Dans le contexte de cette thèse, nous nous focalisons sur des algorithmes pour l’algèbre linéaire numérique, plus précisément sur la résolution de grands systèmes linéaires creux. Nous mettons au point des méthodes de parallélisation pour le solveur linéaire hybride MaPHyS. Premièrement nous considerons l'aproche MPI+threads. Dans MaPHyS, le premier niveau de parallélisme consiste au traitement indépendant des sous-domaines. Le second niveau est exploité grâce à l’utilisation de noyaux multithreadés denses et creux au sein des sous-domaines. Une telle implémentation correspond bien à la structure hiérarchique des supercalculateurs modernes et permet un compromis entre les performances numériques et parallèles du solveur. Nous démontrons la flexibilité de notre implémentation parallèle sur un ensemble de cas tests. Deuxièmement nous considérons un approche plus innovante, où les algorithmes sont décrits comme des ensembles de tâches avec des inter-dépendances, i.e., un graphe de tâches orienté sans cycle (DAG). Nous illustrons d’abord comment une première parallélisation à base de tâches peut être obtenue en composant des librairies à base de tâches au sein des processus MPI illustrer par un prototype d’implémentation préliminaire de notre solveur hybride. Nous montrons ensuite comment une approche à base de tâches abstrayant entièrement le matériel peut exploiter avec succès une large gamme d’architectures matérielles. À cet effet, nous avons implanté une version à base de tâches de l’algorithme du Gradient Conjugué et nous montrons que l’approche proposée permet d’atteindre une très haute performance sur des architectures multi-GPU, multicoeur ainsi qu’hétérogène. / In the context of this thesis, our focus is on numerical linear algebra, more precisely on solution of large sparse systems of linear equations. We focus on designing efficient parallel implementations of MaPHyS, an hybrid linear solver based on domain decomposition techniques. First we investigate the MPI+threads approach. In MaPHyS, the first level of parallelism arises from the independent treatment of the various subdomains. The second level is exploited thanks to the use of multi-threaded dense and sparse linear algebra kernels involved at the subdomain level. Such an hybrid implementation of an hybrid linear solver suitably matches the hierarchical structure of modern supercomputers and enables a trade-off between the numerical and parallel performances of the solver. We demonstrate the flexibility of our parallel implementation on a set of test examples. Secondly, we follow a more disruptive approach where the algorithms are described as sets of tasks with data inter-dependencies that leads to a directed acyclic graph (DAG) representation. The tasks are handled by a runtime system. We illustrate how a first task-based parallel implementation can be obtained by composing task-based parallel libraries within MPI processes throught a preliminary prototype implementation of our hybrid solver. We then show how a task-based approach fully abstracting the hardware architecture can successfully exploit a wide range of modern hardware architectures. We implemented a full task-based Conjugate Gradient algorithm and showed that the proposed approach leads to very high performance on multi-GPU, multicore and heterogeneous architectures.

Au cours des dernières années, s’est développé un intérêt tout particulier pour l’optimisation sans dérivée. Ce domaine de recherche se divise en deux catégories: une déterministe et l’autre stochastique. Bien qu’il s’agisse du même domaine, peu de liens ont déjà été établis entre ces deux branches. Cette thèse a pour objectif de combler cette lacune, en montrant comment les techniques issues de l’optimisation déterministe peuvent améliorer la performance des stratégies évolutionnaires, qui font partie des meilleures méthodes en optimisation stochastique. Sous certaines hypothèses, les modifications réalisées assurent une forme de convergence globale, c’est-à-dire une convergence vers un point stationnaire de premier ordre indépendamment du point de départ choisi. On propose ensuite d’adapter notre algorithme afin qu’il puisse traiter des problèmes avec des contraintes générales. On montrera également comment améliorer les performances numériques des stratégies évolutionnaires en incorporant un pas de recherche au début de chaque itération, dans laquelle on construira alors un modèle quadratique utilisant les points où la fonction coût a déjà été évaluée. Grâce aux récents progrès techniques dans le domaine du calcul parallèle, et à la nature parallélisable des stratégies évolutionnaires, on propose d’appliquer notre algorithme pour résoudre un problème inverse d’imagerie sismique. Les résultats obtenus ont permis d’améliorer la résolution de ce problème. / In recent years, there has been significant and growing interest in Derivative-Free Optimization (DFO). This field can be divided into two categories: deterministic and stochastic. Despite addressing the same problem domain, only few interactions between the two DFO categories were established in the existing literature. In this thesis, we attempt to bridge this gap by showing how ideas from deterministic DFO can improve the efficiency and the rigorousness of one of the most successful class of stochastic algorithms, known as Evolution Strategies (ES’s). We propose to equip a class of ES’s with known techniques from deterministic DFO. The modified ES’s achieve rigorously a form of global convergence under reasonable assumptions. By global convergence, we mean convergence to first-order stationary points independently of the starting point. The modified ES’s are extended to handle general constrained optimization problems. Furthermore, we show how to significantly improve the numerical performance of ES’s by incorporating a search step at the beginning of each iteration. In this step, we build a quadratic model using the points where the objective function has been previously evaluated. Motivated by the recent growth of high performance computing resources and the parallel nature of ES’s, an application of our modified ES’s to Earth imaging Geophysics problem is proposed. The obtained results provide a great improvement for the problem resolution.

La méthode LS-STAG est une méthode de type frontière immergée/cut-cell pour le calcul d’écoulements visqueux incompressibles qui est basée sur la méthode MAC pour grilles cartésiennes décalées, où la frontière irrégulière est nettement représentée par sa fonction level-set, résultant en un gain significatif en ressources informatiques par rapport aux codes MFN commerciaux utilisant des maillages qui épousent la géométrie. La version 2D est maintenant bien établie et ce manuscrit présente son extension aux géométries 3D avec une symétrie translationnelle dans la direction z (configurations extrudées 3D). Cette étape intermédiaire sera considérée comme la clé de voûte du solveur 3D complet, puisque les problèmes de discrétisation et d’implémentation sur les machines à mémoire distribuée sont abordés à ce stade de développement. La méthode LS-STAG est ensuite appliquée à divers écoulements newtoniens et non-newtoniens dans des géométries extrudées 3D (conduite axisymétrique, cylindre circulaire, conduite cylindrique avec élargissement brusque, etc.) pour lesquels des résultats de références et des données expérimentales sont disponibles. Le but de ces investigations est d’évaluer la précision de la méthode LS-STAG, d’évaluer la polyvalence de la méthode pour les applications d’écoulement dans différents régimes (fluides newtoniens et rhéofluidifiants, écoulement laminaires stationnaires et instationnaires, écoulements granulaires) et de comparer ses performances avec de méthodes numériques bien établies (méthodes non structurées et de frontières immergées) / The LS-STAG method is an immersed boundary/cut-cell method for viscous incompressible flows based on the staggered MAC arrangement for Cartesian grids where the irregular boundary is sharply represented by its level-set function. This approach results in a significant gain in computer resources compared to commercial body-fitted CFD codes. The 2D version of LS-STAG method is now well-established and this manuscript presents its extension to 3D geometries with translational symmetry in the z direction (3D extruded configurations). This intermediate step will be regarded as the milestone for the full 3D solver, since both discretization and implementation issues on distributed memory machines are tackled at this stage of development. The LS-STAG method is then applied to Newtonian and non-Newtonian flows in 3D extruded geometries (axisymmetric pipe, circular cylinder, duct with an abrupt expansion, etc.) for which benchmark results and experimental data are available. The purpose of these investigations is to evaluate the accuracy of LS-STAG method, to assess the versatility of method for flow applications at various regimes (Newtonian and shear-thinning fluids, steady and unsteady laminar to turbulent flows, granular flows) and to compare its performance with well-established numerical methods (body-fitted and immersed boundary methods)

Proper metazoan development requires that genes are expressed in a spatiotemporally controlled manner, with tightly regulated levels. Altering the expression of genes that govern development leads mostly to aberrations. However, alterations can also be beneficial, leading to the formation of new phenotypes, which contributes to the astounding diversity of animal forms. In the past the expression of developmental genes has been studied mostly in fixed tissues, which is unable to visualize these highly dynamic processes. We combine genomic fosmid transgenes, expressing genes of interest close to endogenous conditions, with Selective Plane Illumination Microscopy (SPIM) to image the expression of genes live with high temporal resolution and at single cell level in the entire embryo. In an effort to expand the toolkit for studying Drosophila development we have characterized the global expression patterns of various developmentally important genes in the whole embryo. To process the large datasets generated by SPIM, we have developed an automated workflow for processing on a High Performance Computing (HPC) cluster. In a parallel project, we wanted to understand how spatiotemporally regulated gene expression patterns and levels lead to different morphologies across Drosophila species. To this end we have compared by SPIM the expression of transcription factors (TFs) encoded by Drosophila melanogaster fosmids to their orthologous Drosophila pseudoobscura counterparts by expressing both fosmids in D. melanogaster. Here, we present an analysis of divergence of expression of orthologous genes compared A) directly by expressing the fosmids, tagged with different fluorophore, in the same D. melanogaster embryo or B) indirectly by expressing the fosmids, tagged with the same fluorophore, in separate D. melanogaster embryos. Our workflow provides powerful methodology for the study of gene expression patterns and levels during development, such knowledge is a basis for understanding both their evolutionary relevance and developmental function.

<div>As high performance computing systems are advancing from petascale to exascale, scientific workflows to integrate simulation and visualization/analysis are a key factor to influence scientific campaigns. As one of the campaigns to study fluid behaviors, computational fluid dynamics (CFD) simulations have progressed rapidly in the past several decades, and revolutionized our lives in many fields. Lattice Boltzmann method (LBM) is an evolving CFD approach to significantly reducing the complexity of the conventional CFD methods, and can simulate complex fluid flow phenomena with cheaper computational cost. This research focuses on accelerating the workflow of LBM simulation and data analysis.</div><div><br></div><div>I start my research on how to effectively integrate each component of a workflow at extreme scales. Firstly, we design an in-situ workflow benchmark that integrates seven state-of-the-art in-situ workflow systems with three synthetic applications, two real-world CFD applications, and corresponding data analysis. Then detailed performance analysis using visualized tracing shows that even the fastest existing workflow system still has 42% overhead. Then, I develop a novel minimized end-to-end workflow system, Zipper, which combines the fine-grain task parallelism of full asynchrony and pipelining. Meanwhile, I design a novel concurrent data transfer optimization method, which employs a multi-threaded work-stealing algorithm to transfer data using both channels of network and parallel file system. It significantly reduces the data transfer time by up to 32%, especially when the simulation application is stalled. Then investigation on the speedup using OmniPath network tools shows that the network congestion has been alleviated by up to 80%. At last, the scalability of the Zipper system has been verified by a performance model and various largescale workflow experiments on two HPC systems using up to 13,056 cores. Zipper is the fastest workflow system and outperforms the second-fastest by up to 2.2 times.</div><div><br></div><div>After minimizing the end-to-end time of the LBM workflow, I began to accelerate the memory-bound LBM algorithms. We first design novel parallel 2D memory-aware LBM algorithms. Then I extend to design 3D memory-aware LBM that combine features of single-copy distribution, single sweep, swap algorithm, prism traversal, and merging multiple temporal time steps. Strong scalability experiments on three HPC systems show that 2D and 3D memory-aware LBM algorithms outperform the existing fastest LBM by up to 4 times and 1.9 times, respectively. The speedup reasons are illustrated by theoretical algorithm analysis. Experimental roofline charts on modern CPU architectures show that memory-aware LBM algorithms can improve the arithmetic intensity (AI) of the fastest existing LBM by up to 4.6 times.</div>

[ES] En la actualidad, cada vez existen más y más tareas que necesitamos exportar y automatizar en dispositivos portables de bajo consumo que se alimentan de baterías, en los cuales es imprescindible realizar un uso "optimo" de la energía disponible con la finalidad de no drenarlas rápidamente.En la sección primera de esta tesis, "Filtros de señales de audio digital", "optimizamos" las implementaciones de diferentes filtros, tanto generales como específicos, para aplicaciones de sonido digital diseñados e implantados en plataformas basadas en las arquitecturas ARM®. Como filtros generales, trabajamos con los filtros FIR, IIR y Parallel IIR, siendo este tipo de filtros implementados a bajo nivel con instrucciones vectoriales NEON®. Finalmente, se implementa un filtro de separación de señales conocido como "Beamforming", el cual plantea después de su estudio, la problemática de realizar una factorización QR de una matriz relativamente grande en tiempo real, lo cual nos lleva a desarrollar diferentes técnicas de "aceleración" de los cálculos de la misma. En la segunda parte, "Rellenado de mapa de profundidad de una escena", describimos el proceso de rellenado de un mapa de profundidad de una escena capturada a partir del uso de la imagen RGB y de un mapa de profundidad disperso donde únicamente tenemos valores de profundidad en los bordes de los objetos que componen la escena. Estos algoritmos de "rellenado" del mapa de profundidad, también han sido diseñados e implantados en dispositivos basados en la arquitectura ARM®. / [CA] Actualment, cada vegada existixen més i més tasques que tenen la necessitat d'exportar i automatitzar a dispositius portables de baix consum que s'alimenten amb bateríes, als quals es imprescindible realitzar un ús "óptim" de l'energia disponible amb la finalitat de no drenar-les ràpidament. Part I: Filtres de senyals d'àudio digital En aquesta secció "optimitzarem" les implementacions de diferents filtres, tant generals com específics, empreats a aplicacions de so digital disenyats e implantats a plataformes basades a les arquitectures ARM®. Com a filtres generals, treballem amb els filtres FIR, IIR y Parallel IIR, sent aquests tipus de filtres implementats a baix nivell amb instruccions vectorials NEON®. Finalment, s'implementa un filtro de separació de senyals conegut com "Beamforming", el qual planteja després del seu estudi, la problem`atica de realitzar una factorizació QR d'una matriu relativament gran en temps real, i açó ens porta a desenvolupar diferents tècniques "d'acceleració" dels càlculs de la mateixa. Part II: Emplenat del mapa de profunditat d'una escena A la secció d'image per computador, descrivim el procés d'emplenat d'un mapa de profunditat d'una escena capturada fent servir l'image RGB i un mapa de profunditat dispers on únicament tenim valors de profunditat als bordes dels objetes que composen l'escena. Aquests algoritmes "d'emplenat" del mapa de profunditat, també han sigut disenyats e implantats a dispositius basats en l'arquitectura ARM®. / [EN] Currently, there are more and more tasks that we need to export and automate in low-consumption mobile devices that are powered by batteries, in which it is essential to make an "optimum" use of the available energy in order to do not drain them quickly. Part I: Filters of digital audio signals In this section we "optimize" the implementations of different filters, both general and specific, for digital sound applications designed and implemented on platforms based on the ARM®. As general filters, we work with the FIR, IIR and Parallel IIR filters, these types of filters being implemented at a low level with NEON®vector instructions. Finally, a signal separation filter known as "Beamforming" is implemented, which set out after its study, the problem of performing a QR factorization of a relatively large matrix in real time, which leads us to develop different techniques of "acceleration" of the calculations of it. Part II: Filling the depth map of a scene In the computer image section, we describe the process of filling in a depth map of a captured scene using RGB image and a sparse depth map where we only have depth values at the edges of the objects that make up the scene. These depth map "filling" algorithms have also been designed and implemented in devices based on the ARM® architecture. / Alventosa Rueda, FJ. (2022). Algoritmos de imagen y sonido digital con restricciones de tiempo real [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/181573 / TESIS

Air quality assessment plays a crucial role in understanding the impact of air pollution onhuman health and the environment. With the increasing demand for accurate assessment andprediction of air quality, advanced data analytics modelling techniques offer promisingsolutions. This thesis focuses on leveraging advanced data analytics to assess and analyse airpollution concentration levels in Italy over a 4km resolution using the FORAIR_IT datasetsimulated in ENEA on the CRESCO6 infrastructure, aiming to uncover valuable insights andidentifying the most appropriate AI models for predicting air pollution levels. The datacollection, understanding, and pre-processing procedures are discussed, followed by theapplication of big data training and forecasting using Apache Spark MLlib. The research alsoencompasses different phases, including descriptive and inferential analysis to understand theair pollution concentration dataset, hypothesis testing to examine the relationship betweenvarious pollutants, machine learning prediction using several regression models and anensemble machine learning approach and time series analysis on the entire dataset as well asthree major regions in Italy (Northern Italy – Lombardy, Central Italy – Lazio and SouthernItaly – Campania). The computation time for these regression models are also evaluated and acomparative analysis is done on the results obtained. The evaluation process and theexperimental setup involve the usage of the ENEAGRID/CRESCO6 HPC Infrastructure andApache Spark. This research has provided valuable insights into understanding air pollutionpatterns and improving prediction accuracy. The findings of this study have the potential todrive positive change in environmental management and decision-making processes, ultimatelyleading to healthier and more sustainable communities. As we continue to explore the vastpossibilities offered by advanced data analytics, this research serves as a foundation for futureadvancements in air quality assessment in Italy and the models are transferable to other regionsand provinces in Italy, paving the way for a cleaner and greener future.

Cloud-native is a new computing paradigm that takes advantage of key characteristics of cloud computing, where applications are packaged as containers. The lifecycle of containerized applications is typically managed by container orchestration tools such as Kubernetes, the most popular container orchestration system that automates the containers’ deployment, maintenance, and scaling. Kubernetes has become the de facto standard for container orchestrators in the cloud-native era. Meanwhile, with the increasing demand for High-Performance Computing (HPC) over the past years, containerization is being adopted by the HPC community and various processors and special-purpose hardware are utilized to accelerate HPC applications. The architecture of cloud systems has been gradually shifting from homogeneous to heterogeneous with different processors and hardware accelerators, which raises a new challenge: how to exploit different computing resources efficiently? Much effort has been devoted to improving the use efficiency of computing resources in heterogeneous systems from the perspective of task scheduling, which aims to match different types of tasks to optimal computing devices for execution. Existing proposals do not take into account the variation in I/O performance between heterogeneous nodes when scheduling tasks. However, I/O performance is an important but often overlooked factor that can be a potential performance bottleneck for HPC tasks. This thesis proposes an I/O-aware scheduler named cmio-scheduler for containerized data-intensive HPC tasks in Kubernetes-based heterogeneous clusters, which is aware of the I/O throughput of compute nodes when making task placement decisions. In principle, cmio-scheduler assigns data-intensive HPC tasks to the node that fulfills the tasks’ requirements for CPU, memory, and GPU and has the highest I/O throughput. The experimental results demonstrate that cmio-scheduler reduces the execution time by 19.32% for the overall workflow and 15.125% for parallelizable tasks on average. / Cloud-native är ett nytt dataparadigm som drar nytta av de viktigaste egenskaperna hos molntjänster, där applikationer paketeras som behållare. Livscykeln för applikationer i containrar hanteras vanligtvis av verktyg för containerorkestrering, t.ex. Kubernetes, det mest populära systemet för containerorkestrering, som automatiserar installation, underhåll och skalning av containrar. Kubernetes har blivit de facto-standard för containerorkestrar i den molnnativa eran. Med den ökande efterfrågan på högpresterande beräkningar (HPC) under de senaste åren har containerisering antagits av HPC-samhället och olika processorer och specialhårdvara används för att påskynda HPC-tillämpningar. Arkitekturen för molnsystem har gradvis skiftat från homogen till heterogen med olika processorer och hårdvaruacceleratorer, vilket ger upphov till en ny utmaning: hur kan man utnyttja olika datorresurser på ett effektivt sätt? Mycket arbete har ägnats åt att förbättra utnyttjandet av datorresurser i heterogena system ur perspektivet för uppgiftsfördelning, som syftar till att matcha olika typer av uppgifter till optimala datorutrustning för utförande. Befintliga förslag tar inte hänsyn till variationen i I/O-prestanda mellan heterogena noder vid schemaläggning av uppgifter. I/O-prestanda är dock en viktig men ofta förbisedd faktor som kan vara en potentiell flaskhals för HPC-uppgifter. I den här avhandlingen föreslås en I/O-medveten schemaläggare vid namn cmio-scheduler för containeriserade dataintensiva HPC-uppdrag i Kubernetes-baserade heterogena kluster, som är medveten om beräkningsnodernas I/O-genomströmning när den fattar beslut om placering av uppdrag. I princip tilldelar cmio-scheduler dataintensiva HPC-uppgifter till den nod som uppfyller uppgifternas krav på CPU, minne och GPU och som har den högsta I/O-genomströmningen. De experimentella resultaten visar att cmio-scheduler i genomsnitt minskar exekveringstiden med 19,32 % för det totala arbetsflödet och med 15,125 % för parallelliserbara uppgifter.

The race to computing power increases every day in the simulation community. A few years ago, scientists have started to harness the computing power of Graphics Processing Units (GPUs) to parallelize their simulations. As with any parallel architecture, not only the simulation model implementation has to be ported to the new parallel platform, but all the tools must be reimplemented as well. In the particular case of stochastic simulations, one of the major element of the implementation is the pseudorandom numbers source. Employing pseudorandom numbers in parallel applications is not a straightforward task, and it has to be done with caution in order not to introduce biases in the results of the simulation. This problematic has been studied since parallel architectures are available and is called pseudorandom stream distribution. While the literature is full of solutions to handle pseudorandom stream distribution on CPU-based parallel platforms, the young GPU programming community cannot display the same experience yet. In this thesis, we study how to correctly distribute pseudorandom streams on GPU. From the existing solutions, we identified a need for good software engineering solutions, coupled to sound theoretical choices in the implementation. We propose a set of guidelines to follow when a PRNG has to be ported to GPU, and put these advice into practice in a software library called ShoveRand. This library is used in a stochastic Polymer Folding model that we have implemented in C++/CUDA. Pseudorandom streams distribution on manycore architectures is also one of our concerns. It resulted in a contribution named TaskLocalRandom, which targets parallel Java applications using pseudorandom numbers and task frameworks. Eventually, we share a reflection on the methods to choose the right parallel platform for a given application. In this way, we propose to automatically build prototypes of the parallel application running on a wide set of architectures. This approach relies on existing software engineering tools from the Java and Scala community, most of them generating OpenCL source code from a high-level abstraction layer.

A key challenge facing the nuclear fusion community is how to design a reactor that will operate in environmental conditions not easily reproducible in the laboratory for materials testing. Finite element analysis (FEA), commonly used to predict components’ performance, typically uses idealised geometries. An emerging technique shown to have improved accuracy is image based finite element modelling (IBFEM). This involves converting a three dimensional image (such as from X ray tomography) into an FEA mesh. A main advantage of IBFEM is that models include micro structural and non idealised manufacturing features. The aim of this work was to investigate the thermal performance of a CFC Cu divertor monoblock, a carbon fibre composite (CFC) tile joined through its centre to a CuCrZr pipe with a Cu interlayer. As a plasma facing component located where thermal flux in the reactor is at its highest, one of its primary functions is to extract heat by active cooling. Therefore, characterisation of its thermal performance is vital. Investigation of the thermal performance of CFC Cu joining methods by laser flash analysis and X ray tomography showed a strong correlation between micro structures at the material interface and a reduction in thermal conductivity. Therefore, this problem leant itself well to be investigated further by IBFEM. However, because these high resolution models require such large numbers of elements, commercial FEA software could not be used. This served as motivation to develop parallel software capable of performing the necessary transient thermal simulations. The resultant code was shown to scale well with increasing problem sizes and a simulation with 137 million elements was successfully completed using 4096 cores. In comparison with a low resolution IBFEM and traditional FEA simulations it was demonstrated to provide additional accuracy. IBFEM was used to simulate a divertor monoblock mock up, where it was found that a region of delamination existed on the CFC Cu interface. Predictions showed that if this was aligned unfavourably it would increase thermal gradients across the component thus reducing lifespan. As this was a feature introduced in manufacturing it would not have been accounted for without IBFEM.The technique developed in this work has broad engineering applications. It could be used similarly to accurately model components in conditions unfeasible to produce in the laboratory, to assist in research and development of component manufacturing or to verify commercial components against manufacturers’ claims.