It has been proofed that Reconfigurable Computing System possesses the potential to promote system efficiency. Fine-Grain Reconfigurable Computing System, which integrates the co-design of software and hardware, is the prevailing current in system designing with low clock rate and high efficiency. The thesis focuses on computing-oriented Fine-Grain Reconfigurable Computing System of software-hardware interwork, and setting up invoking working model for software program and hardware module as well. The designing of software and hardware can be integrated into one program, which deals with the hardware as a function. By quick computing of hardware, it can promote system efficiency. This interwork mechanism can be combined into traditional instruction execute pipeline. It composes manageable Fine-Grain for Reconfigurable Computing System as a new processor architecture, which brings up new command to support execution of software-hardware interwork mechanism. The model architecture is verified by algorithms of multimedia application, such as Motion Estimation and DCT. In addition, it presents the optimizing model by analysis and comparison of software and hardware efficiency.

At present the scale of multimedia and communication systems has become more and more complicated due to the fast development of them. In order to improve the capability of real-time processing and shorten system development time, the ability to reconfigure system architecture becomes an important and flexible design consideration. In this thesis, we propose a reconfigurable processing unit, FMRPU, which is a fine-grain multi-context reconfigurable processing unit targeting at high-throughput and data-parallel applications. It contains 64 reconfigurable logic arrays, 16 switch boxes, and connects with each other via three hierarchical-level connectivities. To avoid the excessive routing path to be the bottleneck of mapped circuits, we design the data stream switch to rearrange data streams. According to the simulation results, the longest routing path of FMRPU only takes 6.5 ns at 0.35 processes, which is able to construct the required logic circuit efficiently. Compare with same kind devices in dealing with Motion Estimation operations, the performance is raise to 17% and is excellent to other same kind architectures in executing other DSP algorithms.

none

none

This thesis proposes a new approach to formalize refactorings, principally at the UML class diagram design level (but incorporating a limited amount of code-level information—basic access-related information). A set of abstract and atomic fine-grain transformations (FGTs) is defined as prototypical building blocks for constructing refactorings. The semantics of each FGT is specified in terms of its pre- and postcondition conjuncts. Various logical relationships between FGT pre- and postcondition conjuncts are fully catalogued. These include uni- and bidirectional sequential dependency relationships; absorbing and cancelling reduction relationships; and uni- and bi-directional conflict relationships. The principle container for FGTs is an FGT-list in which the ordering of FGTs respects the sequential relationships between them. Such a list is characterised by the set of FGT precondition conjuncts (which a system should satisfy if the FGTs are to be sequentially applied to the system) as well as the resulting postcondition conjuncts (that describe the effect of applying the list). In the thesis, twenty-nine commonly used primitive refactorings are specified as such FGT-lists, together with their associated FGT-enabling precondition conjuncts. Refactoring-level pre- and postconditions are also identified for each primitive refactoring FGT-list. These are, of course, required to guarantee behaviour preservation.<p. An alternative container for FGTs is defined, called an FGT-DAG. It is a directed acyclic graph with FGTs as nodes, and with arcs that reflect the sequential dependency relationships between constituent FGTs. An algorithm is provided to convert a list of FGTs into a corresponding set of FGT-DAGs. Thus design level refactorings specified as FGT-lists can be also be converted to corresponding sets of FGT-DAGs. The precondition for applying such a refactoring to a given system is specified at two levels: the FGT-enabling precondition conjuncts that apply to each FGT-DAG, and the refactoring-level precondition conjuncts. The thesis provides various algorithms that operate on FGT-DAGs. These include an algorithm to remove redundancies from an FGT-DAG. It also includes algorithms that operate on the elements of a set of FGT-DAGs: to detect sequential dependencies between these elements, to detect whether they are in deadlock, and to detect and possibly remove or modify FGTs causing conflicts between them. In addition, an algorithm is provided to build composite refactorings from primitive refactorings. It indicates how composite-level and FGT-enabling precondition conjuncts can be derived and utilised to avoid the rollback problem. A Prolog prototype FGT-based refactoring tool has been implemented. The tool stores all of the above-mentioned catalogued information as Prolog rules and facts. This includes the twenty-nine commonly used primitive refactorings (stored as Prolog FGT-lists) and their associated refactoring-level pre- and postcondition conjuncts. The tool also implements all the previously mentioned algorithms as Prolog procedures. The thesis thus establishes the foundations for a tool in which end users can create (and apply without rollback) not only composite refactorings, but also completely new refactorings whose semantics is constrained only by the fine-grained semantics of FGTs, rather than by the more course-grained semantics of primitive refactorings. Furthermore, using FGTs as refactoring building blocks (i.e. instead of primitive refactorings) means that redundancies and conflicts can be more accurately pin-pointed and removed; and opportunities for parallel execution are exposed at a more fine-grained level. These advantages come at the cost of having to carry out more computations because analysis has to take place at the FGT-level rather than at the refactoring-level. / Thesis (PhD)--University of Pretoria, 2009. / Computer Science / unrestricted

Equal channel angular extrusion (ECAE) has attracted a substantial attention for it provides the opportunity to introduce large plastic strain into the material in the bulk form. Both die angles and processing routes have been recognized as the important parameters in applying ECAE to fabricate ultrafine-grained materials. Unfortunately, studies of different group provided inconsistent conclusions on the effectiveness of processing routes, which are believed to be due to the incomplete microstructural information obtained in each investigation. In the present work, quantitative analysis of the microstructure developed by different processing conditions were conducted using transmission electron microscopy (TEM), in which the morphology, size, and shape of subgrains as well as boundary misorientation were fully characterized. A commercial pure aluminum (AA 1050) was deformed by ECAE to strain of ~ 8 with different routes (A, Bc and C, in terms of reorientation angle 0o, 90o, and 180o respectively of the billet between two extrusion passes) and die angles. The results show that the effectiveness of high angle boundary (HAB) formation is in the sequence of route A¡ÜBc>C. However, in terms of grain refinement, the effectiveness is in the order of route Bc>A>C. In addition, route A produces subgrains with the most elongated shape, while route Bc produces subgrains with the most equiaxed shape. These results may be attributed to the different shear pattern introduced in each route. ECAE die angle determines both the strain per pass and the shear plane orientation. In route C, the shear is maintained in the same plane and the effect of strain per pass can be studied. With route C, both the 90o and 120o die produce microstructure with similar HAB proportions, but they result in different arrangement of HABs. The 120o die produces subgrains with larger size and higher aspect ratio than the 90o die does in route C. Generally speaking, for the die angle range studied, the different values of strain per pass used in ECAE mainly affect the morphology of the subgrains. On the other hand, the effect of die angle is weakened with route Bc as compared to route C, which may be attributed to the intersection of shear planes involved in route Bc.

No description available.

Dans le contexte mécatronique actuel, la simulation en temps réel de la dynamique de systèmes multicorps complexes ainsi que la portabilité des modèles demeurent un challenge. La génération symbolique est le seul moyen d'obtenir des modèles dynamiques efficaces et utilisables dans la plus grande variété d'environnements de calcul. La contribution essentielle de cette thèse est de montrer que l'utilisation conjointe de l'approche symbolique et de la technique du partitionnement des coordonnées permet de produire des modèles dynamiques très efficaces pour l'analyse des systèmes multicorps complexes, dont la topologie contient des boucles cinématiques. Nous déterminons une partition optimale des coordonnées en vue d'obtenir une factorisation, sous forme bloc triangulaire, de la sous-matrice Jacobienne des contraintes associée aux coordonnées dépendantes. Nous calculons explicitement les coordonnées dépendantes, nous produisons un ensemble réduit d'équations du mouvement purement différentielles et nous les résolvons de manière complètement symbolique. Les modèles de systèmes multicorps complexes ainsi générés contiennent un nombre d'opérations nettement moindre que ceux obtenus par les approches numériques pourtant plus répandues, ce qui permet de les utiliser pour des applications temps réel. Ayant pour objectif une réduction maximale du temps de calcul, nous avons étudié deux méthodes de parallélisation automatique par voie symbolique. L'une exploite le taux de parallélisme important présent au niveau des opérations arithmétiques, et vise des architectures parallèles spécifiques. L'autre se base sur la topologie du système pour découper les modèles en vue de leur exécution sur des ordinateurs parallèles conventionnels. Les développements effectués dans le cadre de ce travail ont été implémentés dans le logiciel ROBOTRAN, un outil de génération symbolique d'équations développé dans notre laboratoire et dédicacé à l'étude des systèmes multicorps.

In this talk, we will present the design and evaluation of PeerDB, a peer-to-peer (P2P) distributed data sharing system. PeerDB distinguishes itself from existing P2P systems in several ways. First, it is a full-fledge data management system that supports fine-grain content-based searching. Second, it facilitates sharing of data without shared schema. Third, it combines the power of mobile agents into P2P systems to perform operations at peers' sites. Fourth, PeerDB network is self-configurable, i.e., a node can dynamically optimize the set of peers that it can communicate directly with based on some optimization criterion. / Singapore-MIT Alliance (SMA)

Because of its low density and high strength, magnesium (Mg) and its alloys are being sought after in the automotive industry for structural applications. Although many road-going cars today contain cast Mg parts, in the fabrication of chassis structural members the wrought alloys are required. One of the challenges of fabrication with wrought Mg is welding and joining the formed sheets. Because of the commonly observed difficulties in fusion welding of Mg such as hot cracking and severe Heat Affected Zone (HAZ), this work aimed to establish the feasibility of the solid-state process Magnetic Pulse Welding in producing lap welds of Mg sheet. Mg AZ31 alloy sheets have been lap-welded with Magnetic Pulse Welding (MPW), an Impact Welding technique, using H-shaped symmetric coils connected to a Pulsar MPW-25 capacitor bank. MPW uses the interaction between two opposing magnetic fields to create a high speed oblique collision between the metal surfaces. The oblique impact sweeps away the contaminated surface layers and forces intimate contact between clean materials to produce a solid-state weld. Various combinations of similar and dissimilar metals can be welded using MPW. Other advantages of MPW are high speed, high strength, and the possibility of being mounted on a robotic arm. The present research focuses on the feasibility and mechanical performance of an MPW weld of 0.6 mm AZ31 Mg alloy sheets made in a lap joint configuration. Tensile shear tests were carried out on the joints produced. Load bearing capacity showed linear increase with capacitor bank discharge energy up to a certain value above which a parabolic increase was seen. Strength was estimated to be at least as high as base metal strength by measuring the fracture surface area of selected samples. The fracture surface of samples welded at higher discharge energy showed two regions. In the beginning of the bond a platelet-featured fracture brittle in appearance and a ductile, micro-voiding fracture in the latter part. The joint cross section morphology consisted of a flattened area that had two symmetric bond zones 1 mm wide each separated by an unbonded centre zone ~3mm wide. Reasons for the morphology were presented as a sequence of events based on the transient nature of the oblique collision angle. The interface microstructure was studied by optical and electron microscopy. Examination of the bonds has revealed sound and defect free interfaces. No microcracking, porosity, resolidification, or secondary phase formation was observed. Metallographic examination of the unbonded centre zone revealed anisotropic deformation and a lack of cleaning from the interface. This zone is formed as a result of normal impact in the initial stage of collision. The bond zone interface of the joint was characterized by a smooth interface consisting of refined grains. In samples welded at higher energy interfacial waves developed in the latter half of the bond zone. Transmission electron microscopy (TEM) of the bond zone revealed a continuous interface having an 8-25 μm thick interlayer that coincided with the waves and had a dislocation cell structure and distinct boundaries with adjacent material. Equiaxed 300 nm dynamic recrystallized (DRX) grains were found adjacent to the interlayer as well as other slightly larger elongated grains. The interlayer is thought to be formed in plasticized state at elevated temperature through severe shear strain heating. The interlayer corresponds to a ductile fracture surface and, along with the interfacial waves, is responsible for the joint’s high strength.