FORMAL METHODS AND MODELS FOR SYSTEM DESIGN

Transcription

1 FORMAL METHODS AND MODELS FOR SYSTEM DESIGN

2 Formal Methods and Models for System Design A System Level Perspective Edited by Rajesh Gupta University 0/ California at San Diego Paul Le Guernic INRIA-IRISA Sandeep Kumar Shukla Virginia Tech and Jean-Pierre Talpin INRIA-IRISA SPRINGER - SCIENCE+BUSINESS MEDIA, B.V.

3 A C.I.P. Catalogue record for this book is available from the Library of Congress. ISBN DOI / ISBN (ebook) Printed on acid-free paper All Rights Reserved 2004 Springer- Science+Business Media Dordrecht Origina11y published by Kluwer Academic Publishers in 2004 Softcover reprint of the hardcover Ist edition 2004 No part of this work may be reproduced, stored in a retrievai system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

4 Contents Preface vii Part I Methods and Models for System Level Design 1 Modular Hierarchies of Models for Embedded Systems Manfred Broy 2 Actor-oriented models for codesign Edward A. Lee, and Stephen Neuendorffer 3 Structural Component Compusition for System-level Models Frederic Doucet, Sandeep Shukla, Rajesh Gupta, 4 Truly Heterogeneous modeling with SystemC Hiren D. Patel and Sandeep K. Shukla 5 MoDe: A Method for System-Level Architecture Evaluation fan Romberg, fewgenij Botaschanjan and Oscar Slotosch Part 11 Models and Methods for System Validation 6 A Verification Methodology for Concurrent Software with Synchronous 137 Communication Christoph Sprenger and KrzysztofWorytkiewicz 7 High Level Verification of Control Intensive Systems 159 Edmund Clarke, Oma Grumberg, Muralidhar Talupur, and Dong Wang 8 How to Compute the Refinement Relation far Parameterized Systems Franroise Bellegarde, Cilina Charlet and Olga Kouchnarenko 181

5 vi FORMAL METHODS AND MODELS FOR SYSTEM DESIGN Part m Type Theoretic Models and Methods for System Design 9 Algebraic theory for behavioral type inference Jean-Pierre Talpin and Paul Le Guemic Behavioral type inference for compositional system design 245 Jean-Pierre Talpin, David Bemer, Sandeep Kumar Shukla, Paul Le Guemic, Abdoulaye Gamatie, and Rajesh Gupta Part N Optimizing System Models 11 Optimizations for Faster Execution of Esterel Programs Dumitru Potop-Butucaru and Robert de Simone 12 Optimizing System Models for Simulation Efficiency Shekhar A. Sharad, and Sandeep Kumar Shukla Capturing Formal Specification into Abstract Models 331 David Bemer, Syed Suhaib, Sandeep Kumar Shukla, and Jean-Pierre Talpin, Part V Post-Production Formal Methods 14 Engineering Changes in Field Modifiable Architectures 353 Satoshi Komatsu, Hiroshi Saito, Kenshu Seto, Yoshihisa Kojima, and Masahiro Fujita

6 Preface Perhaps nothing characterizes the inherent heterogeneity in embedded systems than the ability to choose between hardware and software implementations of a given system function. Indeed, most embedded systems at their core represent a careful division and design of hardware and software parts of the system functionality. To do this task effectively, models and methods are necessary to capture application behavior, needs and system implementation constraints. Formal modeling can be valuable in addressing these tasks. As with most engineering domains, co-design practice defines the state of the art, though it seeks to add new capabilities in system conceptualization, modeling, optimization and implementation. These advances - particularly those related to synthesis and verification tasks - direct1y depend upon formal understanding of system behavior and performance measures. Current practice in system modeling relies upon exploiting high-level programming frameworks, such as SystemC, EstereI, to capture design at increasingly higher levels of abstraction and attempts to reduce the system implementation task. While raising the abstraction levels for design and verification tasks, to be really useful, these approaches must also provide for reuse, adaptation of the existing intellectual property (IP) blocks. Currentindustrial trends in use of general-purpose language facilities - driven by the popularity of C/C++/Java and other high-levellanguages - provide an interesting and somewhat urgent backdrop to explore corresponding advances in formal models to reason with these high-level descriptions, to help understand the design, implement formal correctness proofs and predict performances and other metrics. It is important to bridge the gap between theoretical models of computations and the practical experience in general purpose programming based design. Doing so will benefit both the languages (from the rigor of models) and models (from the experience of programming practice). This contrasted picture led to debates and discussions about bringing theory and practice together in an informal mailing list in 2002 which inc1uded various industry experts, and academic researchers. These discussions gave rise to the ACMlIEEE Conference on Formal Models and Methods for Co-Design (MEMOCODE), first organized at Mt. Saint Michel, France, in June This

7 viii FORMAL METHODS AND MODELS FOR SYSTEM DESIGN conference was organized to bring together practitioners of formal methods in the hardware and software industry, and academic researchers to create dialogue on the issues of convergence of formal methods with practice, as weh as to create constructive exchange of ideas and views of people applying formal methods in software and those applying it to hardware design. Through extensive on-line discussions, five areas were identified. These areas were (i) Conformance checking between designs at various abstraction levels, (ii) Hierarchical verification to facilitate verification of designs expressed hierarchically without flattening the hierarchy, (iii) Abstraction vs. optimization, as they lower level optimizations affect abstraction hierarchy, (iv) Incremental verification to harness the existing verification results for effectively verify incrementally changed design, and (v) Post-production patchability of designs and capturing such requirements in formal models. These are by no means the only problems that confront us, but is a good sampling of areas that the members of the MEMOCODE community feit as crucial problems. The idea of this edited volume arose through these discussions and propelled by two very interesting panel discussions at the conference that pointed out the need for documenting challenges and advances focused on high-level modeling and validation. Our goal in putting this volume is to provide a serious technical discussion of the advances in formal modeling and methodologies for embedded systems. The chapters of this book were invited from the attendees of the first MEMOCODE conference, comp1emented by articles written by the editors representing their own views and contributions to this problem area. As mentioned already, the topics addressed in this book are chosen in order to encompass possib1e areas of formal methods that are believed to be useful in making the general purpose programming based methodology more successful and efficient. For organization purposes, we have divided the chapters into five major parts. The first part consists of papers on modeling methodologies, frameworks, and one case study on using formal modeling framework for system design. This part is meant to introduce the reader to the need for formalism and frameworks, the need for behavioral hierarchy, and heterogeneity support, and how certain frameworks such as Ptolemy II, BALBOA, or SystemC enhancements address these. Part two of this volume focus on validation methodologies. The chapters in this part address the issue of abstraction and refinement, and automation

8 PREFACE IX of abstraction and refinement in detail. With the state-space explosion problem hampering the success of formal verification tools, we believe automated abstraction, refinement, and modular proof techniques would be crucial for successful use of formal verification in the industry. This part of the book refiects this view. Also abstract modeling itself is an arduous task, and hence a chapter on a methodology about incremental building of formal models has been added, to illustrate one possible way to tackling the model building problem. Part three introduces how type theory, in particular behavioral type theory is gaining crucial importance in system integration, and composition of preverified IP components in a correct by construction manner. Type inference at the composition time could avoid a number of integration bugs, as is known from statically typed programming languages. However, unlike software procedures and functions, dynamic actor oriented concurrent models require that we extend type theory with behavioral types, and this part will provide the readers with some introduction to such theories. Often times, system level models grow in size, and hence simulation or even implementation synthesis become unmanageable. Under such circumstances, model optimization may lead to efficient simulation or implementation. Part four of this volume introduces two chapters, one of optimizing Esterel programs for efficient execution, and another on SystemC model optimization. These represent a very small sampie of this important research area. Part five has a single paper on formal models to capture the requirements for post production patching of fabricated designs. To our knowledge there has not been much research done on this topic, and hence this paper stands by itself in part five. Overall, we believe that this book will provide a reasonable sampling of the various topics relevant to formal methods in System Design, as understood and advocated by the MEMOCODE community. We invite our reader to participate and contribute to this growing community of researchers. More information on MEMOCODE is available at Rajesh Gupta Paul Le Guernic Sandeep Shukla Jean-Pierre Talpin