The gap between processor speed and memory latency has led to the use of caches in the memory systems of modern computers. Programs must use the caches efficiently and exploit data locality for maximum performance. Multiprocessors, built from many processing units, are becoming commonplace not only in large servers but also in smaller systems such as personal computers. Multiprocessors require careful data locality optimizations since accesses from other processors can lead to invalidations and false sharing cache misses. This thesis explores hardware and software approaches for creating and exploiting temporal and spatial locality in multiprocessors. We propose the capacity prefetching technique, which efficiently reduces the number of cache misses but avoids false sharing by distinguishing between cache lines involved in communication from non-communicating cache lines at run-time. Prefetching techniques often lead to increased coherence and data traffic. The new bundling technique avoids one of these drawbacks and reduces the coherence traffic in multiprocessor prefetchers. This is especially important in snoop-based systems where the coherence bandwidth is a scarce resource. Most of the studies have been performed on advanced scientific algorithms. This thesis demonstrates that a cc-NUMA multiprocessor, with hardware data migration and replication optimizations, efficiently exploits the temporal locality in such codes. We further present a method of parallelizing a multigrid Gauss-Seidel partial differential equation solver, which creates temporal locality at the expense of increased communication. Our conclusion is that on modern chip multiprocessors, it is more important to optimize algorithms for data locality than to avoid communication, since communication can take place using a shared cache.

This thesis develops efficient modeling frameworks via a Partial Differential Equation (PDE) approach for multi-factor financial derivatives, with emphasis on three-factor models, and studies highly efficient implementations of the numerical methods on novel high-performance computer architectures, with particular focus on Graphics Processing Units (GPUs) and multi-GPU platforms/clusters of GPUs. Two important classes of multi-factor financial instruments are considered: cross-currency/foreign exchange (FX) interest rate derivatives and multi-asset options. For cross-currency interest rate derivatives, the focus of the thesis is on Power Reverse Dual Currency (PRDC) swaps with three of the most popular exotic features, namely Bermudan cancelability, knockout, and FX Target Redemption. The modeling of PRDC swaps using one-factor Gaussian models for the domestic and foreign interest short rates, and a one-factor skew model for the spot FX rate results in a time-dependent parabolic PDE in three space dimensions. Our proposed PDE pricing framework is based on partitioning the pricing problem into several independent pricing subproblems over each time period of the swap's tenor structure, with possible communication at the end of the time period. Each of these subproblems requires a solution of the model PDE. We then develop a highly efficient GPU-based parallelization of the Alternating Direction Implicit (ADI) timestepping methods for solving the model PDE. To further handle the substantially increased computational requirements due to the exotic features, we extend the pricing procedures to multi-GPU platforms/clusters of GPUs to solve each of these independent subproblems on a separate GPU. Numerical results indicate that the proposed GPU-based parallel numerical methods are highly efficient and provide significant increase in performance over CPU-based methods when pricing PRDC swaps. An analysis of the impact of the FX volatility skew on the price of PRDC swaps is provided. In the second part of the thesis, we develop efficient pricing algorithms for multi-asset options under the Black-Scholes-Merton framework, with strong emphasis on multi-asset American options. Our proposed pricing approach is built upon a combination of (i) a discrete penalty approach for the linear complementarity problem arising due to the free boundary and (ii) a GPU-based parallel ADI Approximate Factorization technique for the solution of the linear algebraic system arising from each penalty iteration. A timestep size selector implemented efficiently on GPUs is used to further increase the efficiency of the methods. We demonstrate the efficiency and accuracy of the proposed GPU-based parallel numerical methods by pricing American options written on three assets.

The analytical solution of a given inhomogeneous boundary value problem of a linear, elliptic, inhomogeneous partial differential equation and a set of inhomogeneous mixed Dirichlet- and Neumann-type boundary conditions is derived in the present paper. In the context of elasticity theory, the problem arises for a non-conservative symmetric ansatz and an extended constitutive law shown earlier. For convenient user application, the scalar function expressed in cylindrical coordinates is primarily obtained for the general case before being expatiated on a special case of linear boundary conditions.

In addition to previous publications, the paper presents the analytical solution of a special boundary value problem which arises in the context of elasticity theory for an extended constitutive law and a non-conservative symmetric ansatz. Besides deriving the general analytical solution, a specific form for linear boundary conditions is given for user convenience.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / In this thesis we present a proof of the Malgrange-Ehrenpreis theorem, which states that every operator with constant coefficients non identically zero has a fundamental solution. / Nessa dissertação apresentamos uma demonstração do Teorema de Malgrange-Ehrenpreis, que afirma que todo operador de coeficientes constantes não identicamente nulo tem uma solução fundamental.

Made available in DSpace on 2014-06-11T19:26:56Z (GMT). No. of bitstreams: 0 Previous issue date: 2005-02-25Bitstream added on 2014-06-13T20:27:30Z : No. of bitstreams: 1 dippolito_km_me_sjrp.pdf: 838424 bytes, checksum: 9eb5d64b517c6606a595f44d889f6cd5 (MD5) / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / O objetivo deste trabalho þe apresentar um estudo sobre a aplicação de métodos numéricos para a resolução do modelo proposto por Barcelos, Boaventura e Silva Jr. [7], para a eliminação de ruídos em imagens digitais por meio de uma equação diferencial parcial, e propor uma anþalise da estabilidade do mþetodo iterativo comumente aplicado a este modelo. Uma anþalise comparativa entre os vários mþetodos abordados þe realizada atravþes de resultados experimentais em imagens sintéticas e imagens da vida real. / The purpose of this work is to present a study on the application of numerical methods for the resolution of model considered by Barcelos, Boaventura and Silva Jr [7], for image denoising through a partial di erential equation, and to consider a stability analysis of an iterative method usually applied to this model. A comparative analysis among various considered methods is carried out through experimental results for synthetic and real images.

The primary goal of this thesis is to study finite element based a priori and a posteriori error estimates of optimal control problems of various kinds governed by linear elliptic PDEs (partial differential equations) of second and fourth orders. This thesis studies interior and boundary control (Neumann and Dirichlet) problems. The initial chapter is introductory in nature. Some preliminary and fundamental results of finite element methods and optimal control problems which play key roles for the subsequent analysis are reviewed in this chapter. This is followed by a brief literature survey of the finite element based numerical analysis of PDE constrained optimal control problems. We conclude the chapter with a discussion on the outline of the thesis. An abstract framework for the error analysis of discontinuous Galerkin methods for control constrained optimal control problems is developed in the second chapter. The analysis establishes the best approximation result from a priori analysis point of view and delivers a reliable and efficient a posteriori error estimator. The results are applicable to a variety of problems just under the minimal regularity possessed by the well-posedness of the problem. Subsequently, the applications of p p - interior penalty methods for a boundary control problem as well as a distributed control problem governed by the bi-harmonic equation subject to simply supported boundary conditions are discussed through the abstract analysis. In the third chapter, an alternative energy space based approach is proposed for the Dirichlet boundary control problem and then a finite element based numerical method is designed and analyzed for its numerical approximation. A priori error estimates of optimal order in the energy norm and the m norm are derived. Moreover, a reliable and efficient a posteriori error estimator is derived with the help an auxiliary problem. An energy space based Dirichlet boundary control problem governed by bi-harmonic equation is investigated and subsequently a l y - interior penalty method is proposed and analyzed for it in the fourth chapter. An optimal order a priori error estimate is derived under the minimal regularity conditions. The abstract error estimate guarantees optimal order of convergence whenever the solution has minimum regularity. Further an optimal order l l norm error estimate is derived. The fifth chapter studies a super convergence result for the optimal control of an interior control problem with Dirichlet cost functional and governed by second order linear elliptic PDE. An optimal order a priori error estimate is derived and subsequently a super convergence result for the optimal control is derived. A residual based reliable and efficient error estimators are derived in a posteriori error control for the optimal control. Numerical experiments illustrate the theoretical results at the end of every chapter. We conclude the thesis stating the possible extensions which can be made of the results presented in the thesis with some more problems of future interest in this direction.

We study asymptotic properties of evolution partial differential equations posed in unbounded spatial domain in the context of locally uniform spaces. This context allows the use of non-integrable data and carries an inherent non-compactness and non-separability. We establish the existence of a lo- cally compact attractor for non-local parabolic equation and weakly damped semilinear wave equation and provide an upper bound on the Kolmogorov's ε-entropy of these attractors and the attractor of strongly damped wave equation in the subcritical case using the method of trajectories. Finally we also investigate infinite dimensional exponential attractors of nonlinear reaction-diffusion equation in its natural energy setting. 1

Dans nos travaux, nous avons considéré un processus de Lévy X avec une composante brownienne non nulle et dont la partie à sauts est un processus de Poisson composé. Nous avons supposé que la valeur d'une entreprise est modélisée par un processus stochastique de la forme V = Vo exp X et que cette entreprise est mise à défaut dès lors que sa valeur passe sous un certain seuil b déterminé de façon exogène et qui donc, est une donnée du problème. L'instant de défaut T est alors de la forme Tx pour x= ln(Vo) ln((b) où x> 0, Tx = inf{t 2:0: X, 2:x}. Dans un premier temps, nous supposons que des agents observant la valeur V des ac­tifs de la firme souhaitent connaître le comportement de l'instant de défaut. Dans ce modèle, au chapitre 2, nous avons étudié d'une part la régularité de la densité de la loi de l'instant de défaut. D'autre part, nous avons étudié la loi conjointe de l'instant de défaut, de l'overshoot et de l'undershoot. Au chapitre 3, nous avons obtenu une équation à valeurs mesures dont le quadriplet formé par la variable aléatoire X,, le su­ premum du processus X à l'instant t, le supremum du processus X au dernier instant de saut avant l'instant t et le dernier instant de saut à l'instant t est solution au seris faible, puis une équation dont ce quadriplet est une solution forte. Dans un second temps, au chapitre 4, nous avons supposé que des investisseurs souhaitant détenir une part de cette entreprise ne disposent pas de l'information complète. Ils n'observent pas la valeur des actifs de la firme V, mais sa valeur bruitée. Leur information est modélisée par la filtration Ç = (Ç,, t 2: 0) engendrée par cette observation. Dans ce modèle, nous avons montré que la loi conditionnelle de l'instant de défaut sachant la tribu Ç, admet une densité par rapport à la mesure de Lebesgue et obtenu une équation de Volttera dont cette densité est solution. Cette connaissance permet aux investisseurs de prévoir au vu de leur information, quand est-ce que l'instant de défaut va intervenir après l'instant t. Nous avons complété ce travail par des simulations numériques. / In this Ph.D thesis, we consider a jump-diffusion process which the diffusion part is a drifted Brownian motion and the jump part is a compound Poisson process. We assume that a firm value is modelling by a stochastic process V = V0 exp-X. This firm goes to default whenever its value is below a specified tlrreshold b which is exo­ genously determined. For x = ln(Vo) - ln(b) > 0, the default time is of the form Tx = inf{t 2:0: X, 2: x}. First, we suppose that agents observe perfectly the firm value. In this mode, we sho­ wed in chapter 2 that the density of the default time is continuons, then study the joint law of the default time, overshoot an undershoot. We obtained in chapter 3 a valued measure differentia equation which the solution is the quadruplet formed by the random variable X,, the running supremum x; of X at time t, the supremum of X at the last jump time before t and the last jump time before t. Secondly, we assume that investors wishing detain a part of the firm can not observe the firm value. They observe a noisy value of the firm and their information is madel­ ling by the filtration g = (9,,t 2: 0) generated by their observation. In this mode, we have shown that the conditional density of Tx with respect to Ç has a density which is solution of one stochastic integral-differentia equation The knowledge of this density allows investors to predict the default time after time t. This second part is the chapter 4.

Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2018-03-27T18:18:55Z No. of bitstreams: 1 stiwharrisonherrerataipe.pdf: 5886256 bytes, checksum: 4fa85d1d9808790a2f5c85bb6c6c8d8d (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2018-03-28T16:44:29Z (GMT) No. of bitstreams: 1 stiwharrisonherrerataipe.pdf: 5886256 bytes, checksum: 4fa85d1d9808790a2f5c85bb6c6c8d8d (MD5) / Made available in DSpace on 2018-03-28T16:44:30Z (GMT). No. of bitstreams: 1 stiwharrisonherrerataipe.pdf: 5886256 bytes, checksum: 4fa85d1d9808790a2f5c85bb6c6c8d8d (MD5) Previous issue date: 2018-01-26 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Neste trabalho estamos interessados em estudar, mediante simulações computacionais, se o aquecimento eletromagnético é capaz de melhorar o deslocamento do óleo pela água. Nesta direção, nos baseamos nos resultados obtidos pela equipe da TU Delft da Holanda, que desenvolveu experimentos de laboratório que demonstravam a distribuição da temperatura em um meio poroso, onde o óleo está sendo deslocado pela injeção de água, gerada por aquecimento eletromagnético. Para tanto, definimos o modelo matemático que governa o problema em questão regido por equações diferenciais parciais das leis de conservação de massa e energia. Assim, partindo da caracterização do contínuo e estendendo a lei de Darcy para o caso multifásico, através da introdução do conceito de permeabilidades relativas dos fluidos, derivamos um sistema acoplado de equações diferenciais parciais com coeficientes variáveis e termos não lineares formulados em função da velocidade de Darcy para o escoamento bifásico (água, óleo) aquecido por ondas eletromagnéticas. O modelo matemático é discretizado utilizando o método de diferenças finitas no tempo e no espaço e a técnica Splitting. Dessa forma dividimos o sistema de equações diferencias parciais em dois subsistemas. O primeiro subsistema consiste em resolver a parte difusiva e reativa e o segundo subsistema tem por objetivo a resolução do termo convectivo. O método numérico desenvolvido é validado por simulações computacionais que visam a comparação com os resultados obtidos experimentalmente e com soluções semi-analíticas, para este problema, que foram derivadas pelo método do princípio de Duhamel. Além disso, o método proposto quando aplicado para o caso geral da simulação do escoamento bifásico com aquecimento eletromagnético demonstrou um ganho de 1.67%, se comparado ao método sem aquecimento. / In this work we are interested in studying, through computational simulations, if the electromagnetic heating is able to improve the displacement of the oil by water. In this direction, we rely on the results obtained by the TU Delft team from the Netherlands, which developed laboratory experiments that demonstrated the temperature distribution in a porous medium where the oil is being displaced by the injection of water generated by electromagnetic heating. For this, we define the mathematical model that governs the problem in question governed by partial differential equations of the laws of conservation of mass and energy. Thus, starting from the characterization of the continuum and extending Darcy’s law to the multiphase case, by introducing the concept of relative permeabilities of fluids, we derive a coupled system of partial differential equations with variable coefficients and non-linear terms formulated as a function of the velocity of Darcy for two-phase flow (water, oil) heated by electromagnetic waves. The mathematical model is discretized using the finite difference method in time and space and the Splitting technique. In this way we divide the system of partial differential equations into two subsystems. The first subsystem consists of solving the diffusive and reactive part and the second subsystem aims to solve the convective term. The numerical method developed is validated by computational simulations aimed at the comparison with the results obtained experimentally and with semi-analytical solutions, for this problem, which were derived by the Duhamel principle method. In addition, the proposed method when applied to the general case of simulation of the biphasic flow with electromagnetic heating demonstrated a gain of 1.67%, when compared to the non-heating method.