ITR: Design Space Exploration and Synthesis for Multiple-Mode Embedded Systems
University Of Texas At Dallas, Richardson TX
Investigators
Abstract
Sha abstract The complexity of an embedded system is growing rapidly from single mode to multiple modes for achieving adaptability but the algorithms and tools for enhancing design productivity are much under-developed. This ITR project is aimed to develop new techniques and design tools for architectural-level synthesis, code generation, and design space exploration of multimode embedded systems. The goal of multimode system synthesis is to determine the best architecture (including processors with various types, special hardware components, memory, interconnects, etc.) with the minimum cost and generate the optimal codes and schedules for each mode and the transitions of modes. In order to build up a solid foundation of this new area, the project conducts a fundamental study to understand properties and develop provable algorithms to achieve the following objectives: (1) Multiple objective optimization. The project develops optimization algorithms considering multiple objectives, and understands the relationship among them. Scheduling and combined optimization techniques such as retiming, unfolding, and code-size minimization are developed. (2) Design space minimization and exploration. The project utilizes the obtained relationships among various design criteria, derives strict lower bounds and upper bounds, and minimizes the design space to explore. (3) Multiple-mode synthesis. The project analyzes multiple mode system behavioral models, and studies the global optimization and synthesis of multimode system.
View original record on NSF Award Search →