Η πρόσφατη εξάπλωση που είδε το μοντέλο της παράλληλης επεξεργασίας στους μικροεπεξεργαστές γενικής χρήσης με την εισαγωγή περισσότερων από έναν πυρήνες εντός του ολοκληρωμένου κυκλώματος έφερε νέες απαιτήσεις στις μεθόδους προσομοίωσης που παραδοσιακά χρησιμοποιήθηκαν για την εξερεύνηση νέων αρχιτεκτονικών. Στην εργασία αυτή προτείνεται ένα πλαίσιο και ένα προγραμματιστικό μοντέλο που κάνει χρήση της αρχιτεκτονικής υψηλού βαθμού παραλληλίας CUDA για να επιτύχει επιτάχυνση στην αρχιτεκτονική προσομοίωση πρωτοκόλλων συνοχής κρυφής μνήμης. / The recent adoption of the parallel computing model in general-use microprocessors with the inclusion of more than one cores in the IC has raised new demands for the simulation methodologies that have been traditionally used. In this work, a framework and a programming model are proposed that make use of the highly parallel CUDA platform to accelerate architectural simulation of cache coherency protocols.

Viscoelastic fluid flow with immersed boundaries of complex geometry is widely found both in nature and engineering processes. Examples include haemocytes moving in human blood flow, self-propulsion of microscopic organisms in complex liquids, hydraulic fracturing with sand in oil flow, and suspension flow with viscoelastic medium. Computational modelling of such systems is important for understanding complex biological processes and assisting engineering designs. Conventional simulation methods use conformed meshes to resolve the boundaries of complex geometry. Dynamically updating the conformed mesh is computationally expensive and makes parallelization difficult. In comparison, Cartesian grid methods are more promising for large scale parallel simulation. Using Cartesian grid methods to simulate viscoelastic fluid flow with complex boundaries is a relatively unexplored area. In this thesis, a sharp interface Cartesian grid method (SICG) and a smoothed interface immersed boundary method (SIIB) are developed in order to simulate viscoelastic fluids in complex geometries. The SICG method shows a better prediction of the stress on stationary boundaries while the SIIB method shows reduced non-physical oscillations in the computation of drag and lift forces on moving boundaries. Parallel implementations of both solvers are developed. Convergence of the numerical schemes is shown and the implementations are validated with a few benchmark problems with both stationary and moving boundaries. This study also focuses on the simulation of flows past 2D cylindrical or 3D spherical particles. Lateral migration of particles induced by inertial and viscoelastic effects are investigated with different flow types. Equilibrium positions of inertia-induced migration are reported as a function of the particle Reynolds number and the blockage ratio. The migration in the viscoelastic fluid is simulated from zero elastic number to a finite elastic number. The inclusion of both inertial and viscoelastic effects on the lateral migration of a particle is the first of its kind. New findings are reported for the equilibrium positions of a spherical particle in square duct flow, which suggest the need for future experimental and computational work.

Artificial neural networks have shown great potential and have attracted much research interest. One problem faced when simulating such networks is speed. As the number of neurons increases, the time to simulate and train a network increases dramatically. This makes it difficult to simulate and train a large-scale network system without the support of a high-performance computer system. The solution we present is a "real" parallel system - using a parallel machine to simulate neural networks which are intrinsically parallel applications. SpiNNaker is a scalable massively-parallel computing system under development with the aim of building a general-purpose platform for the parallel simulation of large-scale neural systems. This research investigates how to model large-scale neural networks efficiently on such a parallel machine. While providing increased overall computational power, a parallel architecture introduces a new problem - the increased communication reduces the speedup gains. Modeling schemes, which take into account communication, processing, and storage requirements, are investigated to solve this problem. Since modeling schemes are application-dependent, two different types of neural network are examined - spiking neural networks with spike-time dependent plasticity, and the parallel distributed processing model with the backpropagation learning rule. Different modeling schemes are developed and evaluated for the two types of neural network. The research shows the feasibility of the approach as well as the performance of SpiNNaker as a general-purpose platform for the simulation of neural networks. The linear scalability shown in this architecture provides a path to the further development of parallel solutions for the simulation of extremely large-scale neural networks.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Nous développons un modèle basé agents pour reproduire deux anomalies fréquemment observées sur les marchés financiers : distribution leptokurtique des rendements et ampleur de la volatilité irrégulière mais persistante de ces mêmes rendements. Notre but est de montrer de façon probante que ces anomalies pourraient être attribuées à une formation mimétique des anticipations des intervenants sur les marchés. Nous nous éloignons des développements récents dans le domaine des modèles modèles basés agents en finance pour proposer un modèle très simple, estimé à partir des traits statistiques saillants de l’indice français journalier CAC 40. L’hypothèse d’anticipations mimétiques peut ainsi être testée : elle n’est pas rejetée dans notre modélisation. / We develop an agent-based model to replicate two frequently observed anomalies in the financial markets: the fat tails and the clustered volatility of the distribution of the returns. Our goal is to show conclusively that these anomalies could be attributed to a mimetic formation of the expectations of the stakeholders in the markets. We did not follow the rencent developpments in the field of the ACE model in the finance, but we propose a very simple model which is estimated from the stylized facts of the French daily index CAC 40. The hypothesis of mimetic anticipations can thus be tested: it is not rejected in our modeling.