. The model-checking problem for 1-safe Petri nets and lineartime temporal logic (LTL) consists of deciding, given a 1-safe Petri net and a formula of LTL, whether the Petri net satisfies the property encoded by the formula. This paper introduces a semidecision test for this problem. By a semidecision test we understand a procedure which may answer `yes', in which case the Petri net satisfies the property, or `don't know'. The test is based on a variant of the so called automata-theoretic approach to model-checking and on the notion of T-invariant. We analyse the computational complexity of the test, implement it using 2lp -- a constraint programming tool, and apply it to two case studies. This paper is a (very) abbreviated version of [6]. 1 Introduction Linear-time temporal logic (LTL) is a well-known formalism for specifying properties of concurrent systems. The problem of deciding if a concurrent system satisfies a LTL formula is called the model-checking problem (of LTL). In [16] ...

Manufacturers now have the capability to build high performance multiprocessor machines with common PC components. This has created a new market of low cost multiprocessor machines. However, these machines are handicapped unless they have an operating system that can take advantage of their underlying architectures. Presented is a comparison of two such operating systems, Windows NT and Solaris. By focusing on their implementation of threads, we show each system's ability to exploit multiprocessors. We report our results and interpretations of several experiments that were used to compare the performance of each system. What emerges is a discussion on the performance impact of each implementation and its significance on various types of applications. 1. Introduction A few years ago, high performance multiprocessor machines had a price tag of $100,000 and up, see [16]. The multiprocessor market consisted of proprietary architectures that demanded a higher cost due to the scale of econ...

: There are two fundamental themes in constraint programming. One is discrete or finite domain constraint programming based on the constraint satisfaction model. The other is continuous constraint programming based on linear programming and its extensions. In this paper we propose techniques for making constraint solvers cooperate, and we present a scheduling application of the Dutch Railways and a new kind of algorithm for solving disjunctive programming problems, one which could not be developed without cooperating solvers. What emerges is that cooperating solvers, which have old roots in special purpose operations research methods, constitute a central technology with wide applicability. 1.1 INTRODUCTION While discrete constraint satisfaction methods may be ideally suited to solve some problems (e.g. job-shop scheduling problems) and continuous methods to solve others (e.g. airline crew scheduling problems), there are also exam1 ples where each method on its own exhibits certain ...

Machine architecture, see [36], [1]. This architecture is commonly called the WAM. The 2LP interpreter employs an appropriately engineered version of the WAM, which is adapted to support the types int, double and continuous; it is a virtual machine which is emulated in C. A technical point that might be of interest: in the implementation of 2LP of [30], the linear programming module of 2LP does not stack simplex bases for backtracking; instead, a trail of changes to the polytope is kept and the changes are undone as needed. A 2LP program is compiled into code for this virtual machine. The theoretical analysis that underlies the 2LP implementation can be found in [13]. Languages which support logic programming constructs and constraints are called constraint logic programming languages or CLP languages, [23]. The first K. McAloon and C. Tretkoff /Logic, Modeling, and Programming 21 to support constraints on continuous variables were Prolog III, [12] and CLP(R), [25]. For work on CLP ...

. The key idea in Modern Portfolio Theory, as introduced by Markowitz, is to minimize risk which is defined as mean-variance in returns. One method to solve the resulting quadratic programming problem is by converting it into a linear complementarity problem and then solving the latter using specialized algorithms. An analysis reveals that a linear complementarity problem can be understood as a disjunctive linear programming problem which can be modeled and solved using constraint programming systems such as 2LP or ILOG. Representing the problem in this way facilitates the inclusion of additional constraints (integrality requirements, mutually exclusive stocks, personal preferences etc.) which might be required given the specific needs and wishes of an investor. Interesting issues that arise in solving the resulting constraint programs include the use of cooperating constraint solvers and identifying appropriate search strategies. 1 INTRODUCTION The key idea in Modern Portfolio Theor...

The contributions during the last few years on the structural theory of Petri nets can now be applied to formal veri#cation. The structural theory provides methods to #nd e#cient encoding schemes for symbolic representations of the reachable markings. It also provides approximations of the state space that allow to alleviate many bottlenecks in the calculation of the reachability set by breadth or depth #rst search algorithms. The paper reviews some of the results on the structural theory and explains how they can be incorporated inamodel-checking veri#cation framework for concurrent systems. 1

: Researchers have been actively pursuing load balancing schemes for parallel searches in an attempt to achieve linear or near linear speedups. Most of the approaches have used message based distributed models. Although, they have been successfully ported to shared-memory systems, their designs and approaches are counter intuitive and cumbersome for symmetric multiprocessing architectures. Here, we present an asynchronous load balancing scheme that is a natural programming model for symmetric multiprocessor machines. We test the scheme on two classic problems using different search strategies and find it effective in obtaining linear or near linear speedups. Keywords: load balancing, threads, parallelism. 1. INTRODUCTION Parallel searches have proven to be very beneficial for the Artificial Intelligence and Operations Research communities [1][2][3]. Large combinatorial problems, which otherwise are impractical for a single processor, can be solved by exploiting the parallelism of m...

. Liveness properties of parallel systems usually specify that in every execution certain states are eventually reached. Therefore, violation of such a property can only be detected in infinite executions. In this paper we introduce a semi-decision method that is based on structural Petri net analysis and makes use of the constraint programming paradigm. By a semi-decision method we understand a procedure which may answer `yes', so that in this case the parallel system satisfies the property, or `don't know'. We give an implementation of our method in terms of the constraint programming tool 2lp. An application of our approach to a snapshot algorithm demonstrates how constraint programming can beat classical exact methods such as model checking. Keywords: Verification, Petri nets, constraint programming, finite-state systems. 1 Introduction In the field of verification the automatic verification of finite-state parallel systems has become a very interesting and important aspect of co...

Manufacturers now have the capability to build high performance multiprocessor machines with common PC components. This has created a new market of low cost multiprocessor machines. However, these machines are handicapped unless they have an operating system that can take advantage of their underlying architectures. Presented is a comparison of two such operating systems, Windows NT and Solaris. By focusing on their implementation of threads, we show each system’s ability to exploit multiprocessors. We report our results and interpretations of several experiments that were used to compare the performance of each system. What emerges is a discussion on the performance impact of each implementation and its significance on various types of applications. 1.

We consider the situation where the decision procedure of an interactive expert system must repeatedly solve a logic minimization problem. In that setting, the response time is slow if the instances of the logic minimization problem are difficult to solve. We propose an alternate process. It utilizes a logic problem decomposition and replaces some instances of the logic minimization problem by instances of satisfiability subproblems. In practice, instances of satisfiability subproblems tend to be much easier to solve than instances of the original logic minimization problem. Due to that fact, the decomposition approach can be expected to produce a substantial reduction of the required computing effort. We have used the approach in an interactive expert system for the optimal management of industrial chemical exposure. Due to the decomposition approach, the user typically experiences no perceptible waiting during the interactive query phase. Keywords: expert system, logic decomposition...