Model-Implementation Fidelity in Cyber Physical System Design (eBook)

Anca Molnos received her M.Sc. degree in computer science from the “Politehnica” University of Bucharest, Romania and the Ph.D. degree in computer engineering from the Delft University of Technology, The Netherlands, in 2001 and 2009, respectively. Between 2006 and 2009 she was senior scientist at NXP Semiconductors, The Netherlands, working on low-power multi-processors and distributed real-time systems. From 2009 to 2012 she was a researcher with the Delft University of Technology, working on embedded multi-core resource management for low-power and quality of service. In January 2013 she joined CEA LETI, where her research focuses on developing energy-aware software, energy and variability management, and frameworks for adaptable parallel systems. During the years, she (co-)authored more than 40 papers in journals and international conferences and several patents. She serves or served in many program committees, among which ICCD, ICCAD, RTCSA, ICPADS, as well as participated in the organization of several conferences and workshops as program committee chair, or other chair positions.Christian Fabre received his Engineering degree from École nationale supérieure de mathématiques appliquées de Grenoble (ENSIMAG), France, in 1990 while working on a Transputer routing kernel. He joined the OPEN SOFTWARE FOUNDATION Research Institute in 1993 to work on ANDF, an intermediate language. Later he worked on Java Virtual Machines, Java compilation and embedded sotware components. He shared the “Best Embedded Java Product” with other members of the OSF-RI TurboJava team at JavaOne in 2000. After the acquisition of the OSF-RI by Groupe SILICOMP (now ORANGE Business Services) he was part of the corporate SILICOMP R&D team. Since 2004 he has transitioned from compilation and software development to system development by adopting the MDA/MDE approach for top-down co-design of mixed hardware/software systems. He joined CEA LETI in Grenoble, France in 2009 as a Senior Research Engineer. He has been involved in various collaborative projects, such as OMI-GLUE (Esprit 1992-1995), Pastoral (FP5 2000-2012), and Expresso (French RNTL Project, 2001-2003). He was the coordinator of PRO3D (FP7 2012-2012, http://pro3d.eu) and is currently coordinating COPCAMS (COgnitive & Perceptive CAMeraS, 2013-2016, http://copcams.eu), an Artemis/ECSEL project of more than 20 partners.

Preface 5
Introduction 7
Contents 10
1 Building Faithful Embedded Systems Models: Challenges and Opportunities 12
1.1 Introduction 12
1.2 Challenges for ES Performance 15
1.2.1 Memory Contention 16
1.2.2 Memory Architectures 17
1.2.3 Caches and DMAs 17
1.2.4 Multi-processing, Pipelining, and Others 18
1.3 Performance Modeling: State of the Art 18
1.3.1 Techniques for Gathering ES Performance 19
1.3.2 Characterizing ES Performance: Models and Methods 19
1.3.2.1 Detailed Representations 19
1.3.2.2 Abstract Representations 19
1.3.2.3 Probabilistic Representations 21
1.4 The ASTROLABE Approach 21
1.4.1 Overview 22
1.4.2 Distribution Fitting 23
1.4.3 Assumptions and Shortcomings 24
1.5 Performance Modeling Using Probabilistic Models 24
1.5.1 Probabilistic Models 26
1.5.1.1 Mixture Distributions 26
1.5.1.2 Regression Models 27
1.5.1.3 Markov Models 29
1.5.2 Learning and Fitting Techniques 29
1.5.2.1 Fitting Mixture Models 30
1.5.2.2 Fitting Regression Models 30
1.5.2.3 Learning Markov Models 31
1.5.3 Perspectives 31
References 32
2 Resource-Driven Modelling for Managing Model Fidelity 36
2.1 Introduction 36
2.2 Resource-Driven Modelling 38
2.2.1 System Design and Implementation 39
2.2.2 Dynamic Systems and Architectures 41
2.2.3 Resources Quantification and Reward Functions 43
2.2.4 Constrained Architectures 44
2.3 Hierarchical Modelling in Order Graphs 45
2.3.1 Introducing Hierarchies 45
2.3.2 Order Graphs 47
2.3.3 Cross-Layer Cuts 48
2.4 Case Studies 50
2.4.1 Studying the Performance, Energy and Reliability Trade@?????-Offs of Scalable Systems 50
2.4.2 Exploring Concurrency in Many-Core Systems 53
2.4.2.1 Architecture-Open (ArchOn) Simulator 53
2.4.2.2 Benchmark Results 54
2.4.3 Power-Proportional Modelling of Heterogeneous Systems 58
2.4.3.1 Platform Description 58
2.4.3.2 Platform Model 59
2.4.3.3 Power-Proportional Model Sizes 62
2.5 Conclusions 64
References 65
3 Empowering Mixed-Criticality System Engineers in the Dark Silicon Era: Towards Power and Temperature Analysis of Heterogeneous MPSoCs at System Level 67
3.1 Introduction 67
3.2 The CONTREX Project 69
3.3 Considering Power and Temperature of a Full Chip at System Level—The CONTREX Flow 74
3.3.1 Extended Virtual Platform Simulation 75
3.3.2 Primary Traces: Observable Properties 77
3.3.3 Stream Processing 77
3.3.4 Secondary Traces: Power per Component 78
3.3.5 Power Mapping 78
3.3.6 Thermal Model Generation of IC Package 79
3.3.7 Thermal Estimation 80
3.3.8 Tertiary Traces 80
3.3.9 Contract Satisfaction Monitoring 80
3.4 A Mixed-Criticality Use-Case 81
3.4.1 Selected Scenario 81
3.4.2 Fundamentals 82
3.4.3 Payload Setup 83
3.4.4 Hardware 83
3.4.5 Software 86
3.4.5.1 The Safety-Critical Part 87
3.4.5.2 The Mission-Critical Part 87
3.4.6 Ground Control Station 87
3.5 Application of the CONTREX Flow 88
3.5.1 Virtual Platform 88
3.5.2 Timing Model 89
3.5.3 Power Model 91
3.5.4 Temperature Model 94
3.6 Conclusion and Future Work 96
References 98
4 Throughput-Driven Parallel Embedded Software Synthesis from Synchronous Dataflow Models: Caveats and Remedies 101
4.1 Introduction 101
4.2 Streaming Throughput Analysis 102
4.2.1 Overview 102
4.2.2 Preliminaries 103
4.2.2.1 SDF Model 103
4.2.2.2 Target Platform Model 103
4.2.2.3 Buffer-Throughput Tradeoff 104
4.2.3 Inaccuracy in SDF-Based Throughput Analysis 105
4.2.3.1 Throughput Analysis Based on SDF Operational Semantics 105
4.2.3.2 Abstract View of Implementation 106
4.2.4 Proposed Solution: Implementation Aware Throughput Analysis 108
4.2.4.1 Reader and Writer Actors 108
4.2.4.2 Sync Actors 110
4.2.4.3 Properties 111
4.2.5 Empirical Evaluation 113
4.2.5.1 Setup and Benchmark Applications 113
4.2.5.2 Implementation Aware vs. Implementation Oblivious Analysis 114
4.2.5.3 Comparison Against Cycle-Accurate Simulation 116
4.3 Platform-Oriented Throughput Scaling 119
4.3.1 Overview 119
4.3.2 Baseline Software Synthesis 120
4.3.3 SDF Limitations in Throughput Scaling 121
4.3.4 Proposed Solution: FORMLESS Model 121
4.3.4.1 Formalism 122
4.3.4.2 Higher-Order Language 124
4.3.4.3 Exploration of Forming Parameter Space 124
4.3.5 Experimental Evaluation 127
4.3.5.1 Application Case Studies 127
4.3.5.2 Experiment Setup 131
4.3.5.3 Measurement Results 131
4.4 Related Work 133
4.5 Conclusion 135
References 135
5 SimSoC: A Fast, Proven Faithful, Full System Virtual Prototyping Framework 138
5.1 Introduction 138
5.2 The SimSoC Framework 139
5.2.1 Performance Estimate 147
5.3 Faithful Simulation 149
5.3.1 Objective 149
5.3.2 Formal Verification Background 150
5.3.3 Background Tools 151
5.3.3.1 Coq 151
5.3.3.2 Compert-C 152
5.3.4 Verified Simulation 153
5.3.4.1 Constructing the Formal Model 153
5.3.4.2 Proof Structure 154
5.3.4.3 Projection 155
5.3.4.4 Lemmas Library 157
5.3.4.5 Inversion 160
5.3.4.6 Instruction Proofs 161
5.4 Conclusion 162
References 163
6 A Composable and Predictable MPSoC Design Flow for Multiple Real-Time Applications 166
6.1 Introduction 166
6.2 Related Work 168
6.3 The Proposed Design Flow 169
6.4 The Modeling Framework 171
6.5 The Execution Platform 173
6.6 Adapting the Flows by Rapid Performance Evaluation 174
6.7 Case Study 179
6.8 Conclusion and Future Work 181
References 182
7 Analysis and Implementation of Embedded System Models: Example of Tags in Item Management Application 184
7.1 Introduction 184
7.2 A Holistic Design Process 185
7.3 pState Editor 186
7.4 From Hierarchical Charts to Code 187
7.5 Model Checker Input Code 190
7.5.1 MDP 190
7.5.2 PTA 192
7.5.3 Properties Specification 195
7.6 Executable Code 195
7.6.1 PIC C Code 196
7.6.2 PIC Assembly Code 197
7.6.3 Energia, Arduino-Like Code 198
7.7 Contention Resolution in DASH-7 ISO/IEC 18000-7.2 200
7.7.1 Tag Collection and Collision Arbitration 200
7.8 Collision Model 201
7.9 Collection Period Power Consumption 202
7.10 Executable Tag Code 204
7.11 Conclusion 206
References 207
8 Positioning System for Recreated Reality Applications Based on High-Performance Video-Processing 209
8.1 Introduction 209
8.1.1 Recreated Reality 211
8.1.2 State of the Art in Positioning Systems 212
8.2 System Architecture 214
8.2.1 Reference Marker Description 215
8.2.2 Design Methodology and Workflow 216
8.3 Positioning Algorithms 222
8.3.1 Input Data 223
8.3.2 Marker Detection 223
8.3.3 Movement Type 225
8.3.4 Geometry Algorithms 226
8.3.4.1 Perpendicular Movement 227
8.3.4.2 Parallel Movement 230
8.3.4.3 Possible Position 232
8.3.4.4 User Position 234
8.4 Synthesis 236
8.5 Conclusions 237
References 237
Index 239