Virtual Prototype with NEC Microcontrollers
Visualize and Optimize Systems & Software Engineering
Amit Choudhury Automotive Strategic Business Unit NEC Electronics America February 5th 2009
1
Contents
Challenges and solutions
• Industry and Academia needs
Virtual Prototype inherent need for SW and system engineering NEC investment to grow the adoption of Virtual Prototype
2
Growth of Feature-Function in vehicle
Increasing software content in ECU
Solutions by electronics, control and software Vehicle competitive value Exponential growth in software (900%)
Source : Sri Kanajan, GM
Source : Sri Kanajan, GM
3
System & Software complexity, heterogeneity
Multi-disciplinary models and simulation framework
Reliable, Robust Embedded Control Software costly, time consuming Reason
•
• • •
Complexity − Functionality ↑ − Flexibility ↑ − Time-to-market ↓ − Constraints Heterogeneity − Multidisciplinary Lack of predictability Late integration Virtual Prototyping = Simulation Model-level integration discipline-specific issues Stepwise Refinement = Iterative approach Property-preserving code generation
Approach
• • •
•
Source : Dr. Jan F. Broenink, Marcel Groothuis University of Twente
Shorten design time Better quality product
4
Embedded Control System-SW Engineering
Approach
• Concurrent engineering • Local, predictable refinement
−
Focus on aspects
Paper, Huang et al, ACSD 2007
Modeling
• Concurrency model
−
Untimed interactions between concurrent „players‟ − Aspects: deadlock, resource access, synchronization • Model of Computation − Interactions between MoC: discrete event continuous time • Real-time model − Add non functional aspect: timing
soft
hard
Source : Dr. Jan F. Broenink, Marcel Groothuis University of Twente
5
Multi-ECU System and SW Architecture
Optimal Task, Message scheduling & allocation Multi-ECU performances
AUTOSAR: Standardized Distributed Control (Safety) Middleware FlexRay: Time Triggered Architecture with global notion of time HW/SW Architecture incrementally gearing towards client-server Communication becoming dominant performance criteria Computation and communication: end-to-end performances
Source : Klaus D. Muller-Glaser, Universitaet Karlsruhe
Source : A. Sangiovanni Vincentelli, A. Ferrari, PARADES
6
Platform based design
Separation of “application function” from “performance platform” AUTOSAR based on same concept Virtual Prototype enables early Platform hardware/software exploration and performance measurement
Source : A. Sangiovanni Vincentelli, A. Ferrari, PARADES
Source : A. Sangiovanni Vincentelli, A. Ferrari, PARADES Source : Klaus D. Muller-Glaser, Universitaet Karlsruhe
7
Platform Software Components
Platform optimized for performance/cost for control algorithm
Hand-coded: Need to be scheduled early Auto-Code generation: Model based design of application SW Efficient design of platform allows for optimal re-use for performance/cost
Source : A. Sangiovanni Vincentelli, A. Ferrari, PARADES
Source : A. Sangiovanni Vincentelli, A. Ferrari, PARADES
8
How Virtual Prototypes can help ?
Integrated Software Development and Verification
Cycle Accurate virtual prototype advantage over silicon/board prototype
Early software development for product development, achieve time to market even though functionality increases in vehicle • Quality software development needs high visibility of system performance
•
High simulation speed virtual prototype enables iterative simulation delivering software and system engineering
Platform based development adopted by AUTOSAR also needs iterative analysis • Evolving HW/SW Architecture with AUTOSAR and FlexRay directed towards distributed control, can be analyzed
•
High visibility of virtual prototype allows to communicate in pictures both inside and outside your discipline team
•
Multi-disciplinary engineering: software, electronic system, control, electro-mechanical needs co-simulation needs iterative stepwise refinement
9
Virtual Prototype System and Software Engineering
System and software engineer can do “what ifs”
Virtual prototype abstracts platform hardware
• Embedded software runs oblivious of not running on Silicon
Quickly configure multiple platform for target performances
• Depending on (range of) control algorithm needs
Deploy virtual environment globally within minutes, improving logistics on silicon based environment Provides pertinent metrics for analysis for performances for platform hardware, platform software and application software
• Hardware: Power, CPU cycle throughput, Communication Channel
bandwidth/throughput • Platform SW: Distributed/Safety middleware -abilities like reliability, availability, etc. • Application SW: Measure of controllability, performance with plant
10
Virtual Prototype – NEC’s view
Virtual Prototype is model: simplification, abstraction Model‟s purpose provide relevant metric information for analysis Industry stakeholders contribute what to incorporate to enable more benefit from these Virtual Prototype model and simulation framework
V850 System Clock V850 Core Memory Model Reset
Memory
V850
DMA Interrupt controller controller
V850 Memory processor
Timer
Timer
DMA Interrupt controller controller
Platform Creation and Analysis
Bus Model
Embedded Software Development
IRQ Timer
Interrupt Controller
DMA Controller
DMA
UART
CAN
Stream Data
Serial Console
11
Software Development
Metrix Output UML, Simulink, C, C++, …
Create
Compile
Monitor prototype internals
• • • • •
Assemble
Link
HW
Cache hits/misses Bus transactions Processor performance Memory usage Interrupt latency
Load
VaST VSP
Trigger hardware and software debuggers Example usage: Analyze processor and platform power
•
Debug + Monitor SW IDE
Make intelligent tradeoffs between power, performance and cost
12
V Model
VaST helps SW development and system architecture performance
Source : Klaus D. Muller-Glaser, Universitaet Karlsruhe
13
NEC’s VaST releases
Enabling solutions for customers
Three generations of microcontrollers for Safety Body, Safety and Powertrain microcontrollers Today : Pre-Silicon Model Development
• Rx3 delivered in Q1‟2008 • Planning in progress for x4 generation
• Demand-driven releases in safety in 2004/2005 • x3 generation models : 2008 (body/Fx3 & safety/Rx3)
Some Next Step Challenges:
• Globalizing model development strategy • Validating multi-ECU, multi-core, FlexRay networks • Mapping VaST Environment HDL (RTL) and HVL
• Enhanced SystemC peripheral integration
14
NEC Demonstration
AUTOSAR Stack with CAN on 2 NEC FJ3 Body ECU Metric: Latency because of task pre-emption between Application Function call to CAN RX Acknowledge Picture Visualization by dll integration to VaST Model
15
Acknowledgement
NEC thanks papers/presentation from Industry and Academia
• Prof. Dr. Alberto Sangiovanni Vincentelli, PARADES • Dr. Alberto Ferrari, PARADES • Sri Kanajan, General Motors
• Prof. Dr. Ing. Klaus D. Muller-Glaser, Universitaet Karlsruhe • Dr. Juergen Bortolazzi, Daimler Chrysler
• Dr. Jan F. Broenink, University of Twente • Marcel Groothuis, University of Twente
16
17