The design process of embedded systems has changed substantially in recent years. One of the main reasons for this change is the pressure to shorten time-to-market when designing digital systems. To shorten the product cycles, programmable processes are used to implement more and more functionality of the embedded system. Therefore, nowadays, embedded systems are very often implemented by heterogeneous systems consisting of ASICs, processors, memories and peripherals. As a consequence, the research topic of hardware/software co-design, dealing with the problems of designing these heterogeneous systems, has gained great importance. Hardware/Software Co-design for Data Flow Dominated Embedded Systems introduces the different tasks of hardware/software co-design including system specification, hardware/software partitioning, co-synthesis and co-simulation. The book summarizes and classifies state-of-the-art co-design tools and methods for these tasks. In addition, the co-design tool COOL is presented which solves the co-design tasks for the class of data-flow dominated embedded systems. Author Ralf Niemann put this book's primary emphasis on the hardware/software partitioning and the co-synthesis phase and their coupling. In contrast to many other publications in this area, a mathematical formulation of the hardware/software partitioning problem is given. This problem formulation supports target architectures consisting of multiple processors and multiple ASICs. Several novel approaches are presented and compared for solving the partitioning problem, including an MILP approach, a heuristic solution and an approach based on genetic algorithms. The co-synthesis phase is based on the idea of controlling the system by means of a static run-time scheduler implemented in hardware. New algorithms are introduced which generate a complete set of hardware and software specifications required to implement heterogeneous systems. All of these techniques are described in detail and exemplified. This book is intended to serve students and researchers working on hardware/software co-design. At the same time, the variety of presented techniques automating the design tasks of hardware/software systems will be of interest to industrial engineers and designers of digital systems. "Niemann's method should be known by all persons working in the field. Hence, I recommend this book for everyone who is interested in hardware/software co-design." --Prof. Peter Marwedel, Ph.D, University of Dortmund Embedded Systems Group Table of Contents List of Figures List of Tables 1. Introduction Embedded Systems Hardware/Software Co-Design Outline of the Thesis 2. Co-Design Systems Co-Design for System Specification and Modelling Co-Design for Heterogeneous Implementation Processor Synthesis
Single-Processor Architectures with one ASIC
Single-Processor Architectures with many ASICs
Multi-Processor Architectures
Other Co-Design Approaches
Comparison of Co-Design Approaches Overview of COOL Classification of COOL 3. Specification of Embedded Systems Models of Computation State-Oriented Models
Activity-Oriented Models
Structure-Oriented Models
Data-Oriented Models
Heterogeneous Models
Requirements for Embedded System Specification Survey of System Specification Languages VHDL for System Specification System Specification in COOL 4. Hardware/Software Partitioning The Hardware/Software Partitioning Problem Hardware/Software Mapping
Hardware Sharing
Interfacing
Scheduling
Functional Pipelining
Related Work Hardware/Software Partitioning in COOL Special Aspects of HW/SW Partitioning in COOL
Hardware/Software Cost Estimation Software Cost Estimation
Hardware Cost Estimation
Interface Cost Estimation
Generation of the Partitioning Graph Formulation of the HW/SW Partitioning Problem Optimization Linear and Integer Linear Programming
Evolutionary Algorithms
The MILP-Model The Decision Variables
The Objective Function
The Constraints
Hardware Sharing
Interfacing
Scheduling
HW/SW Partitioning based on Heuristic Scheduling Step I: Mapping with Approximated Schedule
Step II: Scheduling solved by MILP
Step III: Scheduling solved by List Scheduling
Results I: Scheduling Algorithms
Results II: HW/SW Partitioning using Heuristic Scheduling
HW/SW Partitioning based on Genetic Algorithms GA Parameters
GA Encoding
GA Fitness Function
Analysis of the GA Approach
Results III: HW/SW Partitioning based on GA
GA Approach for Extended HW/SW Partitioning
5. Hardware/Software Co-Synthesis The Co-Synthesis Problem Communication Synthesis
Specification Refinement
Related Work Co-Synthesis in COOL Hardware/Software Systems generated by COOL System Controller
I/O Controller
Bus Arbiter
Controller for Data Paths
Bus Drivers
Simulation Models for Memory and Processors
State-Transition Graph State-Transition Graph Generation
State-Transition Graph Optimization
Memory Allocation
Refinement and Controller Generation Data-Related Refinements for Hardware and Software
Software Refinement
Hardware Refinement
Controller Generation
Results IV: Application Study of a Fuzzy Controller 6. The COOL Framework Implementation of COOL Description of COOL Graphical User Interface
Validation using Simulation
Specification of Design Constraints
Hardware/Software Partitioning
Co-Synthesis
Co-Simulation
7. Summary and Conclusions Summary and Contribution to Research Future Work Bibliography References Notations Abbreviations Index Hardcover; 244 pages.
|
