Dissertations / Theses on the topic 'Calcul parallèle'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Calcul parallèle.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Mezher, Dany. "Calcul parallèle de pseudo-spectres." Rennes 1, 2001. http://www.theses.fr/2001REN10054.
Full textBarboteu, Mikaël. "Contact, frottement et techniques de calcul parallèle." Montpellier 2, 1999. http://www.theses.fr/1999MON20047.
Full textPlazolles, Bastien. "Plateforme de calcul parallèle « Design for Demise »." Thesis, Toulouse 3, 2017. http://www.theses.fr/2017TOU30055.
Full textThe risk of space debris is now perceived as primordial by government and international space agencies. Since the last decade, international space agencies have developed tools to simulate the re-entry of satellites and orbital stations in order to assess casualty risk on the ground. Nevertheless , all current tools provide deterministic solutions, though models include various parameters that are not well known. Therefore, the provided results are strongly dependent on the as-sumptions made. One solution to obtain relevant and exploitable results is to include uncertainties around those parame-ters in order to perform Monte-Carlo analysis. But such a study is very time consuming due to the large parameter space to explore (that necessitate hundreds of thousands simulations). As part of this thesis work we propose a new satellite atmospheric reentry simulation to perform statistical analysis. To master computing time this tool takes advantage of Taguchi method to restrain the amount of parameter to study and also takes advantage of computing accelerators like Graphic Processing Units (GPUs) and Intel Xeon Phi
Lorong, Philippe. "Sur une approche parallèle pour le calcul des structures : comportement sur calculateurs parallèles." Cachan, Ecole normale supérieure, 1994. http://www.theses.fr/1994DENS0015.
Full textBergogne, Laurent. "Quelques algorithmes parallèles sur des "séquences de" pour différents modèles de calcul parallèle." Amiens, 1999. http://www.theses.fr/1999AMIE0130.
Full textKla, Koué Sylvanus. "Calcul parallèle et en-ligne des fonctions arithmétiques." Lyon 1, 1993. http://www.theses.fr/1993LYO10031.
Full textXiao, Manyu. "Optimisation multidisciplinaire avec modèles réduits et calcul parallèle." Compiègne, 2010. http://www.theses.fr/2010COMP1894.
Full textThe management of the computational effort (CPU time, memory, interfacing) is a major issue in Multidisciplinary Design Optimization (MDO), due to the cost of the “high fidelity” numerical simulations (finite elements, finite volumes, etc. ), as well as the coupling between disciplines (exchange of information, mesh compatibility, file formats). Therefore, to decrease the overall cost of an MDO process, reduced-order models combined with parallel computing strategies constitute an economical and efficient solution. This topic is thoroughly investigated in this thesis. First, a parallel computing environment gathering optimization algorithms, numerical models and process management is analyzed for the case of an airfoil subject to fluid-structure interaction, demonstrating the close connection between the selection of optimization algorithms and the high-performance computing (HPC) platforms, and also revealing the need for reduced-order models. Consequently, the remainder of the thesis is concerned with the development of novel approximation methods combining the advantages of both “general” and “physics-based” metamodels. More precisely, constrained versions of proper orthogonal decomposition techniques (called CPOD/CPOD2) have been carried out, and successfully applied to an academic test case (the aeroelastic airfoil) and to an industrial example (the multiobjective shape optimization of an intake port developed by Renault)
Fedak, Gilles. "XtremWeb : une plate-forme générique pour l'étude expérimentale du calcul global et pair-à-pair." Paris 11, 2003. http://www.theses.fr/2003PA112095.
Full textDefined as a merge between Global Computing systems (SETI@Home, distributed. Net) and Peer-to-Peer applications (Napster, Gnuttella), Peer-to-Peer Global Computing, gathers idle computing resources, spread over Internet or LAN to execute massively parallel applications. In this model, every resource is able to compute or to request a computation from the other participants. XtremWeb is a generic, secured, fault-tolerant environment to execute parallel applications on volatile nodes. The project aims at providing a high-performance computing environment for production use and providing a soft-ware platform for research and experimentation. The thesis presents a state of the art of large scale distributed systems, the fundamental principles, architecture, realization and performances evaluation of XtremWeb. Architecture is based on three components : Worker which controls the computing node, Client which submits applications and Dispatcher which distributes executions on computing no des and insures fault tolerance of workers and clients. With this platform we made two specific studies relative to security of computing nodes and execution of parallel applications (MPICH-V). MPICH-V is an implementation of the MPICH library tolerant to the nodes volatility based on uncoordinated checkpoint and pessimistic log of the messages. XtremWeb is used by several scientific collaborators through ACI, RNTL projects and by industrial partners in a production environment. A Desk-top Grid based on XtremWeb is currently deployed within the University Paris-Sud Campus
Karaseva, Olga. "Déformations élastiques des presses de forgeage et calcul parallèle." Phd thesis, École Nationale Supérieure des Mines de Paris, 2005. http://pastel.archives-ouvertes.fr/pastel-00001513.
Full textGoldman, Alfredo. "Impact des modèles d'exécution pour l'ordonnancement en calcul parallèle." Phd thesis, Grenoble INPG, 1999. http://tel.archives-ouvertes.fr/tel-00004839.
Full textKaci, Ania. "Conception d'une architecture extensible pour le calcul massivement parallèle." Thesis, Paris Est, 2016. http://www.theses.fr/2016PESC1044.
Full textIn response to the growing demand for performance by a wide variety of applications (eg, financial modeling, sub-atomic simulation, bioinformatics, etc.), computer systems become more complex and increase in size (number of computing components, memory and storage capacity). The increased complexity of these systems results in a change in their architecture towards a heterogeneous computing technologies and programming models. The harmonious management of this heterogeneity, resource optimization and minimization of consumption are major technical challenges in the design of future computer systems.This thesis addresses a field of this complexity by focusing on shared memory subsystems where all processors share a common address space. Work will focus on the implementation of a cache coherence and memory consistency on an extensible architecture and methodology for validation of this implementation.In our approach, we selected processors 64-bit ARM and generic co-processor (GPU, DSP, etc.) as components of computing, shared memory protocols AMBA / ACE and AMBA / ACE-Lite and associated architecture "CoreLink CCN" as a starting solution. Generalization and parameterization of this architecture and its validation in the simulation environment GEM5 are the backbone of this thesis.The results at the end of the thesis, tend to demonstrate the achievement of objectives
Cellier, Fabien. "Modélisation et calcul parallèle pour le Web SIG 3D." Thesis, Lyon 1, 2014. http://www.theses.fr/2014LYO10015/document.
Full textThis thesis focuses on displaying and manipulating 3D models from Geographic Information Systems (GIS) in interactive time directly in a web browser. Its main contributions are the visualization of high resolution 3D terrains, the simplification of irregular meshes on the GPU, and the creation of a new API for performing heavy and effective computing in the browser (parallelism GP/GPU) without compromising safety. The first approach proposed for the visualization of terrain models is built on recent browsers efforts to become a versatile platform. With the new 3D pluginless APIs, we have created a visualization client for terrain models “streamed” through HTTP. It fits perfectly into the current Web-GIS ecosystem (desktop and mobile) by the use of the standard protocols provided by OGC Open Geospatial Consortium. This prototype is part of an industrial partnership between ATOS Wordline and its GIS customer, and particularly the IGN (French National Geographic Institute) with the Geoportail application (http://www.geoportail.gouv.fr) and its mapping APIs. The 3D embedded in browsers brings its own challenges which are different from what we know in heavy applications: restrictions and constraints from JavaScript but also problems of data transfer. These constraints, detailed in the next paragraph, led us to rethink the standard algorithms for 3D visualization to take into account the browser specificities. Thus, we have taken advantage of network latency to dynamically manage the connections between the different parts of the mesh without significantly impacting the rendering speed. Beyond 3D visualization, and even if the JavaScript language allows task parallelism, data parallelism remains absent from Web browsers. This observation, added to the slowness of JavaScript processing, constituted a major obstacle in our goal to define a complete and powerful GIS platform integrated in the browser. That is why we have designed and developed the WebCLWorkers, a GP/GPU Web API for high performance computing that meets the criteria of simplicity and security inherent to the Web. We tried to find a trade-off for a language close to the script but secure and efficient, based on the OpenCL API at runtime. This approach is opposite to the existing ones, which are either based on precompiled code or disregard performances. Our API proposal interested the Mozilla Foundation which asked us to participate in the development of the WebCL standard by integrating the Khronos Group (Mozilla, Samsung, Nokia, Google, AMD, and so on). Exploiting these new computing resources, we then suggested an algorithm for parallel simplification of irregular meshes. While the state of the art was mainly based on regular grids for parallelism (and did not take into account Web browsers restrictions) or on simplification and kd-tree clustering, no solution could allow both parallel simplification and progressive visualization using irregular grids. Our solution is based on a three-step algorithm using implicit priorities and local minima to achieve simplification, and its degree of parallelism is linearly related to the number of points and triangles in the mesh to process. We have proposed in the thesis an innovative approach for 3D WebGIS pluglinless visualization, offering tools that bring to the browser a comfortable GP/GPU computing power, and designing a method for irregular meshes parallel simplification allowing to visualize level of details directly in Web browsers. Based on these initial results, it becomes possible to carry all the rich functionalities of desktop GIS clients to Web browsers, on PC as well as mobile phones and tablets
Baladron, Pezoa Javier. "Exploring the neural codes using parallel hardware." Phd thesis, Université Nice Sophia Antipolis, 2013. http://tel.archives-ouvertes.fr/tel-00847333.
Full textJurczuk, Krzysztof. "Calcul parallèle pour la modélisation d'images de résonance magnétique nucléaire." Thesis, Rennes 1, 2013. http://www.theses.fr/2013REN1S089.
Full textThis PhD thesis concerns computer modeling of magnetic resonance imaging (MRI). The main attention is centered on imaging of vascular structures. Such imaging is influenced not only by vascular geometries but also by blood flow which has to been taken into account in modeling. Next to the question about the quality of developed models, the challenge lies also in the demand for high performance computing. Thus, in order to manage computationally complex problems, parallel computing is in use. In the thesis three solutions are proposed. The first one concerns parallel algorithms of vascular network modeling. Algorithms for different architectures are proposed. The first algorithm is based on the message passing model and thus, it is suited for distributed memory architectures. It parallelizes the process of connecting new parts of tissue to existing vascular structures. The second algorithm is designed for shared memory machines. It also parallelizes the perfusion process, but individual processors perform calculations concerning different vascular trees. The third algorithm combines message passing and shared memory approaches providing solutions for hybrid parallel architectures. Developed algorithms are able to substantially speed up the time-demanded simulations of growth of complex vascular networks. As a result, more elaborate and precise vascular structures can be simulated in a reasonable period of time. It can also help to extend the vascular model and to test multiple sets of parameters. Secondly, a new approach in computational modeling of magnetic resonance (MR) flow imaging is proposed. The approach combines the flow computation by lattice Boltzmann method, MRI simulation by following discrete local magnetizations in time and a new magnetization transport algorithm together. Results demonstrate that such an approach is able to naturally incorporate the flow influence in MRI modeling. As a result, in the proposed model, no additional mechanism (unlike in prior works) is needed to consider flow artifacts, what implies its easy extensibility. In combination with its low computational complexity and efficient implementation, the solution is a user-friendly and manageable at different levels tool which facilitates running series of simulations with different physiological and imaging parameters. The goal of the third solution is to apply the proposed MR flow imaging model on complex vascular networks. To this aim, models of vascular networks, flow behavior and MRI are combined together. In all the model components, computations are adapted to be performed at various parallel architectures. The model potential and possibilities of simulations of flow and MRI in complex vascular structures are shown. The model aims at explaining and exploring MR image formation and appearance by the combined knowledge from many processes and systems, starting from vascular geometry, through flow patterns and ending on imaging technology
Jankee, Christopher. "Optimisation par métaheuristique adaptative distribuée en environnement de calcul parallèle." Thesis, Littoral, 2018. http://www.theses.fr/2018DUNK0480/document.
Full textTo solve discrete optimization problems of black box type, many stochastic algorithms such as evolutionary algorithms or metaheuristics exist and prove to be particularly effective according to the problem to be solved. Depending on the observed properties of the problem, choosing the most relevant algorithm is a difficult problem. In the original framework of parallel and distributed computing environments, we propose and analyze different adaptive optimization algorithm selection strategies. These selection strategies are based on reinforcement learning methods automatic, from the field of artificial intelligence, and on information sharing between computing nodes. We compare and analyze selection strategies in different situations. Two types of synchronous distributed computing environment are discussed : the island model and the master-slave model. On the set of nodes synchronously at each iteration, the adaptive selection strategy chooses an algorithm according to the state of the search for the solution. In the first part, two problems OneMax and NK, one unimodal and the other multimodal, are used as benchmarks for this work. Then, to better understand and improve the design of adaptive selection strategies, we propose a modeling of the optimization problem and its local search operator. In this modeling, an important characteristic is the average gain of an operator according to the fitness of the candidate solution. The model is used in the synchronous framework of the master-slave model. A selection strategy is broken down into three main components : the aggregation of the rewards exchanged, the learning scheme and the distribution of the algorithms on the computing nodes. In the final part, we study three scenarios, and we give keys to understanding the relevant use of adaptive selection strategies over naïve strategies. In the framework of the master-slave model, we study the different ways of aggregating the rewards on the master node, the distribution of the optimization algorithms of the nodes of computation and the time of communication. This thesis ends with perspectives in the field of distributed adaptive stochastic optimization
Bordeu, Weldt Felipe Eduardo. "Calcul à l'echelle méso avec interface non locale des composites stratifiés." Phd thesis, École normale supérieure de Cachan - ENS Cachan, 2012. http://tel.archives-ouvertes.fr/tel-00975428.
Full textGouin, Florian. "Méthodologie de placement d'algorithmes de traitement d'images sur architecture massivement parallèle." Thesis, Paris Sciences et Lettres (ComUE), 2019. http://www.theses.fr/2019PSLEM075.
Full textIn industries, the curse of image sensors for higher definitions increases the amount of data to be processed in the image processing domain. The concerned algorithms, applied to embedded solutions, also have to frequently accept real-time constraints. So, the main issues are to moderate power consumption, to attain high performance computings and high memory bandwidth for data delivery.The massively parallel conception of GPUs is especially well adapted for this kind of tasks. However, this achitecture is complex to handle. Some reasons are its multiple memory and computation hierachical levels or the usage of this accelerator inside a global heterogeneous architecture. Therefore, mapping algorithms on GPUs, while exploiting high performance capacities of this architecture, aren’t trivial operations.In this thesis, we have developped a mapping methodology for sequential algorithms and designed it for GPUs. This methodology is made up of code analysis phases, mapping criteria verifications, code transformations and a final code generation phase. Part of the defined mapping criteria has been designed to assure the mapping legality, by considering GPU hardware specifities, whereas the other part are used to improve runtimes. In addition, we have studied GPU memories performances and the capacity of GPU to efficiently support coarse grain parallellism. This complementary work is a foundation for further improvments of GPU resources exploitation inside this mapping methodology.Last, the experimental results have revealed the functional reliability of the codes mapped on GPU and a speedup on the runtime of many C and C++ image processing applications used in industry
Ben, Romdhanne Bilel. "Simulation des réseaux à grande échelle sur les architectures de calculs hétérogènes." Thesis, Paris, ENST, 2013. http://www.theses.fr/2013ENST0088/document.
Full textThe simulation is a primary step on the evaluation process of modern networked systems. The scalability and efficiency of such a tool in view of increasing complexity of the emerging networks is a key to derive valuable results. The discrete event simulation is recognized as the most scalable model that copes with both parallel and distributed architecture. Nevertheless, the recent hardware provides new heterogeneous computing resources that can be exploited in parallel.The main scope of this thesis is to provide a new mechanisms and optimizations that enable efficient and scalable parallel simulation using heterogeneous computing node architecture including multicore CPU and GPU. To address the efficiency, we propose to describe the events that only differs in their data as a single entry to reduce the event management cost. At the run time, the proposed hybrid scheduler will dispatch and inject the events on the most appropriate computing target based on the event descriptor and the current load obtained through a feedback mechanisms such that the hardware usage rate is maximized. Results have shown a significant gain of 100 times compared to traditional CPU based approaches. In order to increase the scalability of the system, we propose a new simulation model, denoted as general purpose coordinator-master-worker, to address jointly the challenge of distributed and parallel simulation at different levels. The performance of a distributed simulation that relies on the GP-CMW architecture tends toward the maximal theoretical efficiency in a homogeneous deployment. The scalability of such a simulation model is validated on the largest European GPU-based supercomputer
Marchand, Philippe. "Système de lecture parallèle de disques optiques : applications au calcul optoélectronique." Mulhouse, 1991. http://www.theses.fr/1991MULH0170.
Full textBrieu, Mathias. "Homogénéisation et endommagement de composites elastomères par techniques de calcul parallèle." Cachan, Ecole normale supérieure, 1999. http://www.theses.fr/1999DENS0005.
Full textGbikpi, benissan Tete guillaume. "Méthodes asynchrones de décomposition de domaine pour le calcul massivement parallèle." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLC071/document.
Full textAn important class of numerical methods features a scalability property well known as the Amdahl’s law, which constitutes the main limiting drawback of parallel computing, as it establishes an upper bound on the number of parallel processing units that can be used to speed a computation up. Extensive research activities are therefore conducted on both mathematical and computer science aspects to increase this bound, in order to be able to squeeze the most out of parallel machines. Domain decomposition methods introduce a natural and optimal approach to solve large numerical problems in a distributed way. They consist in dividing the geometrical domain on which an equation is defined, then iteratively processing each sub-domain separately, while ensuring the continuity of the solution and of its derivative across the junction interface between them. In the present work, we investigate the removal of the scalability bound by the application of the asynchronous iterations theory in various decomposition frameworks, both for space and time domains. We cover various aspects of the development of asynchronous iterative algorithms, from theoretical convergence analysis to effective parallel implementation. Efficient asynchronous domain decomposition methods are thus successfully designed, as well as a new communication library for the quick asynchronous experimentation of existing scientific applications
Abbas-Turki, Lokman. "Calcul parallèle pour les problèmes linéaires, non-linéaires et linéaires inverses en finance." Thesis, Paris Est, 2012. http://www.theses.fr/2012PEST1055/document.
Full textHandling multidimensional parabolic linear, nonlinear and linear inverse problems is the main objective of this work. It is the multidimensional word that makes virtually inevitable the use of simulation methods based on Monte Carlo. This word also makes necessary the use of parallel architectures. Indeed, the problems dealing with a large number of assets are major resources consumers, and only parallelization is able to reduce their execution times. Consequently, the first goal of our work is to propose "appropriate" random number generators to parallel and massively parallel architecture implemented on CPUs/GPUs cluster. We quantify the speedup and the energy consumption of the parallel execution of a European pricing. The second objective is to reformulate the nonlinear problem of pricing American options in order to get the same parallelization gains as those obtained for linear problems. In addition to its parallelization suitability, the proposed method based on Malliavin calculus has other practical advantages. Continuing with parallel algorithms, the last point of this work is dedicated to the uniqueness of the solution of some linear inverse problems in finance. This theoretical study enables the use of simple methods based on Monte Carlo
Léger, Laurent. "Calcul parallèle des racines d'un polynôme à l'aide de la méthode de Weyl." Rouen, 1998. http://www.theses.fr/1998ROUES032.
Full textVialla, Bastien. "Contributions à l'algèbre linéaire exacte sur corps finis et au chiffrement homomorphe." Thesis, Montpellier, 2015. http://www.theses.fr/2015MONTS112.
Full textThis thesis is composed of two independent parts.The first one is related to homomorphic encryption and the second part deal with sparse linear algebra on finite fields.Homomorphic encryption extends traditional encryption in the sense that it becomes feasible to perform operations on ciphertexts, without the knowledge of the secret decryption key. As such, it enables someone to delegate heavy computations on his sensitive data to an untrusted third party, in a secure way. More precisely, with such a system, one user can encrypt his sensitive data such that the third party can evaluate a function on the encrypted data, without learning any information on the underlying plain data. Getting back the encrypted result, the user can use his secret key to decrypt it and obtain, in clear, the result of the evaluation of the function on his sensitive plain data. For a cloud user, the applications are numerous, and reconcile both a rich user experience and a strong privacy protection.The first fully homomorphic encryption (FHE) scheme, able to handle an arbitrary number of additions and multiplications on ciphertexts, has been proposed by Gentry in 2009.In homomorphic encryption schemes, the executed function is typically represented as an arithmetic circuit. In practice, any circuit can be described as a set of successive operation gates, each one being either a sum or a product performed over some ring.In Gentry’s construction, based on lattices, each ciphertext is associated with some noise, which grows at each operation (addition or multiplication) done throughout the evaluation of the function. When this noise reaches a certain limit, decryption is not possible anymore.To overcome this limitation, closely related to the number of operations that the HE.Eval procedure can handle, Gentry proposed in a technique of noise refreshment called“bootstrapping”.The main idea behind this bootstrapping procedure is to homomorphically run the decryptionprocedure of the scheme on the ciphertext, using an encrypted version of the secret key. In this context, our contribution is twofold. We first prove that the lmax-minimizing bootstrapping problem is APX-complete and NP-complete for lmax ≥ 3. We then propose a new method to determine the minimal number of bootstrappings needed for a given FHE scheme and a given circuit.We use linear programming to find the best outcome for our problem. The main advantage of our method over the previous one is that it is highly flexible and can be adapted for numerous types of homomorphic encryption schemes and circuits.Computing a kernel element of a matrix is a fundamental kernel in many computer algebra and cryptography algorithms. Especially, many applications produces matrices with many matrix elements equals to 0.Those matrices are named sparse matrices. Sparse linear algebra is fundamentally relying on iterative approaches such as Wiedemann or Lanczos. The main idea is to replace the direct manipulation of a sparse matrix with its Krylov subspace. In such approach, the cost is therefore dominated by the computation of the Krylov subspace, which is done by successive product of a matrix by a vector or a dense matrix.Modern processor unit characteristics (SIMD, multicores, caches hierarchy, ...) greatly influence algorithm design.In this context our work deal with the best approach to design efficient implementation of sparse matrix vector product for modern processors.We propose a new sparse matrix format dealing with the many +-1 matrix elements to improve performance.We propose a parallel implementation based on the work stealing paradigm that provide a good scaling on multicores architectures.We study the impact of SIMD instructions on sparse matrix operations.Finally, we provide a modular arithmetic implementation based on residue number system to deal with sparse matrix vector product over multiprecision finite fields
Beniamine, David. "Analyzing the memory behavior of parallel scientific applications." Thesis, Université Grenoble Alpes (ComUE), 2016. http://www.theses.fr/2016GREAM088/document.
Full textSince a few decades, to reduce energy consumption, processor vendors builds more and more parallel computers.At the same time, the gap between processors and memory frequency increased significantly.To mitigate this gap, processors embed a complex hierarchical caches architectureWriting efficient code for such computers is a complex task.Therefore, performance analysis has became an important step of the development of applications seeking for performances.Most existing performance analysis tools focuses on the point of view of the processor.Theses tools see the main memory as a monolithic entity and thus are not able to understand how it is accessed.However, memory is a common bottleneck in High Performances Computing, and the pattern of memory accesses can impact significantly the performances.There are a few tools to analyze memory performances, however theses tools are based on a coarse grain sampling.Consequently, they focus on a small part of the execution missing the global memory behavior.Furthermore, these coarse grain sampling are not able to collect memory accesses patterns.In this thesis we propose two different tools to analyze the memory behavior of an application.The first tool is designed specifically for Not Uniform Memory Accesses machines and provides some visualizations of the global sharing pattern inside each data structure between the threads.The second one collects fine grain memory traces with temporal information.We can visualize theses traces either with a generic trace management framework or with a programmatic exploration using R.Furthermore we evaluate both of these tools, comparing them with state of the art memory analysis tools in terms of performances, precision and completeness
Trystram, Denis. "Quelques résultats de complexité en algorithmique parallèle et systolique." Grenoble INPG, 1988. http://tel.archives-ouvertes.fr/tel-00009202.
Full textRey, Christian. "Développement d'algorithmes parallèles de résolution en calcul non-linéaire de structures hétérogènes : application au cas d'une butée acier élastomère." Cachan, Ecole normale supérieure, 1994. http://www.theses.fr/1994DENS0025.
Full textSultan, Ziad. "Algèbre linéaire exacte, parallèle, adaptative et générique." Thesis, Université Grenoble Alpes (ComUE), 2016. http://www.theses.fr/2016GREAM030/document.
Full textTriangular matrix decompositions are fundamental building blocks in computational linear algebra. They are used to solve linear systems, compute the rank, the determinant, the null-space or the row and column rank profiles of a matrix. The project of my PhD thesis is to develop high performance shared memory parallel implementations of exact Gaussian elimination.In order to abstract the computational code from the parallel programming environment, we developed a domain specific language, PALADIn: Parallel Algebraic Linear Algebra Dedicated Interface, that is based on C/C + + macros. This domain specific language allows the user to write C + + code and benefit from sequential and parallel executions on shared memory architectures using the standard OpenMP, TBB and Kaapi parallel runtime systems and thus providing data and task parallelism.Several aspects of parallel exact linear algebra were studied. We incrementally build efficient parallel kernels, for matrix multiplication, triangular system solving, on top of which several variants of PLUQ decomposition algorithm are built. We study the parallelization of these kernels using several algorithmic variants: either iterative or recursive and using different splitting strategies.We propose a recursive Gaussian elimination that can compute simultaneously therow and column rank profiles of a matrix as well as those of all of its leading submatrices, in the same time as state of the art Gaussian elimination algorithms. We also study the conditions making a Gaussian elimination algorithm reveal this information by defining a new matrix invariant, the rank profile matrix
Labbé, Valérie. "Modélisation numérique du chauffage par induction : approche éléments finis et calcul parallèle." Phd thesis, École Nationale Supérieure des Mines de Paris, 2002. http://tel.archives-ouvertes.fr/tel-00443740.
Full text- électromagnétique, - thermique, - éventuellement thermo-mécanique.
Le choix du modèle électromagnétique est primordial. De nombreuses approximations basées sur des hypothèses plus ou moins fortes existent.
Nous avons seulement utilisé l'approximation des régimes quasi-permanents. Nous avons vu que cette première approximation, qui revient à négliger le phénomène de propagation des ondes, est valable dans la gamme de fréquences utilisée lors des procédés de chauffage par induction, les plus hautes fréquences étant largement inférieures au mégahertz. La propagation des ondes est alors considérée comme instantanée, ce qui au vu de la taille caractéristique des installations (quelques mètres) par rapport à la célérité de la lumière (3.105 m/s) est tout à fait raisonnable.
En revanche, nous avons choisi d'écarter l'approximation harmonique des champs électromagnétiques. Cette approximation découple les évolutions spatiales et temporelles du champ et revient à calculer une amplitude complexe pour le champ électromagnétique à partir d'une équation stationnaire. L'avantage d'une telle approximation est le gain souvent important en temps de calcul. Seulement, on perd une précision importante sur l'évolution temporelle et sur la déformation des champs électromagnétiques lorsqu'il s'agit d'un matériau ferromagnétique. En effet, les harmoniques secondaires ne sont pas prises en compte. Afin de pouvoir représenter les phénomènes physiques le plus réellement possible, le modèle électromagnétique utilisé est dépendant du temps. Néanmoins, afin de n'être pas trop pénalisant en temps de calcul, des compromis entre la précision des calculs et le temps de calcul nécessaire ont été étudiés. Ils se situent au niveau :
- du nombre de calculs électromagnétiques nécessaires pour bien décrire l'évolution temporelle d'une période électromagnétique, du nombre de périodes électromagnétiques nécessaires pour arriver à une solution stable,
du nombre de calculs électromagnétiques complets nécessaires au cours de l'évolution du champ de température.
Ces points importants, ainsi que des échelles de temps caractéristiques électromagnétiques et thermiques présentant un rapport allant de 10-2 à 10-6 ont nécessité la mise en place d'un couplage faible, basé sur la stabilisation du terme de puissance Joule moyennée sur une période électromagnétique ainsi que sur la stabilisation des paramètres électromagnétiques au cours de la montée en température.
La méthode numérique employée, de type éléments finis, est fiable et robuste. Néanmoins, elle nécessite une bonne compréhension des phénomènes physiques électromagnétiques inhérents au procédé. En effet, modéliser un espace ouvert par une méthode éléments finis nécessite la fermeture du domaine et l'imposition de conditions aux limites artificielles. L'utilisateur doit estimer la taille du domaine étudié qui doit être assez grand pour ne pas venir tronquer les lignes du champ électromagnétique et ainsi les modifier. Son avantage par rapport à une méthode mixte est que la matrice du système est creuse et symétrique. La résolution du problème est facilitée et se prête mieux à des développements en calcul parallèle.
Enfin, une nouvelle stratégie a été développée pour simuler le déplacement de l'inducteur : ses propriétés se déplacent virtuellement dans l'air. Cette méthode a donné de très bons résultats et ne nécessite aucun remaillage.
Les perspectives de recherche sont multiples.
Au niveau des données, le modèle accepte actuellement une tension ou une densité de courant source uniforme dans l'inducteur. Suite à un calcul électromagnétique complet, la répartition de courants est connue dans l'inducteur et permet une évaluation de l'intensité réelle circulant dans les spires. Il serait intéressant de mettre au point un outil de transfert des données électrotechniques vers nos paramètres d'entrées.
Un autre point, plus académique, serait d'effectuer des comparaisons pour des matériaux ferromagnétiques entre un modèle harmonique et le nôtre, dépendant en temps. En effet nous avons vu que ces deux modèles donnent des solutions identiques pour des matériaux amagnétiques. Tout l'intérêt de notre modèle dépendant en temps apparaît par son analyse beaucoup plus riche des matériaux non linéaires. Nous avons vu que le signal périodique peut être grandement déformé et ne ressemble alors plus du tout à une sinusoïde. Néanmoins, il n'est pas forcément évident que la puissance Joule, issue du calcul électromagnétique et obtenue par intégration sur une période électromagnétique, soit très différente de celle obtenue par une analyse harmonique. Cette différence serait très intéressante à quantifier.
Enfin des comparaisons entre les méthodes numériques 'tout' éléments finis et mixtes permettraient de quantifier la précision des méthodes suivant les tailles des éléments finis, les tailles du domaine de fermeture, ainsi que les différences en temps de calculs.
Un autre axe de ce travail a consisté à étudier et à implémenter une stratégie de parallélisation du modèle direct et de la procédure d'optimisation. Nous avons commencé par tester des solveurs itératifs préconditionnés sur nos différents modèles de type parabolique. Ceux ci donnant des résultats satisfaisants par rapport notamment à un solveur direct, nous avons pu nous orienter vers une méthode de parallélisation SPMD de type partitionnement de domaine. Cette méthode, simple et efficace, donne de très bons résultats au niveau du modèle direct, avec une bonne efficacité et une bonne scalabilité.
La parallélisation de l'optimisation montre une efficacité convenable sur deux et quatre processeurs mais qui tend à chuter rapidement avec le nombre de processeurs: la scalabilité est relativement moyenne. Ce problème fait apparaître une thématique de recherche intéressante en calcul parallèle appliqué aux méthodes adjointes: améliorer la scalabilité de l'optimisation parallèle en développant une meilleure stratégie d'accès aux données, en rééquilibrant les données stockées et les données à recalculer.
Enfin les perspectives à plus long terme consisteraient à développer un modèle analogue tridimensionnel.
Doreille, Mathias. "Athapascan-1 : vers un modèle de programmation parallèle adapté au calcul scientifique." Phd thesis, Grenoble INPG, 1999. http://tel.archives-ouvertes.fr/tel-00004825.
Full textKuhn, Matthieu. "Calcul parallèle et méthodes numériques pour la simulation de plasmas de bords." Thesis, Strasbourg, 2014. http://www.theses.fr/2014STRAD023/document.
Full textThe main goal of this work is to significantly reduce the computational cost of the scientific application Emedge3D, simulating the edge of tokamaks. Improvements to this code are made on two axes. First, innovations on numerical methods have been implemented. The advantage of semi-implicit time schemes are described. Their inconditional stability allows to consider larger timestep values, and hence to lower the number of temporal iteration required for a simulation. The benefits of a high order (time and space) are also presented. Second, solutions to the parallelization of the code are proposed. This study addresses the more general non linear advection-diffusion problem. The hot spots of the application have been sequentially optimized and parallelized with OpenMP. Then, a hybrid MPI OpenMP parallel algorithm for the memory bound part of the code is described and analyzed. Good scalings are observed up to 384 cores. This Ph. D. thesis is part of the interdisciplinary project ANR E2T2 (CEA/IRFM, University of Aix-Marseille/PIIM, University of Strasbourg/ICube)
Inal, A. Kaan. "Modélisation numérique de l'aluminium à grandes déformations plastiques, applications au calcul parallèle." Thesis, National Library of Canada = Bibliothèque nationale du Canada, 1997. http://www.collectionscanada.ca/obj/s4/f2/dsk1/tape11/PQDD_0001/MQ40592.pdf.
Full textInal, A. Kaan. "Modélisation numérique de l'aluminium à grandes déformations plastiques : applications au calcul parallèle." Mémoire, Université de Sherbrooke, 1998. http://savoirs.usherbrooke.ca/handle/11143/1046.
Full textAst, Isabelle d'. "Calcul parallèle en mécanique des fluides et problèmes spécifiques au couplage magnétohydrodynamique." Toulouse, INPT, 1995. http://www.theses.fr/1995INPT041H.
Full textFeyel, Frédéric. "Application du calcul parallèle aux modèles à grand nombre de variables internes." ENSMP, 1998. http://www.theses.fr/1998ENMP0864.
Full textRomagnoli, Emmanuel. "Exploitation efficace de grappes dynamiques de PC indifférenciées pour le calcul parallèle." Grenoble INPG, 2003. http://www.theses.fr/2003INPG0136.
Full textBazile, Alban. "Formulation éléments finis variationnelle adaptative et calcul massivement parallèle pour l’aérothermique industrielle." Thesis, Paris Sciences et Lettres (ComUE), 2019. http://www.theses.fr/2019PSLEM008/document.
Full textBy 2030, considering the progress of HPC, aerospace manufacturers like Safran Aircraft Engines (SAE), hope to be able to simulate a whole aircraft engine, at full scale, using Computational Fluid Dynamic (CFD). The goal of this PhD thesis is to bring a scientific contribution to this research framework. Indeed, the present work is devoted to the development of a variational adaptive finite element method allowing to improve the aerothermal simulations related to the turbine blade cooling. More precisely, our goal is to develop a new multiscale mesh adaptation technique, well suited to the resolution of highly convective heat transfers in turbulent flows. To do so, we propose a hierarchical control of errors based on recently developed subscales VMS error estimators. The first contribution of this work is then to propose a new isotropic mesh adaptation technique based on the previous error estimates. The second contribution is to combine both (i) the coarse scales interpolation error indicator and (ii) the subscales error estimator for anisotropic mesh adaptation. The results on analytic 2D and 3D benchmarks show that the proposed multiscale mesh adaptation technique allows obtaining highly precise solutions with much less elements in comparison with other mesh adaptation techniques. Finally, we propose in this thesis a description of the parallel software capabilities of Cimlib-CFD. Then, we present the two hardware systems used during this PhD thesis. The first one is the lab's cluster allowing the development of numerical methods. The second one however, is the GENCI Occigen II supercomputer which allows producing numerical results using massively parallel computations. In particular, we present a more realistic industrial concerning the cooling of a complete turbine vane composed by 39 holes
Fouladgar, Javad. "Contribution à l'extension du calcul parallèle : Application à l'identification et à la commande des systèmes interconnectés." Nancy 1, 1988. http://www.theses.fr/1988NAN10133.
Full textAit, Mohamed Otmane. "La théorie du π-calcul dans le système HOL." Nancy 1, 1996. http://www.theses.fr/1996NAN10091.
Full textSzczepanski, Nicolas. "SAT en Parallèle." Thesis, Artois, 2017. http://www.theses.fr/2017ARTO0403/document.
Full textThis thesis deals with propositional satisfiability (SAT) in a massively parallel setting. The SAT problem is widely used for solving several combinatorial problems (e.g. formal verification of hardware and software, bioinformatics, cryptography, planning, scheduling, etc.). The first contribution of this thesis concerns the design of efficient algorithms based on the approaches « portfolio » and « divide and conquer ». Secondly, an adaptation of several parallel programming models including hybrid (parallel and distributed computing) to SAT is proposed. This work has led to several contributions to international conferences and highly competitive distributed SAT solvers
Nguyen, Van-Hoa. "Traitement parallèle des comparaisons intensives de séquences génomiques." Rennes 1, 2009. https://tel.archives-ouvertes.fr/tel-00435792.
Full textThe sequence comparison process is one of the main bioinformatics task. The new sequencing technologies lead to a fast increasing of genomic data and strengthen the need of fast and efficient tools to perform this task. In this thesis, a new algorithm for intensive sequence comparison is proposed. It has been specifically designed to exploit all forms of parallelism of today microprocessors (SIMD instructions, multi-core architecture). This algorithm is also well suited for hardware accelerators such as FPGA or GPU boards. The algorithm has been implemented into the PLAST software (Parallel Local Alignment Search Tool). Different versions are available according to the data to process (protein and/or DNA). A MPI version has also been developed. According to the nature of the data and the type of technologies, speed-up from 3 to 20 has been measured compared with the reference software, BLAST, with the same level of quality
Essaïdi, Mohamed. "Echange de données pour le parallélisme à gros grain." Nancy 1, 2004. http://www.theses.fr/2004NAN10007.
Full textWe can distinguish two classes of parallel models: models with fine grain and those with coarse grain. The coarse grain models have a multitude of dedicated libraries facilitating the implementation of their respective algorithms. This thesis focuses on the library SSCRAP which is initially designed for the CGM model. We extended the support of SSCRAP to the model BSP and PRO. Indeed, we proposed and integrated on SSCRAP a new exchange model, a new synchronization model and a new execution model. In order to improve the efficiency of SSCRAP on shared memory architecture, we also designed and implemented a new interface of low level communication. The experimental studies highlight the very good behaviour of SSCRAP and then extend the proof of validity as well to the algorithms as to the coarse grain models
Thomin, Philippe. "Algorithmes parallèles pour la synthèse d'image par radiosité sur calculateur à mémoire distribuée." Valenciennes, 1993. https://ged.uphf.fr/nuxeo/site/esupversions/a9efdd76-820d-4008-ab0e-23f35d428cdf.
Full textDenis, Christophe. "Equilibrage en volume de calcul de la méthode parallèle à fronts multiples : Application au logiciel SIC." Compiègne, 2003. http://www.theses.fr/2003COMP1489.
Full textThe parallel multiple front method is used to solve large sparse linear systems issued from the finite element modeling in mechanical engineering. This direct method is based on a nonoverlapping domain decomposition method. Each subdomain is treated in parallel with a frontal method and then a interface problem is solved. The decomposition is usually built with a graph partitioning approach which is not weIl suited for aIl parallel applications. The graph partitioning approach provides computing lime over the subdomains which can vary from simple to double for our parallel multiple method. However the global method computing time can be decreased by load balancing the subdomain computing time. This thesis proposes a load balancing method to correct in computational volume an initial decomposition issued from graph partitioning tools. The load balancing method is controled by our computation volume estimators of the parallel method
Renouf, Mathieu. "Optimisation numérique et calcul parallèle pour l'étude de milieux divisés bi- et tridimensionnels." Phd thesis, Université Montpellier II - Sciences et Techniques du Languedoc, 2004. http://tel.archives-ouvertes.fr/tel-00007686.
Full textKechid, Mounir. "La programmation dynamique non-serial dans les modèles de calcul parallèle BSP/CGM." Amiens, 2009. http://www.theses.fr/2009AMIE0110.
Full textWe attend this decade a trend (migration) of the parallel hardware towards coarse-grain multiprocessor systems. However, the majority of the traditional parallel software is designed for fine-grain system and for shared memory machines. One of the main current challenges of the researchers in design of parallel algorithms is to reduce this incompatibility said “software-hardware gap”. Big interest is so focused on the design of efficient parallel algorithms for coarse-grain multi-processors systems. It's in this context that this thesis contributes. We use the BSP/CGM parallel computing model (Bulk synchronous parallel Coarse Grained Multicomputer) to design solutions for problems using dynamic programming approach. We are interested in a typical sample of the dynamic programming polyadic non-serial which is characterized by a very strong dependence of calculations. It is about an important class of problems widely used in the high-performance applications (MCOP: the matrix chain ordering problem, OBST: the optimal binary search tree problem, CTP: the convex polygons triangulation). We firstly present a detailed study of the design tools, i. E. BSP/CGM model, as well as a proposition of refinement of the BSP cost model to improve its prediction accuracy. We present then a generic BSP/CGM solution for the aforesaid problems class. At the end, after a study of the constraints of the acceleration of this generic solution in BSP/CGM model for the problems MCOP and OBST, two accelerated BSP/CGM algorithms are proposed
Renouf, Mathieu. "Optimisation numérique et calcul parallèle pour l'étude des milieux divisés bi- et tridimensionnels." Montpellier 2, 2004. http://www.theses.fr/2004MON20131.
Full textPINCHEDEZ, KATIA. "Calcul parallèle pour les équations de diffusion et de transport homogènes en neutronique." Paris 11, 1999. http://www.theses.fr/1999PA112048.
Full textKessy, Edgard. "Décomposition de domaine et calcul parallèle distribué : application à la mécanique des fluides." Rouen, 1997. http://www.theses.fr/1997ROUES052.
Full textSénéchaud, Pascale. "Calcul formel et parallélisme : bases de Gröbner booléennes, méthodes de calcul : applications, parallélisation." Grenoble INPG, 1990. http://tel.archives-ouvertes.fr/tel-00337227.
Full text