Architectural Design Exploration for Real-Time Embedded Systems

点击翻译按钮获取中文摘要

Advances in both hardware and software design have enabled system designers to develop large, complex embedded systems. Such systems may contain a mixture of software programs running on general purpose or special-purpose processors, high-speed memory systems, and application specific hardware devices. One of the key tasks in designing these systems is to find a "best" system architecture including determining the "right" partition between hardware and software components and selecting the "right" hardware components. The focus of this project is to develop a viable methodology as well as a set of tools to assist system designers in this architectural design exploration task. The research proposed performs design exploration at the system level. At this level, a system is specified as a collection of task graphs and an implementation is described by an interconnection of hardware components and an assignment of processes to the hardware components. Though the problem shares some similarities with the hardware synthesis problem, many different issues need to be addressed in order for this approach to be effective. One of the issues is identifying and estimating appropriate parameters to be included in the exploration process such that system timing performance can be adequately predicted. The project emphasizes design exploration for embedded systems with real-time constraints. Such systems can be found in many applications, such as navigation and landing control of aircraft, networks and communications. Attributes that are useful for guiding design exploration will be devised so as to efficiently analyze the timing behavior of a large number of implementations with respect to the system timing specifications and hence permit exploration of a wide range of design options. Other issues including sensitivity analysis and component reuse will also be addressed. To properly handle conflicting performance requirements including both functional and non-functional attributes, an approach based on imprecise value functions from utility theory will be pursued. Furthermore, the design exploration problem is formulated as a multi-attribute optimization problem. Techniques from the evolutionary computation field will be investigated in order to tackle the exponential nature of the optimization problem.