In this thesis we consider the Lanczos algoritm and its behaviour in finite precision. Having summarized theoretical properties of the algorithm and its connection to orthogonal polynomials, we recall the idea of the Lanczos method for approximating the matrix eigenvalues. As the behaviour of the algorithm is strongly influenced by finite precision arithmetic, the linear independence of the Lanczos vectors is usually lost after a few iterations. We use the most im- portant results from analysis of the finite precision Lanczos algorithm according to Paige, Greenbaum, Strakos and others. Based on that, we study formulation and properties of the mathematical model of finite presicion Lanczos computati- ons suggested by Greenbaum. We carry out numerical experiments in Matlab, which support the theoretical results.

Les dépendances fonctionnelles fournissent une information sémantique sur les données d’une table en mettant en lumière les liens de corrélation qui les unient. Dans cette thèse, nous traitons du problème de l’extraction de ces dépendances en proposant un contexte unifié permettant la découverte de n’importe quel type de dépendances fonctionnelles (dépendances de clé, dépendances fonctionnelles conditionnelles, que la validité soit complète ou approximative). Notre algorithme, ParaCoDe, s’exécute en parallèle sur les candidats, réduisant ainsi le temps global de calcul. De ce fait, il est très compétitif vis-à-vis des approches séquentielles connues à ce jour. Les dépendances satisfaites sur une table nous servent à résoudre le problème de la matérialisation partielle du cube de données. Nous présentons une caractérisation de la solution optimale dans laquelle le coût de chaque requête est borné par un seuil de performance fixé préalablement et dont la taille est minimale. Cette spécification de la solution donne un cadre unique pour décrire et donc comparer formellement les techniques de résumé de cubes de données. / Functional dependancies provide a semantic information over data from a table to exhibit correlation links. In this thesis, we deal with the dependancy discovery problem by proposing a unified context to extract any type of functional dependencies (key dependencies, conditional functional dependencies, with an exact or an approximate validity). Our algorithm, ParaCoDe, runs in parallel on candidates there by reducing the global time of computations. Hence, it is very competitive comparated to sequential appoaches known today. Satisfied dependencies on a table are used to solve the problem of partial materiali-zation of data cube. We present a characterization of the optimal solution in which the cost of each query is bounded by a before hand fixed performance threshold and its size is minimal. This specification of the solution gives a unique framework to describe and formally compare summarization techniques of data cubes.

Aerodynamic shape optimization and multidisciplinary optimization algorithms have the potential not only to improve conventional aircraft, but also to enable the design of novel configurations. By their very nature, these algorithms generate and analyze a large number of unique shapes, resulting in high computational costs. In order to improve their efficiency and enable their use in the early stages of the design process, a fast and robust flow solution algorithm is necessary. This thesis presents an efficient parallel Newton-Krylov-Schur flow solution algorithm for the three-dimensional Navier-Stokes equations coupled with the Spalart-Allmaras one-equation turbulence model. The algorithm employs second-order summation-by-parts (SBP) operators on multi-block structured grids with simultaneous approximation terms (SATs) to enforce block interface coupling and boundary conditions. The discrete equations are solved iteratively with an inexact-Newton method, while the linear system at each Newton iteration is solved using the flexible Krylov subspace iterative method GMRES with an approximate-Schur parallel preconditioner. The algorithm is thoroughly verified and validated, highlighting the correspondence of the current algorithm with several established flow solvers. The solution for a transonic flow over a wing on a mesh of medium density (15 million nodes) shows good agreement with experimental results. Using 128 processors, deep convergence is obtained in under 90 minutes. The solution of transonic flow over the Common Research Model wing-body geometry with grids with up to 150 million nodes exhibits the expected grid convergence behavior. This case was completed as part of the Fifth AIAA Drag Prediction Workshop, with the algorithm producing solutions that compare favourably with several widely used flow solvers. The algorithm is shown to scale well on over 6000 processors. The results demonstrate the effectiveness of the SBP-SAT spatial discretization, which can be readily extended to high order, in combination with the Newton-Krylov-Schur iterative method to produce a powerful parallel algorithm for the numerical solution of the Reynolds-averaged Navier-Stokes equations. The algorithm can efficiently solve the flow over a range of clean geometries, making it suitable for use at the core of an optimization algorithm.

Aerodynamic shape optimization and multidisciplinary optimization algorithms have the potential not only to improve conventional aircraft, but also to enable the design of novel configurations. By their very nature, these algorithms generate and analyze a large number of unique shapes, resulting in high computational costs. In order to improve their efficiency and enable their use in the early stages of the design process, a fast and robust flow solution algorithm is necessary. This thesis presents an efficient parallel Newton-Krylov-Schur flow solution algorithm for the three-dimensional Navier-Stokes equations coupled with the Spalart-Allmaras one-equation turbulence model. The algorithm employs second-order summation-by-parts (SBP) operators on multi-block structured grids with simultaneous approximation terms (SATs) to enforce block interface coupling and boundary conditions. The discrete equations are solved iteratively with an inexact-Newton method, while the linear system at each Newton iteration is solved using the flexible Krylov subspace iterative method GMRES with an approximate-Schur parallel preconditioner. The algorithm is thoroughly verified and validated, highlighting the correspondence of the current algorithm with several established flow solvers. The solution for a transonic flow over a wing on a mesh of medium density (15 million nodes) shows good agreement with experimental results. Using 128 processors, deep convergence is obtained in under 90 minutes. The solution of transonic flow over the Common Research Model wing-body geometry with grids with up to 150 million nodes exhibits the expected grid convergence behavior. This case was completed as part of the Fifth AIAA Drag Prediction Workshop, with the algorithm producing solutions that compare favourably with several widely used flow solvers. The algorithm is shown to scale well on over 6000 processors. The results demonstrate the effectiveness of the SBP-SAT spatial discretization, which can be readily extended to high order, in combination with the Newton-Krylov-Schur iterative method to produce a powerful parallel algorithm for the numerical solution of the Reynolds-averaged Navier-Stokes equations. The algorithm can efficiently solve the flow over a range of clean geometries, making it suitable for use at the core of an optimization algorithm.

Multi-core architectures are present throughout a large selection of computing devices from cell phones to super-computers. Parallel applications running on these devices solve bigger problems in a shorter time. Writing those applications is a difficult task for programmers. They need to deal with low-level parallel mechanisms such as data distribution, inter-processor communication, and task placement. The goal of the Standard Template Adaptive Parallel Library (STAPL) is to provide a generic high-level framework to develop parallel applications. One of the first steps of a parallel application is to partition and distribute the data throughout the system. An important data structure for parallel applications to store large amounts of data and model many types of relations is the graph. A mesh, which is a special type of graph, is often used to model a spatial domain in scientific applications. Graph and mesh partitioning has many applications such as VLSI circuit design, parallel task scheduling, and data distribution. Data distribution, significantly impacts the performance of a parallel application. In this thesis, we introduce the STAPL Parallel Graph Partitioner Framework. This framework provides a generic infrastructure to partition arbitrary graphs and meshes and to build customized partitioners. It includes the state of the art parallel k-way multilevel scheme to partition arbitrary graphs, a parallel mesh partitioner with parameterized partition shape, and a customized partitioner used for discrete ordinates particle transport computations. This framework is also part of a generic library, STAPL, allowing the partitioning of the data and development of the whole parallel application to be done in the same environment. We show the user-friendly interface of the framework and its scalability for partitioning different mesh and graph benchmarks on a Cray XE6 system. We also highlight the performance of our customized unstructured mesh partitioner for a discrete ordinates particle transport code. The developed columnar decompositions significantly reduce the execution time of simultaneous sweeps on unstructured meshes.

Polymers is a class of molecules which can have many different structures due to a large number of degrees of freedom. Many biopolymers, e.g. DNA, but also synthetic macromolecules have special structural features due to their backbone stiffness. Since such structural properties are important for e.g. the biological function, a lot of effort has been put into the investigation of the configurational properties of semi-stiff molecules. A theoretical treatment of these systems is often accompanied by computer simulations. The main idea is to compare theoretically derived models with experimental results for real polymers. Using Monte Carlo simulations, I have investigated how this computational technique can build a bridge between theoretical models and experimentally observed phenomena. The effort was mainly directed to develop sampling techniques, for efficiently exploring the configurational space of semi-stiff chains in a wide range of structures. The work was concentrated on compact conformations, since they, as is well known from previous studies, are difficult to sample using conventional methods. In my studies I have shown that the simple and, at a first glance, time consuming method of bead-by-bead regrow as a way of changing the configuration of a semi-stiff chain gave very promising and encouraging results when combined with modern simulation techniques, like Entropic Sampling with the Wang-Landau algorithm. The resulting simulation package was also suitable for parallelization which resulted in a further speed-up of the calculations. In addition to the more elaborate sampling methods, I also investigated external conditions to induce compaction of a semi-stiff polymer. In the case of a polyampholyte the condensing agent could be a multivalent salt, creating effective attraction between the loops of the chain, while for neutral polymers, an external field and the geometry of the confining volume can induce a compaction.

O trabalho constitui-se numa investigação teórica da estrutura ordenada e intuitiva dos espaços coerentes, introduzidos por Girard [GIR 86], na definição do modelo de máquina geométrica para construção e interpretação de estados e processos computacionais rotulados por posições de um espaço geométrico. Esta interpretação poderá ser aplicada às construções determinísticas, incluindo dois tipos especiais de paralelismo - o espacial, com memória e processos infinitos definidos por estruturas matriciais, que operam sobre dimensões independentes, de forma sincronizada; e o temporal, na versão genérica do modelo, com memória global transfinita e processos distribuídos num conjunto enumerável de máquinas geométricas, sincronizadas no tempo. O modelo contempla interpretação para computações não-determinísticas e prevê a aplicação de operadores exponenciais na interpretação do espaço funcional. A noção mais intuitiva deste trabalho está na definição da relação de coerência, que define o grafo sobre o qual se constrói este domínio semántico. Sobre o conjunto de pontos compatíveis de tais grafos, a coerência estrita interpreta a condição implícita para modelar o paralelismo - a concorrência entre posições de memória. Na construção dual, justificada pela presença da negação involutiva no grafo complementar, a incoerência interpreta a condição para o não-determinismo - o conflito de acesso à memória. Para os demais construtores, o produto sequencial e a soma determinística, consideram-se os endofunctores produto e soma direta da categoria CospLin dos espaços coerentes e funções lineares. A estrutura ordenada deste modelo é formalizada pelo espaço coerente D∞ de todos os processos, construído em níveis a partir do espaço coerente D∞ dos processos elementares, seguindo a metodologia proposta por Scott [SCO 76]. Neste sentido, cada nível da construção está identificado por um subespaço Dn que reconstrói todos os objetos do nível anterior, preservando suas propriedades e relações, além de construir os novos objetos. Compatível com a abordagem algébrica, o relacionamento entre os níveis é expresso por funções lineares denominadas imersões e projeções, interpretanto os construtores de processos e seus destrutores, respectivamente. Pelo procedimento de completação, assegura-se a existência do menor ponto fixo para equações recursivas definidas pela composição infinita destes morfismos. Além disso, as interpretações para processos infinitos, construídos por prefixação, apresentadas em D→∞ comprovam que este modelo é compatível com a diversidade dos construtores. O espa¸co coerente D∞2 dos processos transfinitos generaliza a construção e define a estrutura ordenada do modelo de máquina geométrica distribuída. Seus objetos são subconjuntos coerentes de tokens rotulados por posições do espaço geométrico e indexados por subconjuntos isomorfos aos ordinais transfinitos. O espaço coerente S S dos traços lineares de funções definidas sobre o espaço coerente S dos estados computacionais constitui-se no modelo semântico para análise do comportamento associado a cada processo interpretado em D∞. A definição da função de representação introduz um domínio de expressões que formaliza uma linguagem capaz de expressar, de forma mais operacional, as interpretações obtidas neste modelo de m´aquina. Cada uma das expressões válidas na linguagem é compatível com uma expressão gráfica. / This work presents a theoretical investigation of the constructive, intuitive and ordered structure of the coherence spaces, introduced by Girard, in order to define the geometric machine model for interpretation of computational states and processes labelled by positions of a geometric space. This interpretation can be applied to deterministic process constructions, including two special types of parallelism - the temporal parallelism, with infinite memory and infinite processes defined over array structures, that operate over independent dimensions in a synchronized way; and the spatial parallelism, in a generic version of the model, with a transfinite global memory shared by transfinite processes distributed in a enumerable set of geometric machines, synchronized in the time. The work also provides interpretation to the non-deterministic computations and applies the exponential operators in the interpretation of the functional space. The most basic notion of this work is the definition of the coherence relation as the admissibility of parallelism between basic operations (elementary processes). That relation defines the web over which the coherence space of the whole set of deterministic and non-deterministic processes is step-wise and systematically build. Over the set of the compatible points of such graph, the strict coherence interprets the implicity condition to model parallelism - the true concurrence. In the dual construction, justified by the presence of involutive negation in the complementary graph, the incoherence interprets the condition that models non-determinism - the conflict of memory accesses. The other constructors, the sequential product and the deterministic sum, are defined by the endofunctors in the CospLin category of the coherence spaces and linear functions. The ordered structure of this model is formalized by the coherence space D∞ of all processes, constructed by levels from the coherence space D0 of the elementary processes, following the Scott’s methodology [SCO 76]. In this sense, each level is identified by a subspace Dn, which reconstructs all the objects from the level before, preserving their properties and relations, and drives the construction of the new objects. Compatible with the algebraic-theoretic approach to computational processes, the relationship between the levels is expressed by linear functions called embedding and projection-functions, which interpret constructors and destructors of processes, respectively. The completion procedure guarantees the existence of the least fixed point to the recursive equations, defined by infinite composition of these morphisms. In addition, the interpretation for infinite processes constructed by prefix is presented in D→∞ , confirms that the ordered structure of these model is compatible with the diversity of constructors. The coherence space D∞2 of transfinite processes generalizes the construction and defines the ordered structure of the distributed geometric machine model. Its objects are coherent subsets of tokens labelled by the positions of a geometric space and indexed by isomorphic subsets related to the transfinite ordinal numbers. In order to analyze the behavior related to the interpretations in D∞, the coherence space S S of the linear traces of functions, defined over the coherence space S of the computational states, is introduced. The definition of the representation-function induces the construction of the domain Ω of valid expressions and formalizes a (graphic) language which is able to express, in an more operational way, the interpretations obtained in the geometric machine model.

Bien que la vulcanisation soit un procédé de réticulation très répandu dans l’industrie du caoutchouc, les mécanismes réactionnels mis en jeu, la structure du matériau formé, et en particulier les environnements chimiques du soufre restent mal connus. Sonder sélectivement les environnements chimiques du soufre par Résonance Magnétique Nucléaire (RMN) pourrait alors apporter de précieuses informations sur la structure locale du matériau. Cependant, les propriétés intrinsèques du seul isotope du soufre observable par RMN (33S) rendent l’étude de son environnement chimique très délicate et nécessitent la mise en oeuvre d’une méthodologie adaptée. Les travaux présentés dans ce manuscrit montrent que l’utilisation simultanée de très hauts champs magnétiques et de méthodes d’acquisitions appropriées peut permettre l’étude de l’environnement chimique du soufre dans les solides par RMN. Des calculs premier principe des paramètres RMN ont été menés et leur comparaison à l’expérience montre qu’il est possible de prédire avec fiabilité les paramètres RMN et d’attribuer les signaux observés à une structure chimique. Les positions et les largeurs des signaux RMN de soufre-33 correspondant à des ponts soufrés ont été calculées à partir de modèles structuraux. Pour de tels environnements, les couplages quadripolaires attendus sont particulièrement forts (CQ > 40 MHz), et donnent lieu à des signaux RMN extrêmement larges dont l’observabilité est évaluée via l’étude du soufre élémentaire. Dans le cas d’élastomères vulcanisés, les résultats de cette étude montrent que l’observation de l’ensemble des différents environnements chimiques du soufre nécessite à priori l’utilisation de très hauts champs magnétiques et de très basses températures. / Sulfur vulcanization is a widely used crosslinking process of elastomers in the rubber industry, but the involved chemical mechanisms and the structure of the crosslinked materials are still poorly understood. Nuclear Magnetic Resonance (NMR) spectroscopy, which allows selectively probing the sulfur chemical environments, can provide new information about the local structure of the crosslinked material. However, due to its intrinsic properties, the observation of the NMR active isotope of sulfur (33S) is challenging in solid materials and requires the use of a specific methodology. In this work, we show that the use of very high magnetic fields and convenient NMR methods allows studying the chemical environment of sulfur in solid materials. First principle computations of the NMR parameters have been performed and compared to experimental results. This comparison shows that the computations lead to a reliable prediction of the NMR parameters and can be used to assign the observed NMR signals to a chemical structure. The NMR parameters characteristic of sulfur atoms involved in crosslinks have been computed from structural models. For such sulfur local environments, extremely large quadrupolar coupling constants (CQ > 40 MHz) and thus ultra broad resonances are expected. The NMR detection limit of sulfur environments giving rise to such very broad lines has been investigated through the 33S NMR study of elemental sulfur. In the case of vulcanized rubbers, the results of this work suggest that the NMR observation of the distinct sulfur chemical environments present in the crosslinked networks requires the use of both ultra high magnetic field and very low temperature.

O trabalho constitui-se numa investigação teórica da estrutura ordenada e intuitiva dos espaços coerentes, introduzidos por Girard [GIR 86], na definição do modelo de máquina geométrica para construção e interpretação de estados e processos computacionais rotulados por posições de um espaço geométrico. Esta interpretação poderá ser aplicada às construções determinísticas, incluindo dois tipos especiais de paralelismo - o espacial, com memória e processos infinitos definidos por estruturas matriciais, que operam sobre dimensões independentes, de forma sincronizada; e o temporal, na versão genérica do modelo, com memória global transfinita e processos distribuídos num conjunto enumerável de máquinas geométricas, sincronizadas no tempo. O modelo contempla interpretação para computações não-determinísticas e prevê a aplicação de operadores exponenciais na interpretação do espaço funcional. A noção mais intuitiva deste trabalho está na definição da relação de coerência, que define o grafo sobre o qual se constrói este domínio semántico. Sobre o conjunto de pontos compatíveis de tais grafos, a coerência estrita interpreta a condição implícita para modelar o paralelismo - a concorrência entre posições de memória. Na construção dual, justificada pela presença da negação involutiva no grafo complementar, a incoerência interpreta a condição para o não-determinismo - o conflito de acesso à memória. Para os demais construtores, o produto sequencial e a soma determinística, consideram-se os endofunctores produto e soma direta da categoria CospLin dos espaços coerentes e funções lineares. A estrutura ordenada deste modelo é formalizada pelo espaço coerente D∞ de todos os processos, construído em níveis a partir do espaço coerente D∞ dos processos elementares, seguindo a metodologia proposta por Scott [SCO 76]. Neste sentido, cada nível da construção está identificado por um subespaço Dn que reconstrói todos os objetos do nível anterior, preservando suas propriedades e relações, além de construir os novos objetos. Compatível com a abordagem algébrica, o relacionamento entre os níveis é expresso por funções lineares denominadas imersões e projeções, interpretanto os construtores de processos e seus destrutores, respectivamente. Pelo procedimento de completação, assegura-se a existência do menor ponto fixo para equações recursivas definidas pela composição infinita destes morfismos. Além disso, as interpretações para processos infinitos, construídos por prefixação, apresentadas em D→∞ comprovam que este modelo é compatível com a diversidade dos construtores. O espa¸co coerente D∞2 dos processos transfinitos generaliza a construção e define a estrutura ordenada do modelo de máquina geométrica distribuída. Seus objetos são subconjuntos coerentes de tokens rotulados por posições do espaço geométrico e indexados por subconjuntos isomorfos aos ordinais transfinitos. O espaço coerente S S dos traços lineares de funções definidas sobre o espaço coerente S dos estados computacionais constitui-se no modelo semântico para análise do comportamento associado a cada processo interpretado em D∞. A definição da função de representação introduz um domínio de expressões que formaliza uma linguagem capaz de expressar, de forma mais operacional, as interpretações obtidas neste modelo de m´aquina. Cada uma das expressões válidas na linguagem é compatível com uma expressão gráfica. / This work presents a theoretical investigation of the constructive, intuitive and ordered structure of the coherence spaces, introduced by Girard, in order to define the geometric machine model for interpretation of computational states and processes labelled by positions of a geometric space. This interpretation can be applied to deterministic process constructions, including two special types of parallelism - the temporal parallelism, with infinite memory and infinite processes defined over array structures, that operate over independent dimensions in a synchronized way; and the spatial parallelism, in a generic version of the model, with a transfinite global memory shared by transfinite processes distributed in a enumerable set of geometric machines, synchronized in the time. The work also provides interpretation to the non-deterministic computations and applies the exponential operators in the interpretation of the functional space. The most basic notion of this work is the definition of the coherence relation as the admissibility of parallelism between basic operations (elementary processes). That relation defines the web over which the coherence space of the whole set of deterministic and non-deterministic processes is step-wise and systematically build. Over the set of the compatible points of such graph, the strict coherence interprets the implicity condition to model parallelism - the true concurrence. In the dual construction, justified by the presence of involutive negation in the complementary graph, the incoherence interprets the condition that models non-determinism - the conflict of memory accesses. The other constructors, the sequential product and the deterministic sum, are defined by the endofunctors in the CospLin category of the coherence spaces and linear functions. The ordered structure of this model is formalized by the coherence space D∞ of all processes, constructed by levels from the coherence space D0 of the elementary processes, following the Scott’s methodology [SCO 76]. In this sense, each level is identified by a subspace Dn, which reconstructs all the objects from the level before, preserving their properties and relations, and drives the construction of the new objects. Compatible with the algebraic-theoretic approach to computational processes, the relationship between the levels is expressed by linear functions called embedding and projection-functions, which interpret constructors and destructors of processes, respectively. The completion procedure guarantees the existence of the least fixed point to the recursive equations, defined by infinite composition of these morphisms. In addition, the interpretation for infinite processes constructed by prefix is presented in D→∞ , confirms that the ordered structure of these model is compatible with the diversity of constructors. The coherence space D∞2 of transfinite processes generalizes the construction and defines the ordered structure of the distributed geometric machine model. Its objects are coherent subsets of tokens labelled by the positions of a geometric space and indexed by isomorphic subsets related to the transfinite ordinal numbers. In order to analyze the behavior related to the interpretations in D∞, the coherence space S S of the linear traces of functions, defined over the coherence space S of the computational states, is introduced. The definition of the representation-function induces the construction of the domain Ω of valid expressions and formalizes a (graphic) language which is able to express, in an more operational way, the interpretations obtained in the geometric machine model.

O trabalho constitui-se numa investigação teórica da estrutura ordenada e intuitiva dos espaços coerentes, introduzidos por Girard [GIR 86], na definição do modelo de máquina geométrica para construção e interpretação de estados e processos computacionais rotulados por posições de um espaço geométrico. Esta interpretação poderá ser aplicada às construções determinísticas, incluindo dois tipos especiais de paralelismo - o espacial, com memória e processos infinitos definidos por estruturas matriciais, que operam sobre dimensões independentes, de forma sincronizada; e o temporal, na versão genérica do modelo, com memória global transfinita e processos distribuídos num conjunto enumerável de máquinas geométricas, sincronizadas no tempo. O modelo contempla interpretação para computações não-determinísticas e prevê a aplicação de operadores exponenciais na interpretação do espaço funcional. A noção mais intuitiva deste trabalho está na definição da relação de coerência, que define o grafo sobre o qual se constrói este domínio semántico. Sobre o conjunto de pontos compatíveis de tais grafos, a coerência estrita interpreta a condição implícita para modelar o paralelismo - a concorrência entre posições de memória. Na construção dual, justificada pela presença da negação involutiva no grafo complementar, a incoerência interpreta a condição para o não-determinismo - o conflito de acesso à memória. Para os demais construtores, o produto sequencial e a soma determinística, consideram-se os endofunctores produto e soma direta da categoria CospLin dos espaços coerentes e funções lineares. A estrutura ordenada deste modelo é formalizada pelo espaço coerente D∞ de todos os processos, construído em níveis a partir do espaço coerente D∞ dos processos elementares, seguindo a metodologia proposta por Scott [SCO 76]. Neste sentido, cada nível da construção está identificado por um subespaço Dn que reconstrói todos os objetos do nível anterior, preservando suas propriedades e relações, além de construir os novos objetos. Compatível com a abordagem algébrica, o relacionamento entre os níveis é expresso por funções lineares denominadas imersões e projeções, interpretanto os construtores de processos e seus destrutores, respectivamente. Pelo procedimento de completação, assegura-se a existência do menor ponto fixo para equações recursivas definidas pela composição infinita destes morfismos. Além disso, as interpretações para processos infinitos, construídos por prefixação, apresentadas em D→∞ comprovam que este modelo é compatível com a diversidade dos construtores. O espa¸co coerente D∞2 dos processos transfinitos generaliza a construção e define a estrutura ordenada do modelo de máquina geométrica distribuída. Seus objetos são subconjuntos coerentes de tokens rotulados por posições do espaço geométrico e indexados por subconjuntos isomorfos aos ordinais transfinitos. O espaço coerente S S dos traços lineares de funções definidas sobre o espaço coerente S dos estados computacionais constitui-se no modelo semântico para análise do comportamento associado a cada processo interpretado em D∞. A definição da função de representação introduz um domínio de expressões que formaliza uma linguagem capaz de expressar, de forma mais operacional, as interpretações obtidas neste modelo de m´aquina. Cada uma das expressões válidas na linguagem é compatível com uma expressão gráfica. / This work presents a theoretical investigation of the constructive, intuitive and ordered structure of the coherence spaces, introduced by Girard, in order to define the geometric machine model for interpretation of computational states and processes labelled by positions of a geometric space. This interpretation can be applied to deterministic process constructions, including two special types of parallelism - the temporal parallelism, with infinite memory and infinite processes defined over array structures, that operate over independent dimensions in a synchronized way; and the spatial parallelism, in a generic version of the model, with a transfinite global memory shared by transfinite processes distributed in a enumerable set of geometric machines, synchronized in the time. The work also provides interpretation to the non-deterministic computations and applies the exponential operators in the interpretation of the functional space. The most basic notion of this work is the definition of the coherence relation as the admissibility of parallelism between basic operations (elementary processes). That relation defines the web over which the coherence space of the whole set of deterministic and non-deterministic processes is step-wise and systematically build. Over the set of the compatible points of such graph, the strict coherence interprets the implicity condition to model parallelism - the true concurrence. In the dual construction, justified by the presence of involutive negation in the complementary graph, the incoherence interprets the condition that models non-determinism - the conflict of memory accesses. The other constructors, the sequential product and the deterministic sum, are defined by the endofunctors in the CospLin category of the coherence spaces and linear functions. The ordered structure of this model is formalized by the coherence space D∞ of all processes, constructed by levels from the coherence space D0 of the elementary processes, following the Scott’s methodology [SCO 76]. In this sense, each level is identified by a subspace Dn, which reconstructs all the objects from the level before, preserving their properties and relations, and drives the construction of the new objects. Compatible with the algebraic-theoretic approach to computational processes, the relationship between the levels is expressed by linear functions called embedding and projection-functions, which interpret constructors and destructors of processes, respectively. The completion procedure guarantees the existence of the least fixed point to the recursive equations, defined by infinite composition of these morphisms. In addition, the interpretation for infinite processes constructed by prefix is presented in D→∞ , confirms that the ordered structure of these model is compatible with the diversity of constructors. The coherence space D∞2 of transfinite processes generalizes the construction and defines the ordered structure of the distributed geometric machine model. Its objects are coherent subsets of tokens labelled by the positions of a geometric space and indexed by isomorphic subsets related to the transfinite ordinal numbers. In order to analyze the behavior related to the interpretations in D∞, the coherence space S S of the linear traces of functions, defined over the coherence space S of the computational states, is introduced. The definition of the representation-function induces the construction of the domain Ω of valid expressions and formalizes a (graphic) language which is able to express, in an more operational way, the interpretations obtained in the geometric machine model.