Doctoral School

Reading date: 27/09/2024

• CIRAUQUI GARCÍA, DAVID: Optimization with spin glass models
  
  Author: CIRAUQUI GARCÍA, DAVID
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN PHOTONICS
  Department: Institute of Photonic Sciences (ICFO)
  Mode: Normal
  Deposit date: 20/06/2024
  Reading date: 27/09/2024
  Reading time: 10:30
  Reading place: ICFO, Mediterranean Technology Park, Avinguda Carl Friedrich Gauss, 3, 08860 Castelldefels, Barcelona
  Thesis director: LEWENSTEIN, MACIEJ | MARTÍNEZ SAAVEDRA, JOSÉ RAMÓN | RYSZARD GRZYBOWSKI, PRZEMYSLAW
  Committee:
       PRESIDENT: MAZZANTI CASTRILLEJO, FERNANDO PABLO
       SECRETARI: PRUNERI, VALERIO
       VOCAL: DELLANTONIO, LUCA
  Thesis abstract: With applicability on almost every aspect of our lives, optimization problems are ubiquitous to a broad range of fields within both scientific research and industrial environments. As such, these are growing in size and complexity at a fast pace, and are only expected to continue to do so. Accordingly, the urgency for better methods that can yield more optimal solutions in shorter times is increasing and, while the development of quantum computing technologies that are capable of tackling these problems evolves steadily, it does so too slowly for the challenges that nowadays society's demands represent. Consequently, a lot of effort is being invested to further develop classical methods and machines that are specially designed to solve optimization problems of relevant enough sizes. The present thesis is framed within this paradigm: classical optimization techniques are studied from various different perspectives, with the goal of improving their efficiency. To this end, we first dive into basic concerns related to the physical properties of the systems that allow for the convenient formulation of industrially-relevant optimization problems, namely spin glasses with quenched disorders. The understanding of such properties is of utmost importance for the correct designing of the annealing schedules used by thermally-based optimization methods. We then study the impact that the hidden correlations of the pseudo random number streams used in their simulations have in the results by comparing simulations using PRNGs of various qualities and perfectly random QRNGs. To conclude, we investigate novel ways, inspired by quantum-mechanical systems, to efficiently navigate the energy landscapes of spin glasses in classical algorithms, which has the potential of preventing the simulations getting stuck into local energy minima and thus reaching more optimal solutions.
  

• JAMET, ALEXANDRE VALENTIN: Interaction between computer architecture and artificial intelligence
  
  Author: JAMET, ALEXANDRE VALENTIN
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN COMPUTER ARCHITECTURE
  Department: Department of Computer Architecture (DAC)
  Mode: Normal
  Deposit date: 19/07/2024
  Reading date: 27/09/2024
  Reading time: 09:30
  Reading place: Defensa pública a la Sala E101 - Edifici C6 (FIB) - Campus Nord - Barcelona
  Thesis director: ÁLVAREZ MARTÍ, LLUC | CASAS GUIX, MARC
  Committee:
       PRESIDENT: ROS BARDISA, ALBERTO
       SECRETARI: CANAL CORRETGER, RAMON
       VOCAL: PERAIS, ARTHUR
  Thesis abstract: Since its inception with the first computing systems, computer architecture has lived through many revolutions and saw plenty of technological innovations. However, a significant challenge has persisted throughout the evolution of computing systems: the Memory Wall. To address this challenge, architects have devised various latency tolerance techniques, including cache hierarchy, cache replacement policies, hardware prefetching, and off-chip prediction. Cache hierarchy involves the use of intermediate memories, such as caches, to store frequently accessed data close to the processor, thereby reducing memory access latencies. Cache replacement policies determine which data blocks should be stored or evicted from caches based on predictions of future reuse. Hardware prefetching mechanisms aim to bring data blocks that are likely to be needed in the near future into the cache proactively. Off-chip prediction predicts whether a load demand request will benefit from cache access or if it will require a DRAM access, allowing for speculative fetching of data blocks from DRAM to hide memory access latencies. This thesis addresses the challenges of cache management and memory access optimization in modern computer architectures, focusing on improving performance and energy efficiency across a variety of workloads. It presents three main contributions. The first contribution critically assesses the effectiveness of contemporary Last Level Cache (LLC) replacement policies across a diverse spectrum of workloads, encompassing graph processing, scientific, industrial applications, as well as standard benchmark suites like SPEC CPU 2006 and SPEC CPU 2017. Despite exhibiting notable performance enhancements in conventional benchmark scenarios, these existing LLC replacement policies often falter in capturing the nuanced access patterns characteristic of modern High-Performance Computing (HPC) and big data workloads. In response to this challenge, two novel LLC replacement policies, namely Multi-Sampler Multiperspective (MS-MPPPB) and Multiperspective with Dynamic Features Selector (DS-MPPPB), are introduced and rigorously evaluated. Demonstrating superior efficacy across a broad array of workloads, these innovative policies offer heightened performance benefits tailored specifically for HPC and big data applications. The second contribution is dedicated to enhancing memory access patterns, specifically for graph-processing workloads. These workloads are renowned for their irregular memory access patterns and suboptimal data locality. This contribution targets the first level of the cache hierarchy, as careful analysis reveals that, when considering graph-processing workloads, the vast majority of L1D misses eventually require a DRAM access. Introducing the innovative Large Predictor (LP), this endeavor aims to discern between regular and irregular memory accesses, channeling irregular accesses efficiently through a dedicated Side Data Cache (SDC). By synergizing LP with SDC, notable performance enhancements are achieved, surpassing conventional cache hierarchies and state-of-the-art cache replacement policies, particularly within the realm of graph-processing applications. The third contribution presents the Two Level Perceptron (TLP) predictor, a sophisticated approach that integrates off-chip prediction with adaptive prefetch filtering within the first-level data cache (L1D). Leveraging a dual-layered structure composed of the First Level Predictor (FLP) and Second Level Predictor (SLP), TLP effectively mitigates average DRAM transactions while enhancing overall performance across both single-core and multi-core workloads. Collectively, these contributions advance the state-of-the-art in cache management and memory access optimization, providing insights and techniques to enhance the performance and energy efficiency of modern computer architectures across a variety of workloads.
  

• MORAGA GONZÁLEZ, GRECO ALONSO: Contribution to the structural modal behaviour of reaction hydraulic turbines
  
  Author: MORAGA GONZÁLEZ, GRECO ALONSO
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN MECHANICAL, FLUIDS AND AEROSPACE ENGINEERING
  Department: Department of Fluid Mechanics (MF)
  Mode: Article-based thesis
  Deposit date: 11/07/2024
  Reading date: 27/09/2024
  Reading time: 11:00
  Reading place: Aula L-LS4 de la ETSEIB
  Thesis director: PRESAS BATLLÓ, ALEXANDRE
  Committee:
       PRESIDENT: HOUDE, SEBASTIEN
       SECRETARI: VALENTIN RUIZ, DAVID
       VOCAL: ALVAREZ ALVAREZ, EDUARDO
  Thesis abstract: Current environmental challenges have driven the importance of renewable energies worldwide. Consequently, solar and wind generation plants, in particular, have become increasingly popular and integrated into the electrical grid. Although these plants have increased the worldwide generation capacity, their intermittent nature has generated the need for power regulation mechanisms that allow maintaining a constant energy supply to the grid. In this line, hydroelectric plants have played a crucial role not only in the generation of renewable energies but also in the regulation of the total energy supplied to the electrical grid. Thanks to their quick response and their ability to work in an extended operating range, hydraulic turbines have been used to generate and store available energy (as in the case of pump turbines) based on energy demand.To regulate the energy needed by the grid, units operate under off-design conditions and experience frequent start and stop cycles. With this, machines undergo intense pressure pulsations due to hydraulic instabilities, transient conditions and intense periodic excitations such as rotor-stator interactions. This can lead to potential structural problems and resonances in critical turbine structures, such as the runner, which can result in mechanical failures caused by structural fatigue. To avoid these problems, it is of paramount importance to have an accurate knowledge of the modal behaviour of these structures. Nevertheless, this implies to analyze very complex phenomena such as the influence of heavy and viscous fluids, boundary conditions and structural characteristics on the modal parameters of the structure, and overcoming all of the challenges that can arise during the analysis. In this thesis, the structural modal behavior of reaction hydraulic turbines is studied through the analysis of modal parameters of simple structures and real prototype runners. First, the influence of vibration amplitude and confinement on the damping of a runner in resonance is examined. This study includes an experimental and theoretical analysis of the influence of vibration amplitude on the fluid added damping in a submerged disc vibrating in resonance. The effect of a nearby rigid wall on the correlation between the vibration amplitude and the fluid damping is also analyzed. Analytical explanation and formulation for the complex energy dissipation mechanism are given.Secondly, the mode shapes of bladed runners similar to axial flow machines, including Kaplan turbines, are characterized. This study includes a numerical and experimental analysis of the mode shapes of simple disks and bladed disks. Later, the relationship between the characteristics of the global mode shapes and the number of blades of the structure is examined. Finally, the numerical results of the bladed disks are contrasted with experimental results of a real runner. Both the global and local mode shapes are discussed in this section.Lastly, a real case of a real prototype turbine is analyzed. In this case, strong vibrations and noise were detected in the machine during its commissioning. Through this study, the natural frequencies and mode shapes of the machine are analyzed experimentally and numerically. To determine the origin of the vibrations, a transmissibility analysis is also carried out. Finally, operation tests are conducted to identify the point at which the levels of vibration and noise start to increase. The effectiveness of the proposed measurement strategy, along with the importance of the measurement techniques used, such as the transmissibility test, is demonstrated through a real case. The cause of the problem is found and a solution is proposed and implemented.This thesis is based on three papers published during its development period, in scientific journals indexed in first and second quartiles of the JCR. The details of these articles are explained throughout the thesis and are attached at the end of it.
  

• ORTEGA DONOSO, SARA ISABEL: Luz y aprendizaje dentro de los espacios educativos. Aproximaciones desde el color.
  
  Author: ORTEGA DONOSO, SARA ISABEL
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN ARCHITECTURAL, BUILDING CONSTRUCTION AND URBANISM TECHNOLOGY
  Department: Department of Architectural Technology (TA)
  Mode: Normal
  Deposit date: 01/07/2024
  Reading date: 27/09/2024
  Reading time: 11:00
  Reading place: Sala de Graus ETSAB. Planta Baixa (E.T.S. Arquitectura de Barcelona)Presencialment
  Thesis director: MUROS ALCOJOR, ADRIAN | BAUTISTA PEREZ, GUILLERMO
  Committee:
       PRESIDENT: ESCOFET ROIG, ANNA MARIA
       SECRETARI: DAUMAL DOMENECH, FRANCESC DE PAULA
       VOCAL: HIGUERA TRUJILLO, JUAN LUIS
  Thesis abstract: We live in a visually demanding environment. Most of our cognitive stimuli come from visual perception. In this context of widespread dependence on the sense of sight, studying the link between artificial light and cognitive tasks is key to interacting with our memory and attention, which are increasingly accustomed to dynamic environments.The objective of this research is to analyze the disciplines that can most influence learning: cognitive sciences, architecture, and education, to:First, analyze how some light parameters can affect learning, paying special attention to the impact of light on human perception and attention and how these affect memory.Next, analyze innovative teaching methodologies and how communication and learning technologies have been incorporated into the classroom.Finally, examine the evolution of educational spaces through case studies of schools worldwide over recent decades, delving into the ways lighting and control technologies have been integrated according to the needs of educational methods within the classroom.From the analysis of the implications of light on attention and memory, we will derive the possibilities that working with different light parameters in the classroom offers us; from the analysis of teaching methodologies, we will obtain the real needs of the current classroom; and from the case study analysis, we will gain the capacity for innovation in the classroom with new lighting strategies.The intersection of these three disciplines results in three alternative uses of lighting in educational spaces, which are subjected to an experimental analysis comparing performance on a task in the proposed scenes with performance under traditional lighting conditions to conclude new classroom solutions and future lines of research.A broad narrative review methodology has been used for the case studies, providing an overview of the problem of adapting lighting to usage and users within schools. To this end, sources published in architectural journals and significant examples with educational spaces that develop innovative teaching methodologies have been reviewed, and criteria for analyzing natural and artificial lighting considered throughout these cases have been established.An experimental methodology has been used to analyze the impact of three alternative lighting proposals based on situations not previously considered. In all three cases, a control group and a study group from two public schools in the province of Barcelona are considered. These proposals arise from the observation of a lack of consideration of an important light parameter in the transmission of information in a visual environment: color.Finally, the results reflect the need to adapt and make classrooms more flexible according to the different learning needs within the classroom and the possibility of doing so with proposals not previously considered regarding color and dynamism parameters.
  

Reading date: 01/10/2024

• ALI, OMAR SHAABAN IBRAHIM: Optimization techniques for distributed task-based programming models
  
  Author: ALI, OMAR SHAABAN IBRAHIM
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN COMPUTER ARCHITECTURE
  Department: Department of Computer Architecture (DAC)
  Mode: Normal
  Deposit date: 18/07/2024
  Reading date: 01/10/2024
  Reading time: 14:00
  Reading place: Defensa pública a la Sala E101 - Edifici C6 (FIB) - Campus Nord - Barcelona
  Thesis director: CARPENTER, PAUL MATTHEW
  Committee:
       PRESIDENT: BULL, JONATHAN MARK
       SECRETARI: JIMENEZ GONZALEZ, DANIEL
       VOCAL: ARAUJO, GUIDO
  Thesis abstract: In HPC, task-based models have gained prominence via the adoption of tasks in OpenMP, as an asynchronous and platform-agnostic high-performance and productive model by annotating existing code, transforming it into a parallel version. A program is expressed as a Directed Acyclic Graph (DAG), whose vertices are units of code called tasks. Edges between tasks represent dependencies between them, and tasks with no logical relationship can be executed concurrently. The task graph is independent of the targeted platform architecture, making these models suitable for concurrent execution on a broad spectrum of platforms such as multi-core SMPs or offloaded to GPUs, FPGAs, or accelerators. Several initiatives explore distributed-tasking approaches, which use the task model to decompose the application across multiple nodes with distributed memory. The advantage is that the application is expressed in a simple and clean manner that reflects only the computations and dependencies. Unfortunately, the distributed tasking approaches suffer from poor efficiency and scalability, which hinder their adoption by the HPC community. This is mainly due to the overhead of task creation and dependency graph construction, which are usually sequential. This thesis proposes two techniques to address this problem. Our first approach relates to task nesting, which mitigates the sequential bottleneck by building the full dependency graph in parallel using multiple concurrently executing parent tasks. A key limitation of task nesting is that a task cannot be created until all its accesses and its descendants’ accesses are known. Current approaches to work around this limitation either halt task creation and execution using an explicit taskwait barrier or substitute dependencies with artificial accesses known as sentinels. We introduce the auto clause, which indicates that the task may create subtasks that access unspecified memory regions, or it may allocate and return memory at addresses that are not yet known. Contrary to taskwait, our approach does not prevent the concurrent creation and execution of tasks, maintaining parallelism and allowing the scheduler to optimize load balance and data locality. In addition, all tasks can be given a precise specification of their own data accesses, unlike sentinels, resulting in a unified mechanism governing task ordering, program data transfers on distributed memory, and optimizing data locality, e.g. on NUMA systems. The auto clause, therefore, provides an incremental path to develop programs with nested tasks by removing the need for every parent task to have a complete specification of the accesses of its descendent tasks while reducing redundant information that can be time-consuming and error-prone to describe. Our second approach takes advantage of the iterative behaviour of many HPC applications, such as those that employ iterative methods or multi-step simulations. Most models construct the full unrolled task graph sequentially despite the fact that these applications create the same directed acyclic graph of tasks on each timestep. We define the programming model based on the taskiter clause, a recently introduced construct in the literature for iterative applications on SMP. We also describe the full runtime implementation to exploit this information to eliminate the sequential bottleneck and control messages while retaining the simplicity and productivity of the existing approach. We integrate both techniques into OmpSs-2@Cluster, the distributed tasking variant of OmpSs-2, and evaluate the performance on the MareNostrum 4 supercomputer.