Model-Integrated Development Environment for Adaptive Computing

Institute for Software Integrated Systems
Vanderbilt University

Introduction:

The goals of this research are to develop high-level system design tools for the construction of dynamically reconfigurable, high performance embedded systems using adaptive computing technology. The tools will implement a domain-specific environment that is customized to the application arena.

The figure below summarizes project goals and approach. Our approach is to apply Model-Integrated Computing (MIC) technology, (developed under the DARPA Evolutionary Design of Complex Software program) to the domain of Adaptive Computing Systems. The MIC environment has been customized for this domain, and extended with hardware and software synthesis mechanisms.  The result is the Adaptive Computing Systems Model-Integrated Development Environment (ACSMIDE), pictured below.

 The MIDE consists of  Model Builder tools, which allow the designer to construct graphical models that specify system design information, Generators, which interpret the models and generate executable computational structures (software and hardware), and a Runtime Environment.

The graphical models capture design information in a domain-specfic format.  The MIDE is customized to capture this information from 3 related perspectives:

• Behavioral Models specify the system modes and the conditions for transition between modes.
• Algorithm/Structural Models capture the computational structure of the algorithms that perform the necessary mathematical and contol operations.  Rather capturing a fixed data flow, the structural algorithms can represent design alternatives, creating a flexible design space for potential system implementations.
• Resource Models represent the target hardware architecture and capabilities.

These three modeling perspectives are augmented with a constraint language to manage design space and algorithm selection/resource mapping.

The Generators process the models to assess system behavioral modes, system implementation possibilities, and system resources.

For additional information, please contact Ted Bapty at bapty@vuse.vanderbilt.edu or at (615) 343-6709.