Embedded System Design: Models, Validation and SynthesisAlberto Sangiovanni VincentelliSlide Number 2eMerging Societal-Scale SystemsEmbedded SystemsObservationsCourse overviewBehavior Vs. ArchitectureBehavior Vs. CommunicationAdministrationGradingScheduleLinksOutline of the courseOutline for the IntroductionElectronics and the CarAutomotive IndustryThree Levels of PlayersChallenge: Electronics, Controls and Software Shifting the Basis of Competition in VehiclesGM SAC Vehicular Electronics, Controls and Software StudySlide Number 19Slide Number 20Slide Number 21Slide Number 22Slide Number 23Consumer segmentsCommon technology elementsCommon Situation in IndustryOutline for the IntroductionConcurrency and HeterogeneityChallenge: The Physical InternetExponentials Bound to ContinueSlide Number 30Slide Number 31Slide Number 32Challenge: PowerSlide Number 34Smart DustWireless Sensor NetworksCreating a Whole New World of ApplicationsEnergy Management and ConservationSlide Number 39Tire to Vehicle Industrial PlantsTemperature TrackingWeyerhaeuser 20 Million Seed ManagementTree Growth Rate Variability Preventative Maintenance Program on Oil TankersPersonal ThemesSilicon-Processed Micro-needlesApplicationsWhat is Disaster Response?DiscussionSlide Number 51Secure Network Embedded SystEms (SENSE)Outline for the IntroductionOpportunity: Electronic Systems Design ChainDisaggregation:Complex Design Chain ManagementSupply Chain: Design Roles-> Methodology->ToolsAutomotive Supply Chain:Car ManufacturersAutomotive Supply Chain:Tier 1 Subsystem ProvidersAutomotive Supply Chain:Subsystem ProvidersAutomotive Supply Chain:Platform & IP ProvidersOutline for the IntroductionHow Safe is Our Real-Time Software?Computing for Embedded SystemsSlide Number 64Slide Number 65Complexity, Quality, & Time To Market today Software Bugs Cost $59.5 Billion a YearEmbedded Software Architecture TodayWe Live in an Imperfect World!How is Embedded Software Different from Ordinary Software?Devils AdvocateExample: RegistersThe Standard TrickThe Standard TrickExamples of the Standard TrickWhats Wrong with This?Revisiting the AssumptionsAnother Traditional Systems Science - Computation, Languages, and SemanticsProcesses and Process CalculiInteracting Processes Concurrency as AfterthoughtInteracting Processes Not CompositionalCompositionalityWhat About Real Time?First Challenge on the Cyber Side:Real-Time and Power-aware SoftwareSecond Challenge on the Cyber Side:ConcurrencyCommon FeaturesThe Intellectual AgendaChess: Center for Hybrid and Embedded Software SystemsSlide Number 89Slide Number 90Slide Number 91Slide Number 92Slide Number 93Slide Number 94Slide Number 95Slide Number 96Slide Number 97Slide Number 98The CHESS ChallengeSlide Number 100Slide Number 101The Embedded Software SCIENCE DilemmaSoftware Architecture TodaySoftware Architecture Tomorrow?