Abstract:The message passing system is crucialto the computational performance of workstation clusters．With the fast growth in the performance of networking device，the unnecessary overhead in traditionalTCP／IPprotocolhas become the bottleneck of communication speed．To fully exploitthe communication potentialofthefast hardware，this paper introduces FMP，a new communication protocol specifically designed for Myrinet． FMP，which consists of a network part and a localpart，is a concise protocol．The physicallayer is assumed tobe error free and thus the protocolbecomes extremely simple and highly efficient．Compared with TCP／IP，theperformance ofFMPis much better．In this article the essentialtechniques used in the protocolare discussed andsome results are presented．

Abstract:Mathematically a 2Dconstrained design system can be modeled by m independent nonlinear equa－tions with n design variables and the design process can be viewed as a process ofsolving a geometric constraintsystem．Design decomposition is a highly effective way to improve a geometric constraint solver to make it effi－cient and robust．This paper reports a graph based decomposing approach and gives the correctness proof oftheapproach：（1）this approach can dealwith the decomposition ofstructurally under－constrained systems，（2）thisapproach can detect structurally over－constrained systems，（3）the approach can terminate within finite numberofsteps，and（4）the solving steps obtained through the decomposing approach are structurally consistent．

Abstract:Popularization and acceptance of electronic commerce mainly depend on the following properties：security，atomicity，privacy and anonymity．There are no electronic commerce protocols appropriate forelectron－ic transactions ofphysicalgoods in which three properties are needed：security，atomicity and privacy．An elec－tronic commerce modelis suggested in this paper．The modelis named ELCwhich simulates L／Cin internationaltrade．Then a secure and atomic electronic commerce protocolis proposed．Finally the protocolis analyzed foritsstrength and correctness by proving the desired properties using BAN style logic in the presence of an intruder．

Abstract:This paper discusses how to completely debug indeterminate MPI／PVM parallelprograms．Due tothe indeterminacy，the previous bugs may be non－repeatable in successive executions during a cyclic debuggingsession．Based on the FIFO communication model of MPI／PVM，an implementation of record and replay tech－nique is presented．Moreover，users are provided with an easy way to completely debug their programs by cover－ing all possible execution paths through controllable replay．Comparied with other solutions，the proposedmethod produces much less temporaland spatialoverhead．The implementation has been completed on two kindsof message passing architectures：one is Dawning－2000 super server（that was developed by the National Re－search Center for Intelligent Computing Systems ofChina）with single－processor（PowerPC）nodes which are in－terconnected by a custom－built wormhole mesh network；the other is a cluster ofworkstations（PowerPC／AIX）which has been built in NationalHigh Performance Computing Center at Hefei．

Abstract:The problem on liveness judgementis stillopen in Petrinets．The paper studies liveness on Asym－metric Choice nets（AC nets）by structure analysis theory．Firstly，some known results on liveness are dis－cussed．Then a sufficient condition by S－invariant for the liveness ofACnets is presented and the correspondingpolynomial－time algorithm is got．Finally a simple necessary－and－sufficientcondition on monotonicity ofboundedACnets is presented．

Abstract:Integrating infrastructure is the basic supporting environmentforCIM（computer integrated manu－facturing）systems．This paperintroduces the design and implementation ofan opening distributed CIM integrat－ing infrastructure，HIT－IIS．2，developed by HIT（Harbin Institute of Technology）．The HIT－IIS．2 prototypeincludes the following service functions for CIM systems：business process control，enterprise activity control，resource management，definition and description，system information，communication managementand front－endinterface．Each service function is described in the paper．

Abstract:The Internet is undergoing substantial changes from a communication and browsing infrastructure to a medium for conducting business and services. These changes require that Web servers should support the preferential services such as those in E-commerce and also achieve good fairness among different types of requests. In this paper, some integrated schemes of request dispatching and selecting for Web server clusters is provided, with the goal of achieving load balance and meeting the requirements of both WebQoS and fairness. The stochastic high level Petri net (SHLPN) models for them are given. To cope with the well-known state space explosion problem, this paper a novel approximate analysis technique is proposed, which can significantly reduce the complexity of the model solution. Moreover, the numerical results of the performance analysis for those schemes are presented, and the best one among them is recommended that is suited to the preferential services and can achieve high system performance for Web server clusters.

Abstract:In order to improve the throughput, response speed and scalability of Web servers efficiently, many famous Web sites have thrown away single servers and turned to Web server clusters. Web server clusters with different distribution modes of data copy have different data availability levels. After discussing many data copy distribution modes, an optimized data copy distribution is proved in this paper.

Abstract:After changes are made to a previously tested program, a goal of regression testing is to retest the program based only on the modification while maintaining the same testing coverage as the complete retesting of the program. Several regression techniques for structural programs using data flow or partial data flow (also known as program slicing) have been proposed. With the object oriented (OO for short) method growing mature, there is an urgent need of testing techniques for OO programs. In this paper, based on the analysis of characteristics of OO system, firstly, the dependency relations among objects are defined, from which series of objects' methods are deduced. Secondly, program slicing techniques are used to identify test cases to be applied to modification. Finally, a complete regression approach to OO program is presented.

Abstract:News video analysis is an important field of video analysis. In this paper a knowledge-based analysis method for news video——a two-stage template matching method-to detect the anchorperson shot of news video is presented, which is a fundamental step for segmenting news video units. The method is both generic and real-time, and can be used in automatic news video analyzing and indexing systems.

Abstract:Most existing algorithms of three-dimensional distance transform are extensions of two-dimensional approximate Euclidean distance transform algorithms such as the city block/chessboard. Such algorithms can only get the approximate Euclidean distance. A new method of three-dimensional true Euclidean distance transform is presented in this paper. The proposed method can get the true Euclidean distance with time complexity O(n3*log n). Moreover, this method is used to render the soft tissue in three-dimensional medical CT images, and good result has been obtained.

Abstract:The study of parallelism for Java program is one of the most important subjects at present. In this paper, a kind of automatic parallel conversion technique is given. The serial Java program is transformed to parallel program utilizing the multithreading machanism and testing technique of commutativity operations. The parallel program after transforming can run in the supercomputer with multi-CPU under the multi-processor operating system, which will enhance the programs' efficiency.

Abstract:Composition is essential in the research of formal specifications. It facilitates the design, synthesis, testing and reuse of large programs. In order to facilitate users to specify the temporal specification at the time of authoring a multimedia program, a composition model is needed. Recently various such models have been proposed in literature, which include language-based model, graphical model, time-interval based model, object-oriented model, etc. However, most of these models are lowly abstract and it is difficult to support the composition of two multimedia programs. In this paper, by introducing a new concept unit stream and extending two temporal relations, the structuring technique is studied in order to allow complex multimedia programs to be structured, making them understandable for designers, and to allow multimedia programs to be authored in terms of composition of general purpose multimedia programs that can be re-used in authoring various different multimedia programs.

Abstract:HLA (high level architecture) is a new generation of technical framework for distributed interactive simulation which supports simulation interoperability and reusability. But distributed interactive simulations under Internet environment face the challenge of system scalability due to the limited resources of bandwidth and computation. HLA provides the possibility for system scalability through data distribution management (DDM). In this paper, the technical approach to DDM is discussed and a new strategy is proposed for multicast address allocation. This new strategy uses a hierarchical method to allocate addresses to solve the problem of contradiction between number of multicast addresses supported by hosts and the redundant data received by hosts. The new strategy also supports for the reliable transferring and data bundling of simulation data.

Abstract:The rough logic defined on neighbor-valued decision tables and its truth values of the formulas are discussed in this paper. It is more general than the decision logic defined by Pawlak in the applications of data reduction. At present, the methods used are often Pawlak's data analysis and Skowron's discernible matrix methods. The former is informal, no ease mechanization. The latter is intuitive, easy to understand, but it requires to generate a medial link of discernible matrix, to make unnecessary expenses on time and space. Therefore, in the paper, one side extracts the attributes of attribute neighbor-valued discernible from the neighbor-valued decision table and discernible Conjunctive Normal Form is constructed. The other side simplifies the formula to use absorbable laws and other calculus of logical formulas. It obtains directly all reductions in the neighbor-valued decision table. Since it doesn't need to generate the medial link of discernible matrix, so it can spare space and time, and raise the efficiency of the program run. Thus, reduction of the tables is handled to possess 6 attributes (4 conditional attributes and 2 decision attributes) and 102 objects to use two methods respectively, and to obtain the same results. It uses one side to extract formulas from the tables, and the other side to reduce the formulas in DELPHI 3.0 on PⅠⅠ 233/64 M. The time of program running is about 1 minute 54 seconds; while time of spending is about 1 minute 55 seconds to use the discernible matrix method. Due to the increase of an array (discernible matrix), its space degree of complexity is O(m×n²), where m is the number of attributes, n is the number of objects. So, the space and time occupied will also increase rapidly along with the increment of attributes and objects. The strong points and shortcomings of two methods are quite clear from space and time used.

Abstract:The fundamental matrix is a basic tool in the analysis of scenes taken with two uncalibrated cameras. The 8-point algorithm and the improved 8-point algorithm are widely used linear methods for estimating the fundamental matrix. They have advantages of simplicity in implementation. But they are extremely sensitive to noise and outliers. Hence in most cases, they are useless virtually. A new robust linear method——weighted normalization algorithm is developed by introducing a cost function related to residual errors. Firstly, the matching points with a weight factor are normalized. Secondly, the eight parameters of fundamental matrix are calculated by using the 8-point algorithm. Experiments on simulated and real image data are conducted. The results show that this algorithm is very robust to noises and outliers, and the fundamental matrix with high accuracy can be found.

Abstract:In this paper, an object-oriented common data model PDMM (petroleum data model for management) is presented, which is constructed for petroleum resource management domain. PDMM is characterizedly its strong capacities of describing data with its definition of superclasses “activity” and “property”. The constructing techniques in PDMM include meta-model construction and its expressions, class organization, entity definition, as well as attributes and constraints. The representation of the model in EXPRESS language is also discussed.

Abstract:The multidegree reduction of Bézier curves is studied in this paper. The elevation property of Bézier curves themselves is combined with the least squares theory of generalized inverse matrices, and a new approximation method of multidegree reduction is presented. This method overcomes the weakness of the general method, which can reduce only one degree for each time, and obtains good approximation effects.

Abstract:Based on the object-oriented data structure of Voronoi diagram for normal curvilinear polygon, an improved Divide-and-Conquer algorithm is presented in this paper for Voronoi diagram generation called Main-Chain Divide-and-Conquer algorithm. It is easier to implement compared with the classical one. Meanwhile, because the boundary of a Voronoi cell contains parabolic or hyperbolic curves in the Euclidean metric, it is difficult to compute the Voronoi cell area in practice. For this reason, an area computation lemma of the Voronoi cell is presented. The lemma proof and an example are also given. Thus, a way is provided for the area calculation in some engineering applications.

Abstract:In this paper, a new gene learning algorithm for optimum search problem is proposed, which extended the binary population-based incremental learning (PBIL) and selfish algorithm (SA) by allowing a gene's allele to be multi-valued. In this new algorithm, the entropy of probability distribution as used as the criterion of termination, and the evolution process is combined with local heuristic search. Three typical combinatorial optimization problems (maximum cut problem, scheduling problem and travelling salesman problem) are solved and some results are better than the best result of existing algorithm.

Abstract:In this paper, based on the discussion of SASOS (single-address space operating system), as well as its advantages and disadvantages, the concept and principle of NASOS (no-address space operating system) is put forword and a NASOS prototype is discussed. In the NASOS, there is no process virtual address space, instructions access files directly and processes run on files. Compared with SASOS, not only does NASOS have the advantages of SASOS, but also avoids its disadvantages.

Abstract:In this paper, a kind of dynamic genetic algorithm based on the neural network with constraints is presented, which combines the local searching ability of neural network with the global searching ability of genetic algorithm. The dynamic algorithm is used to decide the initial point of the neural network, and the neural network is employed to decide the fitness of the dynamic genetic algorithm. The proposed algorithm has some theoretical and biological meanings. Compared with the standard genetic algorithm, it can decrease the searching scale and get the approximate solution of quadratic programming problems that are non-definite.

Abstract:Automatic video segmentation, which is based on SBD (shot boundary detection), is one necessary step in establishing video databases. Existing SBD algorithms can detect abrupt scene changes quite accurately but may fail for gradual scene changes. This is due to the fact that there are no apparent peak values of frame difference during gradual scene change periods and hence it is more difficult to detect them. Wipes are frequently used to edit video in spatial domain to generate various kinds of scene changes. Through analysis of various wipes, a new algorithm is proposed in this paper based on a spatial editing model for wipe scene change detection. This method is better than the statistic-based method proposed by Alattar. Test results for video sequences generated by Adobe Premiere 5.1 containing wipe scene changes show that the algorithm is well suited for various types and durations of wipes.