CAREER: Designing Embedded Systems with Domain-Specific Languages

Columbia University, New York NY

Investigators

Abstract

Designing Embedded Systems with Domain-Specific Languages Stephen A. Edwards This goal of this project is to provide tools to designers of embedded systems that will allow them to more quickly create correct systems. It proposes to do this by providing domain-specific languages and compilers. Such succinct, abstract languages reduce development and testing time by providing more formal, correct-by-construction methodologies and by enabling automatic implementation techniques. The result will enable the development of these systems to proceed beyond the current ad hoc C-and-assembly approach. The project addresses three issues in the design of many embedded systems: real-time software, software for communication, and software for communicating across the hardware/software boundary. These issues are addressed through the following specific investigations: New Compilation Techniques for Synchronous Languages Synchronous languages such as Esterel provide very predictable real-time behavior but are challenging to compile. This project includes developing new, more efficient compilation techniques for large systems described in the Esterel language and its variants. New Languages and Compilers for Communication Protocols Network communication is becoming a crucial part of every embedded system, but the growing complexity of the protocols and the need for bandwidth (e.g., for video) makes this software difficult to develop. This project includes the development of a new language for describing such protocol software and a compiler for producing efficient implementations. New Languages and Compilers for Device Drivers Writing software that communicates directly with hardware is currently tedious and error-prone, and unfortunately even more important in embedded systems since they tend to have application-specific peripherals. This project aims to simplify this task by creating a language and compiler for writing device drivers. The results of this research will enable more diverse languages for communicating across the hardware/software boundary. This project is also addressing the recurring complaint that academia does not produce students who understand both hardware and software well enough to be effective embedded systems designers by developing new courses along these lines. At the undergraduate level, a new embedded systems project course is being developed that will give students the opportunity to learn about and use some of these new techniques. At the graduate level, a more sophisticated course is being developed that focuses on languages for describing embedded systems and isses with their semantics and implementation. This will give future designers more tools for developing embedded systems and will prepare students for research in automating embedded system design. The results of this project are being disseminated in three ways: through conference and journal papers, through the release of all developed software to the public domain, and through the education of students in the use of the languages and compilers developed. Just as computer-aided design tools have enabled the rapid, correct design of fantastically complex integrated circuits, this work is enabling the quick, correct design of tomorrow's complex embedded systems.