Efficient Implementation of 3D Finite Difference Schemes on Recent Processors Abstract In this paper a solver is introduced that solves a problem set modelled by the Burgers equation using the finite difference method: forward in time and central in space. The solver is parallelized and optimized for Intel Xeon Phi 7120P as well as Intel Xeon E5-2699v3 processors to investigate differences in terms of performance between the two architectures. Optimized data access and layout have been implemented to ensure good cache utilization. Loop tiling strategies are used to adjust data access with respect to the L2 cache size. Compiler hints describing aligned memory access are used to support vectorization on both processors. Additionally, prefetching strategies and streaming stores have been evaluated for the Intel Xeon Phi. Parallelization was done using OpenMP and MPI. The parallelisation for native execution on Xeon Phi is based on OpenMP and yielded a raw performance of nearly 100 GFLOP/s, reaching a speedup of almost 50 at a 83\% parallel efficiency. An OpenMP implementation on the E5-2699v3 (Haswell) processors produced up to 292 GFLOP/s, reaching a speedup of almost 31 at a 85\% parallel efficiency. For comparison a mixed implementation using interleaved communications with computations reached 267 GFLOP/s at a speedup of 28 with a 87\% parallel efficiency. Running a pure MPI implementation on the PDC's Beskow supercomputer with 16 nodes yielded a total performance of 1450 GFLOP/s and for a larger problem set it yielded a total of 2325 GFLOP/s, reaching a speedup and parallel efficiency at resp. 170 and 33,3\% and 290 and 56\%. An analysis based on the roofline performance model shows that the computations were memory bound to the L2 cache bandwidth, suggesting good L2 cache utilization for both the Haswell and the Xeon Phi's architectures. Xeon Phi performance can probably be improved by also using MPI. Keeping technological progress for computational cores in the Haswell processor in mind for the comparison, both processors perform well. Improving the stencil computations to a more compiler friendly form might improve performance more, as the compiler can possibly optimize more for the target platform. The experiments on the Cray system Beskow showed an increased efficiency from 33,3\% to 56\% for the larger problem, illustrating good weak scaling. This suggests that problem sizes should increase accordingly for larger number of nodes in order to achieve high efficiency. Frederick Ceder / Effektiv implementering av finita differensmetoder i 3D på moderna processorarkitekturer Sammanfattning Denna uppsats diskuterar implementationen av ett program som kan lösa problem modellerade efter Burgers ekvation numeriskt. Programmet är byggt ifrån grunden och använder sig av finita differensmetoder och applicerar FTCS metoden (Forward in Time Central in Space). Implementationen paralleliseras och optimeras på Intel Xeon Phi 7120P Coprocessor och Intel Xeon E5-2699v3 processorn för att undersöka skillnader i prestanda mellan de två modellerna. Vi optimerade programmet med omtanke på dataåtkomst och minneslayout för att få bra cacheutnyttjande. Loopblockningsstrategier används också för att dela upp arbetsminnet i mindre delar för att begränsa delarna i L2 cacheminnet. För att utnyttja vektorisering till fullo så används kompilatordirektiv som beskriver minnesåtkomsten, vilket ska hjälpa kompilatorn att förstå vilka dataaccesser som är alignade. Vi implementerade också prefetching strategier och streaming stores på Xeon Phi och disskuterar deras värde. Paralleliseringen gjordes med OpenMP och MPI. Parallelliseringen för Xeon Phi:en är baserad på bara OpenMP och exekverades direkt på chipet. Detta gav en rå prestanda på nästan 100 GFLOP/s och nådde en speedup på 50 med en 83% effektivitet. En OpenMP implementation på E5-2699v3 (Haswell) processorn fick upp till 292 GFLOP/s och nådde en speedup på 31 med en effektivitet på 85%. I jämnförelse fick en hybrid implementation 267 GFLOP/s och nådde en speedup på 28 med en effektivitet på 87%. En ren MPI implementation på PDC's Beskow superdator med 16 noder gav en total prestanda på 1450 GFLOP/s och för en större problemställning gav det totalt 2325 GFLOP/s, med speedup och effektivitet på respektive 170 och 33% och 290 och 56%. En analys baserad på roofline modellen visade att beräkningarna var minnesbudna till L2 cache bandbredden, vilket tyder på bra L2-cache användning för både Haswell och Xeon Phi:s arkitekturer. Xeon Phis prestanda kan förmodligen förbättras genom att även använda MPI. Håller man i åtanke de tekniska framstegen när det gäller beräkningskärnor på de senaste åren, så preseterar både arkitekturer bra. Beräkningskärnan av implementationen kan förmodligen anpassas till en mer kompilatorvänlig variant, vilket eventuellt kan leda till mer optimeringar av kompilatorn för respektive plattform. Experimenten på Cray-systemet Beskow visade en ökad effektivitet från 33,3% till 56% för större problemställningar, vilket visar tecken på bra weak scaling. Detta tyder på att effektivitet kan uppehållas om problemställningen växer med fler antal beräkningsnoder. Frederick Ceder

No description available.

No description available.

The author presents some generalized Burgers' equations for incompressible and isothermal flow of viscous non-Newtonian fluids based on the Cross model, the Carreau model, and the Power-Law model and some simple assumptions on the flows. The author numerically solves the traveling wave equations for the Cross model, the Carreau model, the Power-Law model by using industrial data. The author proves existence and uniqueness of solutions to the traveling wave equations of each of the three models. The author also provides numerical estimates of the shock thickness as well as maximum strain $\varepsilon_{11}$ for each of the fluids.

This PhD thesis treats some problems concerning nonlinear differential equations. In the first two papers computer-assisted proofs are used. The differential equations there are rewritten as fixed point problems, and the existence of solutions are proved. The problem in the first paper is one-dimensional; with one boundary condition given by an integral. The problem in the second paper is three-dimensional, and Dirichlet boundary conditions are used. Both problems have their origins in fluid dynamics. Paper III describes an inverse problem for the heat equation. Given the solution, a solution dependent diffusion coefficient is estimated by intervals at a finite set of points. The method includes the construction of set-valued level curves and two-dimensional splines. In paper IV we prove that there exists a unique, globally attracting fixed point for a differential equation system. The differential equation system arises as the number of peers in a peer-to-peer network, which is described by a suitably scaled Markov chain, goes to infinity. In the proof linearization and Dulac's criterion are used.

Submitted by Leonardo Cavalcante (leo.ocavalcante@gmail.com) on 2018-05-03T13:44:12Z No. of bitstreams: 1 Arquivototal.pdf: 1150863 bytes, checksum: a7e25ab87986c9d088c0fe224303f97f (MD5) / Made available in DSpace on 2018-05-03T13:44:12Z (GMT). No. of bitstreams: 1 Arquivototal.pdf: 1150863 bytes, checksum: a7e25ab87986c9d088c0fe224303f97f (MD5) Previous issue date: 2017-03-31 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / In this thesis we presents results on the exact controllability of the partial Di erential Equations (PDEs) of the parabolic and hyperbolic type, in the context of hierarchic control, using the Stackelberg-Nash strategy. In every problems we consider a main control (leader) and two secondary controls (followers). To each leader we obtain a correnponding Nash equilibrium, associated to a bi-objective optimal control problem; then we look for a leader of minimal cost that solves the exact controllability problem. For the parabolic problems we have distributed and boundary controls, now in the hyperbolics every controls are distributed. We consider linear and semilinear cases, which we solve using observability inequality obtained combining right Carleman inequalities. Also we use a xed point method. / Nesta tese apresentamos resultados sobre controlabilidade exata de Equações Diferenciais Parciais (EDPs) dos tipos parabólico e hiperbólico, no contexto de controle hierárquico, usando a estratégia de Stackelberg-Nash. Em todos os problemas consideramos um controle principal (líder) e dois controles secundários (seguidores). Para cada líder obtemos um equil íbrio de Nash correspondente, associado a um problema de controle ótimo bi-objetivo; então buscamos o líder de custo que resolve o problema de controlabilidade. Para os problemas parabólicos temos controles distribuídos e na fronteira, já nos hiperbólico todos os controles são distribuídos. Consideramos casos lineares e semilineares, os quais resolvemos usando desigualdade de observabilidade obtidas combinando desigualdades de Carleman adequadas. Também usamos um método de ponto xo.

Turbulence is ubiquitous in the flows of fluids and plasmas. This thesis is devoted to studies of the statistical properties of turbulence in the three-dimensional (3D) Hall magnetohydrodynamic (Hall-MHD) equations, the two-dimensional (2D) MHD equations, the one-dimensional (1D) hyperviscous Burgers equation, and the 3D Navier-Stokes equations. Chapter 1 contains a brief introduction to statistically homogeneous and isotropic turbulence. This is followed by an over-view of the equations we study in the subsequent chapters, the motivation for the studies and a summary of problems we investigate in chapters 2-6. In Chapter 2 we present our study of Hall-MHD turbulence [1]. We show that a shell-model version of the 3D Hall-MHD equations provides a natural theoretical model for investigating the multiscaling behaviors of velocity and magnetic structure functions. We carry out extensive numerical studies of this shell model, obtain the scaling exponents for its structure functions, in both the low-k and high-k power-law ranges of 3D Hall-MHD, and find that the extended-self-similarity procedure is helpful in extracting the multiscaling nature of structure functions in the high-k regime, which otherwise appears to display simple scaling. Our results shed light on intriguing solar-wind measurements. In Chapter 3 we present our study of the inverse-cascade regime in two-dimensional magnetohydrodynamic turbulence [2]. We present a detailed direct numerical simulation (DNS) of statistically steady, homogeneous, isotropic, two-dimensional magnetohydrodynamic (2D MHD) turbulence. Our study concentrates on the inverse cascade of the magnetic vector potential. We examine the dependence of the statistical properties of such turbulence on dissipation and friction coefficients. We extend earlier work significantly by calculating fluid and magnetic spectra, probability distribution functions (PDFs) of the velocity, magnetic, vorticity, current, stream-function, and magnetic-vector-potential fields and their increments. We quantify the deviations of these PDFs from Gaussian ones by computing their flatnesses and hyperflatnesses. We also present PDFs of the Okubo-Weiss parameter, which distinguishes between vortical and extensional flow regions, and its magnetic analog. We show that the hyperflatnesses of PDFs of the increments of the stream-function and the magnetic vector potential exhibit significant scale dependence and we examine the implication of this for the multiscaling of structure functions. We compare our results with those of earlier studies. In Chapter 4 we compare the statistical properties of 2D MHD turbulence for two different energy injection scales. We present systematic DNSs of statistically steady 2D MHD turbulence. Our two DNSs are distinguished by kinj, the wave number at which we inject energy into the system. In our first DNS (run R1), kinj = 2 and, in the second (run R2) kinj = 250. We show that various statistical properties of the turbulent states in the runs R1 and R2 are strikingly different The nature of energy spectrum, probability distribution functions, and topological structures are compared for the two runs R1 and R2 are found to be strikingly different. In Chapter 5 we study the hyperviscous Burgers equation for very high α, order of hyperviscosity [3]. We show, by using direct numerical simulations and theory, how, by increasing α in equations of hydrodynamics, there is a transition from a dissipative to a conservative system. This remarkable result, already conjectured for the asymptotic case α →∞ [U. Frisch et al., Phys. Rev. Lett. 101, 144501 (2008)], is now shown to be true for any large, but finite, value of α greater than a crossover value α crossover. We thus provide a self-consistent picture of how dissipative systems, under certain conditions, start behaving like conservative systems, and hence elucidate the subtle connection between equilibrium statistical mechanics and out-of-equilibrium turbulent flows. In Chapter 6 we show how to use asymptotic-extrapolation and Richardson extrapolation methods to extract the exponents ξ p that characterize the dependence of the order-p moments of the velocity gradients on the Reynolds number Re. To use these extrapolation methods we must have high-precision data for such moments. We obtain these high-precision data by carrying out the most extensive, quadruple precision, pseudospectral DNSs of the Navier-Stokes equation.

L'objet de cette thèse est de proposer un cadre complet, du haut niveau au bas niveau, de génération de trajectoires pour un groupe de systèmes dynamiques indépendants. Ce cadre, basé sur la résolution de l'équation de Burgers pour la génération de trajectoires, est appliqué à un modèle original de drone trirotor et utilise la platitude des deux systèmes différentiels considérés. La première partie du manuscrit est consacrée à la génération de trajectoires. Celle-ci est effectuée en créant formellement, par le biais de la platitude du système considéré, des solutions à l'équation de la chaleur. Ces solutions sont transformées en solution de l'équation de Burgers par la transformation de Hopf-Cole pour correspondre aux formations voulues. Elles sont optimisées pour répondre à des contraintes spécifiques. Plusieurs exemples de trajectoires sont donnés.La deuxième partie est consacrée au suivi autonome de trajectoire par un drone trirotor. Ce drone est totalement actionné et un contrôleur en boucle fermée non-linéaire est proposé. Celui-ci est testé en suivant, en roulant, des trajectoires au sol et en vol. Un modèle est présenté et une démarche pour le contrôle est proposée pour transporter une charge pendulaire. / This thesis is dedicated to the creation of a complete framework, from high-level to low-level, of trajectory generation for a group of independent dynamical systems. This framework, based for the trajectory generation, on the resolution of Burgers equation, is applied to a novel model of trirotor UAV and uses the flatness of the two levels of dynamical systems.The first part of this thesis is dedicated to the generation of trajectories. Formal solutions to the heat equation are created using the differential flatness of this equation. These solutions are transformed into solutions to Burgers' equation through Hopf-Cole transformation to match the desired formations. They are optimized to match specific requirements. Several examples of trajectories are given.The second part is dedicated to the autonomous trajectory tracking by a trirotor UAV. This UAV is totally actuated and a nonlinear closed-loop controller is suggested. This controller is tested on the ground and in flight by tracking, rolling or flying, a trajectory. A model is presented and a control approach is suggested to transport a pendulum load.

Transportphänomene treten in biologischen und künstlichen Systemen auf allen Längenskalen auf. In dieser Arbeit untersuchen wir sie für verschiedene Systeme aus einer mesoskopischen Perspektive, in der Fluktuationen physikalischer Größen um ihre Mittelwerte eine wichtige Rolle spielen. Im ersten Teil untersuchen wir die persistente Bewegung aktiver Brownscher Teilchen mit zusätzlichem Drehmoment, wie sie z.B. für Spermien oder Janus Teilchen auftritt. Wird ihre Bewegung auf einen Tunnel variierender Breite beschränkt, so setzt im thermischen Nichtgleichgewicht Transport ein; ungerichtete Fluktuationen des rauschhaften Antriebs werden gleichgerichtet. Hierdurch wird ein neuer Ratschentyp realisiert. Im zweiten Teil untersuchen wir den intrazellulären Cargotransport in den Axonen von Nervenzellen mithilfe molekularer Motoren. Sie werden als asymmetrischer Ausschlussprozess simuliert. Zusätzlich können die Cargos zwischen benachbarten Motoren ausgetauscht werden. Dadurch lassen sich charakteristische Eigenschaften des langsamen axonalen Transports mit einer einzigen Motorspezies reproduzieren. Bewerkstelligt wird dies durch die transiente Anbindung der Cargos an rückwärtslaufende Motorstaus. Im dritten Teil diskutieren wir resistive switching, die nicht volatile Widerstandsänderung eines Dielektrikums durch elektrische Impulse. Es wird für Anwendungen im Computerspeicher ausgenutzt, dem resistive RAM. Wir schlagen ein auf Sauerstoffvakanzen basierendes stochastisches Gitterhüpfmodell vor. Wir definieren binäre logische Zustände mit Hilfe der zugrunde liegenden Vakanzenverteilung und definieren Schreibe- und Leseoperationen durch Spannungsimpulse für ein solches Speicherelement. Überlegungen über die Unterscheidbarkeit dieser Operationen unter Fluktuationen zusammen mit der Deutlichkeit der unterschiedlichen Widerstandszustände selbst ermöglichen es uns, eine optimale Vakanzenzahl vorherzusagen. / Transport phenomena occur in biological and artificial systems at all length scales. In this thesis, we investigate them for various systems from a mesoscopic perspective, in which fluctuations around their average properties play an important role. In the first part, we investigate the persistent diffusive motion of active Brownian particles with an additional torque. It can appear in many real life systems, for example in sperm cells or Janus particles. If their motion is confined to a tunnel of varying width, transport arises out of thermal equilibrium; unbiased fluctuations of the noisy drive are rectified. This way, we have realized a novel kind of ratchet. In the second part, we study intracellular cargo transport in the axons of nerve cells by molecular motors. They are modeled by an asymmetric exclusion process. In a new approach, we add a cargo exchange interaction between the motors. This way, the characteristics of slow axonal transport can be accounted for with a single motor species. It is explained by the transient attachment of cargos to reverse walking motors jams. In the third part, we discuss resistive switching, the non-volatile change of resistance in a dielectric due to electric pulses. It is exploited for applications in computer memory, the resistive random access memory (ReRAM). We propose a stochastic lattice hopping model based on the on oxygen vacancies. We define binary logical states by means of the underlying vacancy distributions, and establish a framework of writing and reading such a memory element with voltage pulses. Considerations about the discriminability of these operations under fluctuations together with the markedness of the resistive switching effect itself enable us to predict an optimal vacancy number.

Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro / O processo de recuperação secundária de petróleo é comumente realizado com a injeção de água no reservatório a fim de manter a pressão necessária para sua extração. Para que o investimento seja viável, os gastos com a extração têm de ser menores do que o retorno financeiro obtido com o petróleo. Para tanto, tornam-se extremamente importantes as simulações dos processos de extração. Neste trabalho são estudados os problemas de Burgers e de Buckley-Leverett visando o escoamento imiscível água-óleo em meios porosos, onde o escoamento é incompressível e os efeitos difusivos (devido à pressão capilar) são desprezados. Com o objetivo de incorporar conhecimento matemático mais avançado, para em seguida utilizá-lo no entendimento do problema estudado, abordou-se com razoável profundidade a teoria das leis de conservação. Foram consideradas soluções fracas que, fisicamente, podem ser interpretadas como ondas de choque ou rarefações, então, para que fossem distinguidas as fisicamente admissíveis, foi utilizado o princípio de entropia, nas suas diversas formas. Inicialmente consideramos alguns exemplos clássicos de métodos numéricos para uma lei de conservação escalar, os quais podem ser vistos como esquemas conservativos de três pontos. Entre eles, o método de Lax-Friedrichs (LF) e o método de Lax-Wendroff (LW). Em seguida, um esquema composto foi testado, o qual inclui na sua formulação os métodos LF e LW (chamado de LWLF-4). Respeitando a condição CFL, foram obtidas soluções numéricas de todos os problemas tratados aqui. Com o objetivo de validar tais soluções, foram utilizadas soluções analíticas oriundas dos problemas de Burgers e Buckley- Leverett. Também foi feita uma comparação com os métodos do tipo TVDs com limitadores de fluxo, obtendo resultado satisfatório. Vale à pena ressaltar que o esquema LWLF-4, pelo que nos consta, nunca foi antes utilizado nas resoluções das equações de Burgers e Buckley- Leverett. / The secondary recovery of petroleum is usually performed with injection of water through an oil reservoir to keep the oil pressure for the exploration. In order to make the exploration profitable, the extraction cost must be less than the financial return, which means that the simulation of the exploration process is extremely relevant. In this work, the Burgers- and- Buckley-Leverett problems are studied seeking a two-phase displacement in porous media. The flow is considered incompressible and capillary effects are ignored. In order to analyze the problem, it was necessary to use the theory of conservation law in a spatial variable. Weak solutions, which can be understood as shock or rarefaction waves, are studied with the entropy condition, so that only the physically correct solutions are considered. Some classical numerical methods, which can be seen as conservative schemes of three points, are studied, among them the Lax-Friedrichs (LF) and Lax-Wendroff (LW) methods. A composite scheme, called LWLF-k, is tested using LF and LW methods, being respected the CFL condition, with satisfactory results. In order to validate the numerical schemes, we consider analytical solutions of the Burgers-and-Buckley-Leverett equations. Was also made a comparison with TVDs methods with flux limiters, obtaining satisfactory results. We emphasize that to the best of our knowledge, the LWLF-4 scheme has never been used to solve the Buckley-Leverett equation.