Sistemas de predição de desempenho de cadeias de suprimento são constituídos por indicadores que visam estimar o desempenho da empresa-foco em decorrência também do desempenho dos indicadores dos fornecedores. Na literatura são encontrados apenas dois modelos quantitativos (GANGA; CARPINETTI, 2011; AGAMI; SALEH; RASMY, 2014) que permitem predizer o desempenho de cadeias de suprimento usando os indicadores do modelo SCOR® (Supply Chain Operations Reference). Uma limitação de ambos modelos é a dificuldade de se ajustar ao ambiente de uso, uma vez que sua implementação e atualização requerem a parametrização manual de muitas regras de decisão. Tanto o uso de redes neurais quanto de sistemas neuro-fuzzy têm o potencial de contornar essa dificuldade por utilizarem um mecanismo de aprendizagem que possibilita a adaptação ao ambiente de uso usando dados numéricos. Todavia, na literatura não são encontradas aplicações dessas técnicas no apoio à predição de desempenho de cadeias de suprimento, tampouco estudos que discutam qual dessas técnicas se mostra mais adequada para lidar com este problema. Diante disso, o objetivo desta pesquisa é construir e a avaliar a adequabilidade de dois sistemas de predição de desempenho, ambos baseados nos indicadores do modelo SCOR®, mas usando alternativamente as técnicas redes neurais e sistemas neuro-fuzzy, para apoiar a gestão de desempenho da empresa-foco e de sua cadeia imediata. A execução desta pesquisa envolveu o uso de simulação computacional e de testes estatísticos. Os resultados mostram que, embora ambas as técnicas apresentem capacidade de predição satisfatória, as redes neurais são mais adequadas em relação à complexidade da definição da configuração topológica, enquanto os sistemas neuro-fuzzy se sobressaíram em relação à capacidade de predição, complexidade do treinamento, quantidade de variáveis de entrada, suporte à tomada de decisão sob incerteza e interpretabilidade dos dados. Outros resultados desta pesquisa estão relacionados à identificação de particularidades do processo de modelagem das técnicas avaliadas, à elaboração de um panorama sobre o uso de técnicas quantitativas na avaliação de desempenho de cadeias de suprimento e à identificação de algumas oportunidades de pesquisa. / Supply chain performance prediction systems are composed by indicators that aim to estimate the performance of a focal company considering also indicators related to their suppliers. There are two quantitative models in the literature (GANGA; CARPINETTI, 2011; AGAMI; SALEH; RASMY, 2014) that enable to predict the supply chain performance using the indicators proposed by the SCOR® model (Supply Chain Operations Reference). Nevertheless, there is a drawback of both models that refers to the difficulty in adapting to the environment of use, since implementation and updating of these models require parameterization of many decision rules that must be done by an expert. The application of artificial neural networks as well as neuro-fuzzy systems can overcome this drawback by using a learning mechanism that enables the adaptation to the environment of use using numerical data on supply chain performance. However, there are neither studies in the literature that propose the use of these techniques in order to support supply chain performance prediction nor studies that discuss which of these techniques seem to be more appropriate to deal with this problem. Thus, the objective of this study is to propose and evaluate the adequability of the two types of performance prediction systems based on the performance indicators of the SCOR® model, and both using alternatively artificial neural networks and neuro-fuzzy systems to support performance management of a focal company and their supply chain. The implementation of this research involved the use of computer simulation and statistical tests. The results show that although both techniques present a satisfactory predictive capacity, neural networks are more appropriate in relation to the complexity of defining the topological configuration, whereas the neuro-fuzzy systems are more adequate regarding the predictive capacity, complexity of the training, amount of input variables, support to decision-making under uncertainty and interpretability of data. Other results of this research refer to the identification of characteristics of the modeling process of the evaluated techniques, as well as to the review on the use of quantitative techniques for supply chain performance evaluation and to the identification of some research opportunities.

Since the late 19th century, the average temperature on Earth has risen by approximately 1.1 °C because of the increased carbon dioxide (CO2) and other man-made emissions to the atmosphere. The transportation sector is responsible for approximately 33% of the global CO2 emissions and 14% of the overall greenhouse gas emissions. Therefore, increasingly stringent regulations in the European Union require CO2 emissions to be lower than 95 gCO₂/km by 2020. In this regard, improvements in internal combustion engines (ICEs)must be achieved in terms of fuel consumption and CO2 emissions. Given that only up to 35% of fuel energy is converted into mechanical power, the wasted energy can be reused through waste heat recovery (WHR) technologies. Consequently, organic Rankine cycle (ORC) has received significant attention as a WHR technology because of its ability to recover wasted heat in low- to medium-heat sources. The Expansion machine is the key component in ORC systems, and its performance has a direct and significant impact on overall cycle efficiency. However, the thermal efficiencies of ORC systems are typically low due to low working temperatures. Moreover, supersonic conditions at the high pressure ratios are usually encountered in the expander due to the thermal properties of the working fluids selected which are different to water. Therefore, this thesis aims to design an efficient radial-inflow turbine to avoid further efficiency reductions in the overall system. To fulfil this aim, a novel design and optimisation methodology was developed. A design of experiments technique was incorporated in the methodology toexplorethe effects of input parameters on turbine performance and overall size. Importantly, performance prediction modelling by means of 1D mean-line modelling was employed in the proposed methodology to examine the performance of ORC turbines at constant geometries. The proposed methodology was validated by three methods: computational fluid dynamics analysis, experimental work available in the literature, and experimental work in the current project. Owing to the lack of actual experimental works in ORC-ICE applications, a test rig was built around a heavy-duty diesel engine at Brunel University London and tested at partial load conditions due to the requirement for a realistic off-high representation of the performance of the system rather than its best (design) point, while taking into account the limitation of the engine dynamometer employed. Results of the design methodology developed for this projectpresented an efficient single-stage high-pressure ratio radial-inflow turbine with a total to static efficiency of 74.4% and an output power of 13.6 kW.Experimental results showed that the ORC system had a thermal efficiency of 4.3%, and the brake-specific fuel consumption of the engine was reduced by 3%. The novel meanlineoff designcode (MOC) was validated with the experimental works from three turbines. In comparison with the experimental results conducted at Brunel University London, the predicted and measured results were in good agreement with a maximum deviation of 2.8%.

Sistemas de predição de desempenho de cadeias de suprimento são constituídos por indicadores que visam estimar o desempenho da empresa-foco em decorrência também do desempenho dos indicadores dos fornecedores. Na literatura são encontrados apenas dois modelos quantitativos (GANGA; CARPINETTI, 2011; AGAMI; SALEH; RASMY, 2014) que permitem predizer o desempenho de cadeias de suprimento usando os indicadores do modelo SCOR® (Supply Chain Operations Reference). Uma limitação de ambos modelos é a dificuldade de se ajustar ao ambiente de uso, uma vez que sua implementação e atualização requerem a parametrização manual de muitas regras de decisão. Tanto o uso de redes neurais quanto de sistemas neuro-fuzzy têm o potencial de contornar essa dificuldade por utilizarem um mecanismo de aprendizagem que possibilita a adaptação ao ambiente de uso usando dados numéricos. Todavia, na literatura não são encontradas aplicações dessas técnicas no apoio à predição de desempenho de cadeias de suprimento, tampouco estudos que discutam qual dessas técnicas se mostra mais adequada para lidar com este problema. Diante disso, o objetivo desta pesquisa é construir e a avaliar a adequabilidade de dois sistemas de predição de desempenho, ambos baseados nos indicadores do modelo SCOR®, mas usando alternativamente as técnicas redes neurais e sistemas neuro-fuzzy, para apoiar a gestão de desempenho da empresa-foco e de sua cadeia imediata. A execução desta pesquisa envolveu o uso de simulação computacional e de testes estatísticos. Os resultados mostram que, embora ambas as técnicas apresentem capacidade de predição satisfatória, as redes neurais são mais adequadas em relação à complexidade da definição da configuração topológica, enquanto os sistemas neuro-fuzzy se sobressaíram em relação à capacidade de predição, complexidade do treinamento, quantidade de variáveis de entrada, suporte à tomada de decisão sob incerteza e interpretabilidade dos dados. Outros resultados desta pesquisa estão relacionados à identificação de particularidades do processo de modelagem das técnicas avaliadas, à elaboração de um panorama sobre o uso de técnicas quantitativas na avaliação de desempenho de cadeias de suprimento e à identificação de algumas oportunidades de pesquisa. / Supply chain performance prediction systems are composed by indicators that aim to estimate the performance of a focal company considering also indicators related to their suppliers. There are two quantitative models in the literature (GANGA; CARPINETTI, 2011; AGAMI; SALEH; RASMY, 2014) that enable to predict the supply chain performance using the indicators proposed by the SCOR® model (Supply Chain Operations Reference). Nevertheless, there is a drawback of both models that refers to the difficulty in adapting to the environment of use, since implementation and updating of these models require parameterization of many decision rules that must be done by an expert. The application of artificial neural networks as well as neuro-fuzzy systems can overcome this drawback by using a learning mechanism that enables the adaptation to the environment of use using numerical data on supply chain performance. However, there are neither studies in the literature that propose the use of these techniques in order to support supply chain performance prediction nor studies that discuss which of these techniques seem to be more appropriate to deal with this problem. Thus, the objective of this study is to propose and evaluate the adequability of the two types of performance prediction systems based on the performance indicators of the SCOR® model, and both using alternatively artificial neural networks and neuro-fuzzy systems to support performance management of a focal company and their supply chain. The implementation of this research involved the use of computer simulation and statistical tests. The results show that although both techniques present a satisfactory predictive capacity, neural networks are more appropriate in relation to the complexity of defining the topological configuration, whereas the neuro-fuzzy systems are more adequate regarding the predictive capacity, complexity of the training, amount of input variables, support to decision-making under uncertainty and interpretability of data. Other results of this research refer to the identification of characteristics of the modeling process of the evaluated techniques, as well as to the review on the use of quantitative techniques for supply chain performance evaluation and to the identification of some research opportunities.

The modeling and evaluation calculus FMC-QE, the Fundamental Modeling Concepts for Quanti-tative Evaluation [1], extends the Fundamental Modeling Concepts (FMC) for performance modeling and prediction. In this new methodology, the hierarchical service requests are in the main focus, because they are the origin of every service provisioning process. Similar to physics, these service requests are a tuple of value and unit, which enables hierarchical service request transformations at the hierarchical borders and therefore the hierarchical modeling. Through reducing the model complexity of the models by decomposing the system in different hierarchical views, the distinction between operational and control states and the calculation of the performance values on the assumption of the steady state, FMC-QE has a scalable applica-bility on complex systems. According to FMC, the system is modeled in a 3-dimensional hierarchical representation space, where system performance parameters are described in three arbitrarily fine-grained hierarchi-cal bipartite diagrams. The hierarchical service request structures are modeled in Entity Relationship Diagrams. The static server structures, divided into logical and real servers, are de-scribed as Block Diagrams. The dynamic behavior and the control structures are specified as Petri Nets, more precisely Colored Time Augmented Petri Nets. From the structures and pa-rameters of the performance model, a hierarchical set of equations is derived. The calculation of the performance values is done on the assumption of stationary processes and is based on fundamental laws of the performance analysis: Little's Law and the Forced Traffic Flow Law. Little's Law is used within the different hierarchical levels (horizontal) and the Forced Traffic Flow Law is the key to the dependencies among the hierarchical levels (vertical). This calculation is suitable for complex models and allows a fast (re-)calculation of different performance scenarios in order to support development and configuration decisions. Within the Research Group Zorn at the Hasso Plattner Institute, the work is embedded in a broader research in the development of FMC-QE. While this work is concentrated on the theoretical background, description and definition of the methodology as well as the extension and validation of the applicability, other topics are in the development of an FMC-QE modeling and evaluation tool and the usage of FMC-QE in the design of an adaptive transport layer in order to fulfill Quality of Service and Service Level Agreements in volatile service based environments. This thesis contains a state-of-the-art, the description of FMC-QE as well as extensions of FMC-QE in representative general models and case studies. In the state-of-the-art part of the thesis in chapter 2, an overview on existing Queueing Theory and Time Augmented Petri Net models and other quantitative modeling and evaluation languages and methodologies is given. Also other hierarchical quantitative modeling frameworks will be considered. The description of FMC-QE in chapter 3 consists of a summary of the foundations of FMC-QE, basic definitions, the graphical notations, the FMC-QE Calculus and the modeling of open queueing networks as an introductory example. The extensions of FMC-QE in chapter 4 consist of the integration of the summation method in order to support the handling of closed networks and the modeling of multiclass and semaphore scenarios. Furthermore, FMC-QE is compared to other performance modeling and evaluation approaches. In the case study part in chapter 5, proof-of-concept examples, like the modeling of a service based search portal, a service based SAP NetWeaver application and the Axis2 Web service framework will be provided. Finally, conclusions are given by a summary of contributions and an outlook on future work in chapter 6. [1] Werner Zorn. FMC-QE - A New Approach in Quantitative Modeling. In Hamid R. Arabnia, editor, Procee-dings of the International Conference on Modeling, Simulation and Visualization Methods (MSV 2007) within WorldComp ’07, pages 280 – 287, Las Vegas, NV, USA, June 2007. CSREA Press. ISBN 1-60132-029-9. / FMC-QE (Fundamental Modeling Concepts for Quantitative Evaluation [1]) ist eine auf FMC, den Fundamental Modeling Concepts, basierende Methodik zur Modellierung des Leistungsverhaltens von Systemen mit einem dazugehörenden Kalkül zur Erstellung von Leistungsvorhersagen wie Antwortzeiten und Durchsatz. In dieser neuen Methodik steht die Modellierung der hierarchischen Bedienanforderungen im Mittelpunkt, da sie der Ursprung aller dienstbasierenden Systeme sind. Wie in der Physik sind in FMC-QE die Bedienanforderungen Tupel aus Wert und Einheit, um Auftragstransformationen an Hierarchiegrenzen zu ermöglichen. Da die Komplexität durch eine Dekomposition in mehreren Sichten und in verschiedene hierarchische Schichten, die Unterscheidung von Operations- und Kontrollzuständen, sowie dazugehörige Berechungen unter Annahme der Stationarität reduziert wird, skaliert die Anwendbarkeit von FMC-QE auf komplexe Systeme. Gemäß FMC wird das zu modellierende System in einem 3-dimensionalen hierarchischen Beschreibungsraum dargestellt. Die quantitativen Kenngrößen der Systeme werden in drei beliebig frei-granularen hierarchischen bi-partiten Graphen beschrieben. Die hierarchische Struktur der Bedienanforderungen wird in Entity Relationship Diagrammen beschrieben. Die statischen Bedienerstrukturen, unterteilt in logische und reale Bediener, sind in Aufbaudiagrammen erläutert. Außerdem werden Petri Netze, genauer Farbige Zeit-behaftete Petri Netze, dazu verwendet, die dynamischen Abläufe, sowie die Kontrollflüsse im System zu beschreiben. Anschließend wird eine Menge von hierarchischen Gleichungen von der Struktur und den Parametern des Modells abgeleitet. Diese Gleichungen, die auf dem stationären Zustand des Systems beruhen, basieren auf den beiden Fundamental Gesetzen der Leistungsanalyse, dem Gesetz von Little und dem Verkehrsflussgesetz. Das Gesetz von Little definiert hierbei Beziehungen innerhalb einer hierarchischen Schicht (horizontal) und das Verkehrsflussgesetz wiederum Beziehungen zwischen hierarchischen Schichten (vertikal). Die Berechungen erlauben Leistungsvorhersagen für komplexe Systeme durch eine effiziente Berechnung von Leistungsgrößen für eine große Auswahl von System- und Lastkonfigurationen. Innerhalb der Forschungsgruppe von Prof. Dr.-Ing Werner Zorn am Hasso Plattner Institut an der Universität Potsdam ist die vorliegende Arbeit in einen größeren Forschungskontext im Bereich FMC-QE eingebettet. Während hier ein Fokus auf dem theoretischen Hintergrund, der Beschreibung und der Definition der Methodik als auch der Anwendbarkeit und Erweiterung gelegt wurde, sind andere Arbeiten auf dem Gebiet der Entwicklung einer Anwendung zur Modellierung und Evaluierung von Systemen mit FMC-QE bzw. der Verwendung von FMC-QE zur Entwicklung einer adaptiven Transportschicht zur Einhaltung von Dienstgüten (Quality of Service) und Dienstvereinbarungen (Service Level Agreements) in volatilen dienstbasierten Systemen beheimatet. Diese Arbeit umfasst einen Einblick in den Stand der Technik, die Beschreibung von FMC-QE sowie die Weiterentwicklung von FMC-QE in repräsentativen allgemeinen Modellen und Fallstudien. Das Kapitel 2: Stand der Technik gibt einen Überblick über die Warteschlangentheorie, Zeit-behaftete Petri Netze, weitere Leistungsbeschreibungs- und Leistungsvorhersagungstechniken sowie die Verwendung von Hierarchien in Leistungsbeschreibungstechniken. Die Beschreibung von FMC-QE in Kapitel 3 enthält die Erläuterung der Grundlagen von FMC-QE, die Beschreibung einiger Grundannahmen, der graphischen Notation, dem mathematischen Modell und einem erläuternden Beispiel. In Kapitel 4: Erweiterungen von FMC-QE wird die Behandlung weiterer allgemeiner Modelle, wie die Modellklasse von geschlossenen Netzen, Synchronisierung und Mehrklassen-Modelle beschrieben. Außerdem wird FMC-QE mit dem Stand der Technik verglichen. In Kapitel 5 werden Machbarkeitsstudien beschrieben. Schließlich werden in Kapitel 6 eine Zusammenfassung und ein Ausblick gegeben. [1] Werner Zorn. FMC-QE - A New Approach in Quantitative Modeling. In Hamid R. Arabnia, editor, Proceedings of the International Conference on Modeling, Simulation and Visualization Methods (MSV 2007) within WorldComp ’07, 280 – 287, Las Vegas, NV, USA, Juni 2007. CSREA Press. ISBN 1-60132-029-9.

GPGPU Computing using CUDA is rapidly gaining ground today. GPGPU has been brought to the masses through the ease of use of CUDA and ubiquity of graphics cards supporting the same. Although CUDA has a low learning curve for programmers familiar with standard programming languages like C, extracting optimum performance from it, through optimizations and hand tuning is not a trivial task. This is because, in case of GPGPU, an optimization strategy rarely affects the functioning in an isolated manner. Many optimizations affect different aspects for better or worse, establishing a tradeoff situation between them, which needs to be carefully handled to achieve good performance. Thus optimizing an application for CUDA is tough and the performance gain might not be commensurate to the coding effort put in. I propose to simplify the process of optimizing CUDA programs using a CUDA Performance Analyzer. The analyzer is based on analytical modeling of CUDA compatible GPUs. The model characterizes the different aspects of GPU compute unified architecture and can make prediction about expected performance of a CUDA program. It would also give an insight into the performance bottlenecks of the CUDA implementation. This would hint towards, what optimizations need to be applied to improve performance. Based on the model, one would also be able to make a prediction about the performance of the application if the optimizations are applied to the CUDA implementation. This enables a CUDA programmer to test out different optimization strategies without putting in a lot of coding effort.

A two-dimensional Boundary Element Method (BEM) is developed to study the strongly nonlinear interaction between a surface-piercing body and the free-surface. The scheme is applied to problems with and without the possibility of ventilation resulting from the motion and geometric configuration of the surface-piercing body. The main emphasis of this research work is on the development of numerical methods to improve the performance prediction of surface-piercing propellers by including the whole range of free-surface nonlinearities. The scheme is applied to predict the ventilated cavity shapes resulting from the vertical and rotational motion of a blade-section with fully nonlinear free-surface boundary conditions. The current method is able to predict the ventilated cavity shapes for a wide range of angles of attack and Froude numbers, and is in good agreement with existing experimental results. Through a comparison with a linearized free-surface method, the current method highlights the shortcomings of the negative image approach used commonly in two-dimensional and three-dimensional numerical methods for surface-piercing hydrofoils or propellers. The current method with all its capabilities makes it a unique contribution to improving numerical tools for the performance prediction of surface-piercing propellers. The scheme is also applied to predict the roll and heave dynamics of two-dimensional Floating Production Storage and Offloading (FPSO) vessel hull sections within a potential flow framework. The development of the potential flow model is aimed at validating the free-surface dynamics of an independently developed Navier Stokes Solver for predicting the roll characteristics of two-dimensional hull sections with bilge keels. / text

In this thesis, we design, develop and implement parallel algorithms for irregular problems on heterogeneous multi-core architectures. Irregular problems exhibit random and unpredictable memory access patterns, poor spatial locality and input dependent control flow. Heterogeneous multi-core processors vary in: clock frequency, power dissipation, programming model (MIMD vs. SIMD), memory design and computing units, scalar versus vector units. The heterogeneity of the processors makes designing efficient parallel algorithms for irregular problems on heterogeneous multicore processors challenging. Techniques of mapping tasks or data on traditional parallel computers can not be used as is on heterogeneous multi-core processors due to the varying hardware. In an attempt to understand the efficiency of futuristic heterogeneous multi-core architectures on applications we study several computation and bandwidth oriented irregular problems on one heterogeneous multi-core architecture, the IBM Cell Broadband Engine (Cell BE). The Cell BE consists of a general processor and eight specialized processors and addresses vector/data-level parallelism and instruction-level parallelism simultaneously. Through these studies on the Cell BE, we provide some discussions and insight on the performance of the applications on heterogeneous multi-core architectures. Verifying these experimental results require some performance modeling. Due to the diversity of heterogeneous multi-core architectures, theoretical performance models used for homogeneous multi-core architectures do not provide accurate results. Therefore, in this thesis we propose an analytical performance prediction model that considers the multitude architectural features of heterogeneous multi-cores (such as DMA transfers, number of instructions and operations, the processor frequency and DMA bandwidth). We show that the execution time from our prediction model is comparable to the execution time of the experimental results for a complex medical imaging application.

In this thesis, we design, develop and implement parallel algorithms for irregular problems on heterogeneous multi-core architectures. Irregular problems exhibit random and unpredictable memory access patterns, poor spatial locality and input dependent control flow. Heterogeneous multi-core processors vary in: clock frequency, power dissipation, programming model (MIMD vs. SIMD), memory design and computing units, scalar versus vector units. The heterogeneity of the processors makes designing efficient parallel algorithms for irregular problems on heterogeneous multicore processors challenging. Techniques of mapping tasks or data on traditional parallel computers can not be used as is on heterogeneous multi-core processors due to the varying hardware. In an attempt to understand the efficiency of futuristic heterogeneous multi-core architectures on applications we study several computation and bandwidth oriented irregular problems on one heterogeneous multi-core architecture, the IBM Cell Broadband Engine (Cell BE). The Cell BE consists of a general processor and eight specialized processors and addresses vector/data-level parallelism and instruction-level parallelism simultaneously. Through these studies on the Cell BE, we provide some discussions and insight on the performance of the applications on heterogeneous multi-core architectures. Verifying these experimental results require some performance modeling. Due to the diversity of heterogeneous multi-core architectures, theoretical performance models used for homogeneous multi-core architectures do not provide accurate results. Therefore, in this thesis we propose an analytical performance prediction model that considers the multitude architectural features of heterogeneous multi-cores (such as DMA transfers, number of instructions and operations, the processor frequency and DMA bandwidth). We show that the execution time from our prediction model is comparable to the execution time of the experimental results for a complex medical imaging application.

La programmation parallèle et les algorithmes data-parallèles sont depuis plusieurs décennies les principales techniques de soutien l'informatique haute performance. Comme toutes les propriétés non-fonctionnelles du logiciel, la conversion des ressources informatiques dans des performances évolutives et prévisibles implique un équilibre délicat entre abstraction et automatisation avec une précision sémantique. Au cours de la dernière décennie, de plus en plus de professions ont besoin d'une puissance de calcul très élevée, mais la migration des programmes existants vers une nouvelle configuration matérielle et le développement de nouveaux algorithmes à finalité spécifique dans un environnement parallèle n'est jamais un travail facile, ni pour les développeurs de logiciel, ni pour les spécialistes du domaine. Dans cette thèse, nous décrivons le travail qui vise à simplifier le développement de programmes parallèles, en améliorant également la portabilité du code de programmes parallèles et la précision de la prédiction de performance d'algorithmes parallèles pour des environnements hétérogènes. Avec ces objectifs à l'esprit, nous avons proposé un modèle de transition nommé SGL pour la modélisation des architectures parallèles hétérogènes et des algorithmes parallèles, et une mise en œuvre de squelettes parallèles basés sur le modèle SGL pour le calcul haute performance. SGL simplifie la programmation parallèle à la fois pour les machines parallèles classiques et pour les nouvelles machines hiérarchiques. Il généralise les primitives de la programmation BSML. SGL pourra plus tard en utilisant des techniques de Model-Driven pour la génération de code automatique á partir d'une fiche technique sans codage complexe, par exemple pour le traitement de Big-Data sur un système hétérogène massivement parallèle. Le modèle de coût de SGL améliore la clarté de l'analyse de performance des algorithmes, permet d'évaluer la performance d'une machine et la qualité d'un algorithme / Parallel programming and data-parallel algorithms have been the main techniques supporting high-performance computing for many decades. Like all non-functional properties of software, the conversion of computing resources into scalable and predictable performance involves a delicate balance of abstraction and automation with semantic precision. During the last decade, more and more professions require a very high computing power. However, migrating programs to new hardware configuration or developing new specific-purpose algorithms on a parallel environment is never an easy work, neither for software developers nor for domain specialists. In this thesis we describe work that attempts to improve the simplicity of parallel program development, the portability of parallel program code, and the precision of parallel algorithm performance prediction for heterogeneous environments. With these goals in mind we proposed a bridging model named SGL for modelling heterogeneous parallel architectures and parallel algorithms, and an implementation of parallel skeletons based on SGL model for high-performance computing. SGL simplifies the parallel programming either on the classical parallel machines or on the novel hierarchical machines. It generalizes the BSML programming primitives. SGL can be served later with model-driven techniques for automatic code generation from specification sheet without any complex coding, for example processing Big Data on the heterogeneous massive parallel systems. The SGL cost model improves the clarity of algorithms performance analysis; it allows benchmarking machine performance and algorithm quality

No description available.