Internet-Based Micro-Electronic Design Automation (IMEDA) Framework

Introduction

As the complexity of VLSI systems continues to increase, the micro-electronic industry must possess an ability to reconfigure design and manufacturing resources and integrate design activities so that it can quickly adapt to the market changes and new technology. Gaining this ability imposes a two-fold challenge: (1) to coordinate design activities that are geographically separated and (2) to represent an immense amount of knowledge from various disciplines in a unified format. The Internet can provide the catalyst by abridging many design activities with the resources around the world not only to exchange information but also to communicate ideas and methodologies.

In this chapter, we present a collaborative engineering framework that coordinates distributed design activities through the Internet. Engineers can represent, exchange, and access the design knowledge and carry out design activities. The crux of the framework is the formal representation of process flow using the process grammar, which provides the theoretical foundation for representation, abstraction, manipulation, and execution of design processes. The abstraction of process representation provides mechanisms to represent hierarchical decomposition and alternative methods, which enable designers to manipulate the process flow diagram and select the best method. In the framework, the process information is layered into separate specification and execution levels so that designers can capture processes and execute them dynamically. As the framework is being executed, a designer can be informed of the current status of design such as updating and tracing design changes and be able to handling exception. The framework can improve design productivity by accessing, reusing, and revising the previous process for a similar. The cockpit of our framework interfaces with engineers to perform design tasks and to negotiate design tradeoff. The framework has the capability to launch whiteboards that enable the engineers in a distributed environment to view the common process flows and data and to concurrently execute dynamic activities such as process refinement, selection of alternative process, and design reviews. The proposed framework has a provision for various browsers where the tasks and data used in one activity can be organized and retrieved later for other activities.

One of the predominant challenges for micro-electronic design is to handle the increased complexity of VLSI systems. At the turn of the century, it is expected that there will be 100 million transistors in a single chip with 0.1 micron features, which will require an even shorter design time (Spiller, 1997). This increase of chip complexity has given impetus to trends such as system on a chip, embedded system, and hardware/software co-design. To cope with this challenge, industry uses custom-off-the-shelf (COTS) components, relies on design reuse, and practices outsourcing design. In addition, design is highly modularized and carried out by many specialized teams in a geographically distributed environment. Multi-facets of design and manufacturing, such as manufacturability and low power, should be considered at the early stage of design. It is a major challenge to coordinate these design activities (Fair, 94). The difficulties are caused by due to the interdependencies among the activities, the delay in obtaining distant information, the inability to respond to errors and changes quickly, and general lack of communications. At the same time, the industry must contend with decreased expenditures on manufacturing facilities while maintaining rapid responses to market and technology changes.

To meet this challenge, the U.S. government has launched several programs. The Rapid Prototyping of Application Specific Signal Processor (RASSP) Program was initiated by the Department of Defense to bring about the timely design and manufacturing of signal processors. One of the main goals of the RASSP program was to provide an effective design environment to achieve a four-time improvement in the development cycle of digital systems (Chung, 1996). DARPA also initiated a program to develop and demonstrate key software elements for Integrated Product and Process Development (IPPD) and agile manufacturing applications. One of the foci of the earlier program was the development of infrastructure for distributed design and manufacturing. Recently, the program is continued to Rapid Design Explo- ration & Optimization (RaDEO) to support research, development, and demonstration of enabling technologies, tools, and infrastructure for the next generation of design environments for complex electro-mechanical systems. The design environment of RaDEO is planned to provide cognitive support to engineers by vastly improving their ability to explore, generate, track, store, and analyze design alternatives (Lyons, 1997).

The new information technologies, such as the Internet and mobile computing, are changing the way we communicate and conduct business. More and more design centers use PCs, and link them on the Internet/intranet. The web-based communication allows people to collaborate across space and time, between humans, humans and computers, and computers in a shared virtual world (Berners-Lee, 1994). This emerging technology holds the key to enhance design and manufacturing activities. The Internet can

be used as the medium of a virtual environment where concepts and methodologies can be discussed, accessed, and improved by the participating engineers. Through the medium, resources and activities can be reorganized, reconfigured, and integrated by the participating organizations. This new paradigm certainly impacts the traditional means for designing and manufacturing a complex product. Using Java, programs can be implemented in a platform-independent way so that they can be executed in any machine with a Web browser. Common Object Request Broker Architecture (CORBA) (Yang, 1996) provides distributed services for tools to communicate through the Internet (Vogel). Designers may be able to execute remote tools through the Internet and see the visualization of design data (Erkes, 1996; Chan, 1998; Chung, 1998). Even though the potential impact of this technology will be great on computer aided design, Electronic Design Automation (EDA) industry has been slow in adapting this new technology (Spiller, 1997). Until recently, EDA frameworks used to be a collection of point tools. These complete suites of tools are integrated tightly by the framework using their proprietary technology. These frameworks have been suitable enough to carry out a routine task where the process of design is fixed. However, new tools appear constantly. To mix and match various tools outside of a particular framework is very difficult. Moreover, tools, expertise, and materials for design and manufacturing of a single system are dispersed geographically. Now we have reached the stage where a single tool or framework is not sufficient enough to handle the increasing complexity of a chip and emerging new technology. A new framework is necessary which is open and scalable. It must support collaborative design activities so that designers can add new tools to the framework, and interface them with other CAD systems. There are two key functions of the framework: (1) managing the process and (2) maintaining the relationship among many design repre- sentations. For design data management, refer to (Katz, 1987). In this chapter, we will focus on the

process management aspect.

To cope with the complex process of VLSI system design, we need a higher level of viewing of a complete process, i.e., the abstraction of process by hiding all details that need not to be considered for the purpose at hand. As pointed out in National Institute of Standards and Technology reports (Schlenoff, 1996; Knutilla, 1998), a “unified process specification language” should have the following major require- ments: abstraction, alternative task, complex groups of tasks, and complex sequences.

In this chapter, we first review the functional requirements of the process management in VLSI system design. We then present the Internet-based Micro-Electronic Design Automation (IMEDA) System. IMEDA is a web-based collaborative engineering framework where engineers can represent, exchange, and access design knowledge and perform the design activities through the Internet. The crux of the framework is a formal representation of process flow using process grammar. Similar to the language grammar, production rules of the process grammar map tasks into admissible process flows (Baldwin, 1995a). The production rules allow a complex activity to be represented more concisely with a small number of high-level tasks. The process grammar provides the theoretical foundation for representation, abstraction, manipulation, and execution of design and manufacturing processes. It facilitates the com- munication at an appropriate level of complexity. The abstraction mechanism provides a natural way of browsing the process repository and facilitates process reuse and improvement. The strong theoretical foundation of our approach allows users to analyze and predict the behavior of a particular process. The cockpit of our framework interfaces with engineers to perform design tasks and to negotiate design tradeoff. The framework guides the designer in selecting tools and design methodologies, and it generates process configurations that provide optimal solutions with a given set of constraints. The just-in-time binding and the location transparency of tools maximize the utilization of company resources. The framework is equipped with whiteboards so that engineers in a distributed environment can view the common process flows and data and concurrently execute dynamic activities such as process refinement, selection of alternative processes, and design reviews. With the grammar, the framework gracefully handles exceptions and alternative productions. A layered approach is used to separate the specification of design process and execution parameters. One of the main advantages of this separation is freeing designers from the over- specification and graceful exception handling. The framework, implemented using Java, is open and extensible. New process, tools, and user-defined process knowledge and constraints can be added easily.