Skip to Main Content
 

Global Search Box

 
 
 
 

ETD Abstract Container

Abstract Header

High Level Design Methodology for Reconfigurable Systems

Ding, Mingwei

Abstract Details

2005, Doctor of Philosophy (PhD), Ohio University, Electrical Engineering & Computer Science (Engineering and Technology).

The rise of reconfigurable computing systems presents great challenges to traditional electronic system design mainly due to the dynamic reconfigurability requirements of these systems. At the same time, the chip capacity is growing at a faster pace than that of design productivity, bringing up the problem of a “productivity crisis.” Thus, there is a strong demand in the microelectronic design industry for electronic design automation tools and design paradigms that will improve the design productivity and utilize the full power of reconfigurable architectures.

This dissertation addresses these needs and includes research on high-level design methodology for reconfigurable systems and studies applications of such systems.

Two ongoing projects – SOLAR and DRAW – which are dedicated to machine learning and next generation mobile station design, respectively, were selected as examples of reconfigurable system applications.

First of all, a novel design methodology and design language, Unified Algorithmic Design Language (UADL), have been proposed and developed. UADL allows the designer to enter a high-level algorithmic description once and automatically generate executable code in different target languages, such as Matlab and VHDL. The proposed UADL tool has been applied to the design process of both projects with about 2 to 3 times savings in code sizes of their algorithmic parts.

Regarding the selected projects’ needs, the following topics have been studied: optimum interconnection, reconfigurable routing, mapping of the Turbo decoder algorithm and dimensionality reduction. An optimum interconnection scheme has been derived and a novel pipeline structure has been devised to realize the reconfigurable routing with linear hardware costs. Dimensionality reduction is an indispensable step for an efficient learning, feature extraction, classification, and image/pattern processing. Two postmapping algorithms have been developed and tested against real world data.

An important contribution of this dissertation is the proposed new design methodology based on UADL. The UADL methodology provides a high-level, unified, tool-independent design interface for system designers, as Java has provided a platform independent programming language for software developers. It is my hope and desire that this design paradigm will gain such popularity as Java did.

Janusz Starzyk (Advisor)
147 p.

Recommended Citations

Citations

  • Ding, M. (2005). High Level Design Methodology for Reconfigurable Systems [Doctoral dissertation, Ohio University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1131117147

    APA Style (7th edition)

  • Ding, Mingwei. High Level Design Methodology for Reconfigurable Systems. 2005. Ohio University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1131117147.

    MLA Style (8th edition)

  • Ding, Mingwei. "High Level Design Methodology for Reconfigurable Systems." Doctoral dissertation, Ohio University, 2005. http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1131117147

    Chicago Manual of Style (17th edition)