Formal Refinement Checking in a System-level Design Methodology

  • Authors:

  • Jean-Pierre Talpin;Paul Le Guernic;Sandeep Kumar Shukla;Frédéric Doucet;Rajesh Gupta

  • Affiliations:

  • Inria-Irisa, Campus de Beaulieu, 35042 Rennes, France;Inria-Irisa, Campus de Beaulieu, 35042 Rennes, France;Virginia Tech, Electrical and Computer Engineering, Blacksburg, VA 24061, USA;University of California at San Diego, Department of Computer Science and Engineering AP&M 3111, 9500 Gilman Drive, La Jolla, CA 92093-0114, USA;University of California at San Diego, Department of Computer Science and Engineering AP&M 3111, 9500 Gilman Drive, La Jolla, CA 92093-0114, USA

  • Venue:
  • Fundamenta Informaticae - Application of Concurrency to System Design (ACSD'03)
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Rising complexity, increasing performance requirements, and shortening time-to-market demands necessitate newer design paradigms for embedded system design. Such newer design methodologies require raising the level of abstraction for design entry, reuse of intellectual property blocks as virtual components, refinement based design, and formal verification to prove correctness of refinement steps. The problem of combining various components from different designers and companies, designed at different levels of abstraction, and embodying heterogeneous models of computation is a difficult challenge for the designer community today. Moreover, one of the gating factors for widespread adoption of the system-level design paradigm is the lack of formal models, method and tools to support refinement. In the absence of provably correct and adequate behavioral synthesis techniques, the refinement of a system-level description towards its implementation is primarily a manual process. Furthermore, proving that the implementation preserves the properties of the higher system-level design-abstraction is an outstanding problem. In this paper, we address these issues and define a formal refinement-checking methodology for system-level design. Our methodology is based on a polychronous model of computation of the multi-clocked synchronous formalism SIGNAL. This formalism is implemented in the POLYCHRONY workbench. We demonstrate the effectiveness of our approach by the experimental case study of a SPECC modeling example. First, we define a technique to systematically model SPECC programs in the signal formalism. Second, we define a methodology to compare system-level models of SPECC programs and to validate behavioral equivalence relations between these models at different levels of abstraction. Although we use SPECC modeling examples to illustrate our technique, our methodology is generic and language-independent and the model that supports it conceptually minimal by offering a scalable notion and a flexible degree of abstraction.