This dissertation proves the feasibility of accurate runtime prediction of processor performance under frequency scaling. The performance predictors developed in this dissertation allow processors capable of dynamic voltage and frequency scaling (DVFS) to improve their performance or energy efficiency by dynamically adapting chip or core voltages and frequencies to workload characteristics. The dissertation considers three processor configurations: the uniprocessor capable of chip-level DVFS, the private cache chip multiprocessor capable of per-core DVFS, and the shared cache chip multiprocessor capable of per-core DVFS. Depending on processor configuration, the presented performance predictors help the processor realize 72–85% of average oracle performance or energy efficiency gains. / text

A tomada de decisão em gerência de pavimentos depende, dentre outros fatores, da estimativa da evolução da condição do pavimento ao longo do tempo. Tal estimativa é obtida por uma função que relaciona as causas e os efeitos da deterioração dos pavimentos, denominada modelo de desempenho. Neste trabalho são desenvolvidos modelos estatísticos para previsão do desempenho de pavimentos, mediante o estabelecimento de equações de regressão tendo por base dados históricos de avaliações da condição da malha rodoviária do Estado do Paraná. A análise do comportamento dos pavimentos é efetuada utilizando-se uma programação fatorial que, através de análise de variância (ANOVA), permite a determinação do nível de significância de fatores pré-selecionados (variáveis independentes: tráfego, idade e estrutura do pavimento) e de suas interações, bem como a modelagem do desempenho dos pavimentos (variáveis dependentes: irregularidade longitudinal e condição estrutural). Para cada uma das células da matriz fatorial, que correspondem às combinações dos fatores considerados, também são desenvolvidos modelos probabilísticos para previsão do desempenho de pavimentos, a partir de avaliações realizadas por especialista (engenheiros do DER-PR) e mediante o estabelecimento de matrizes de probabilidade de transição de Markov. Este trabalho mostra que é possível o desenvolvimento de modelos de desempenho sem dados históricos de avaliação da condição dos pavimentos ou tendo-se apenas dados coletados por um curto período de tempo. Observa-se, também, boa concordância entre os modelos estatísticos e probabilísticos, particularmente para previsão do desempenho funcional dos pavimentos. Os modelos de desempenhos desenvolvidos neste trabalho, quando comparados com equações desenvolvidas por pesquisadores e órgãos rodoviários brasileiros e estrangeiros, apresentaram melhores resultados, evidenciando as limitações de modelos de desempenho desenvolvidos e calibrados sob condições específicas. / The decision-making in pavement management systems depends, among other factors, of the prediction of the pavement condition during the service life. This prediction is obtained through a relation between causes and effects of pavement deterioration, called performance prediction model. This work develops statistic models for the predicion of pavement performance, based on regression equations from data of pavement evaluation performed in the highway network of the State of Paraná-Brazil. The pavement behavior is evaluated from an Analysis of Variance (ANOVA) of a factorial array, which calculates the level of significance of preselected factors (independent variables: traffic, age, and pavement structure) and their interactions and gives the performance models(dependent variables: roughness and structural condition). For each cell of the factorial array, that corresponds to combinations of the considered factors, it is also developed probabilistic models for the prediction of pavement performance, based on evaluations of pavement condition performed by specialists (State of Parana DOT engineers) and the definition of Markov transition matrices. This work shows that it is possible to develop performance prediction models without historic data of pavement evaluation or having just data colected in a short period of time. It is observed good correspondence between both models, statistic and probabilistic, particularly for the prediction of thefunctional behavior. The performance prediction models developed in this work show better results than equations developed by Brazilian and foreign researches and highway agencies, in a clear evidence of the limitation of models developed and calibrated under specific conditions.

The knowledge available to be exploited by text classification and information retrieval systems has significantly changed, both in nature and quantity, in the last years. Nowadays, there are several sources of information that can potentially improve the classification process, and systems should be able to adapt to incorporate multiple sources of available data in different formats. This fact is specially important in environments where the required information changes rapidly, and its utility may be contingent on timely implementation. For these reasons, the importance of adaptability and flexibility in information systems is rapidly growing. Current systems are usually developed for specific scenarios. As a result, significant engineering effort is needed to adapt them when new knowledge appears or there are changes in the information needs. This research investigates the usage of knowledge within text classification from two different perspectives. On one hand, the application of descriptive approaches for the seamless modelling of text classification, focusing on knowledge integration and complex data representation. The main goal is to achieve a scalable and efficient approach for rapid prototyping for Text Classification that can incorporate different sources and types of knowledge, and to minimise the gap between the mathematical definition and the modelling of a solution. On the other hand, the improvement of different steps of the classification process where knowledge exploitation has traditionally not been applied. In particular, this thesis introduces two classification sub-tasks, namely Semi-Automatic Text Classification (SATC) and Document Performance Prediction (DPP), and several methods to address them. SATC focuses on selecting the documents that are more likely to be wrongly assigned by the system to be manually classified, while automatically labelling the rest. Document performance prediction estimates the classification quality that will be achieved for a document, given a classifier. In addition, we also propose a family of evaluation metrics to measure degrees of misclassification, and an adaptive variation of k-NN.

Dans le domaine du calcul haute performance, de nombreux programmes étalons ou benchmarks sont utilisés pour mesurer l’efficacité des calculateurs,des compilateurs et des optimisations de performance. Les benchmarks de référence regroupent souvent des programmes de calcul issus de l’industrie et peuvent être très longs. Le processus d’´étalonnage d’une nouvelle architecture de calcul ou d’une optimisation est donc coûteux.La plupart des benchmarks sont constitués d’un ensemble de noyaux de calcul indépendants. Souvent l’´étalonneur n’est intéressé que par un sous ensemble de ces noyaux, il serait donc intéressant de pouvoir les exécuter séparément. Ainsi, il devient plus facile et rapide d’appliquer des optimisations locales sur les benchmarks. De plus, les benchmarks contiennent de nombreux noyaux de calcul redondants. Certaines opérations, bien que mesurées plusieurs fois, n’apportent pas d’informations supplémentaires sur le système à étudier. En détectant les similarités entre eux et en éliminant les noyaux redondants, on diminue le coût des benchmarks sans perte d’information.Cette thèse propose une méthode permettant de décomposer automatiquement une application en un ensemble de noyaux de performance, que nous appelons codelets. La méthode proposée permet de rejouer les codelets,de manière isolée, dans différentes conditions expérimentales pour pouvoir étalonner leur performance. Cette thèse étudie dans quelle mesure la décomposition en noyaux permet de diminuer le coût du processus de benchmarking et d’optimisation. Elle évalue aussi l’avantage d’optimisations locales par rapport à une approche globale.De nombreux travaux ont été réalisés afin d’améliorer le processus de benchmarking. Dans ce domaine, on remarquera l’utilisation de techniques d’apprentissage machine ou d’´echantillonnage. L’idée de décomposer les benchmarks en morceaux indépendants n’est pas nouvelle. Ce concept a été aappliqué avec succès sur les programmes séquentiels et nous le portons à maturité sur les programmes parallèles.Evaluer des nouvelles micro-architectures ou la scalabilité est 25× fois plus rapide avec des codelets que avec des exécutions d’applications. Les codelets prédisent le temps d’exécution avec une précision de 94% et permettent de trouver des optimisations locales jusqu’`a 1.06× fois plus efficaces que la meilleure approche globale. / In high performance computing, benchmarks evaluate architectures, compilers and optimizations. Standard benchmarks are mostly issued from the industrial world and may have a very long execution time. So, evaluating a new architecture or an optimization is costly. Most of the benchmarks are composed of independent kernels. Usually, users are only interested by a small subset of these kernels. To get faster and easier local optimizations, we should find ways to extract kernels as standalone executables. Also, benchmarks have redundant computational kernels. Some calculations do not bring new informations about the system that we want to study, despite that we measure them many times. By detecting similar operations and removing redundant kernels, we can reduce the benchmarking cost without loosing information about the system. This thesis proposes a method to automatically decompose applications into small kernels called codelets. Each codelet is a standalone executable that can be replayed in different execution contexts to evaluate them. This thesis quantifies how much the decomposition method accelerates optimization and benchmarking processes. It also quantify the benefits of local optimizations over global optimizations. There are many related works which aim to enhance the benchmarking process. In particular, we note machine learning approaches and sampling techniques. Decomposing applications into independent pieces is not a new idea. It has been successfully applied on sequential codes. In this thesis we extend it to parallel programs. Evaluating scalability or new micro-architectures is 25× faster with codelets than with full application executions. Codelets predict the execution time with an accuracy of 94% and find local optimizations that outperform the best global optimization up to 1.06×.

A tomada de decisão em gerência de pavimentos depende, dentre outros fatores, da estimativa da evolução da condição do pavimento ao longo do tempo. Tal estimativa é obtida por uma função que relaciona as causas e os efeitos da deterioração dos pavimentos, denominada modelo de desempenho. Neste trabalho são desenvolvidos modelos estatísticos para previsão do desempenho de pavimentos, mediante o estabelecimento de equações de regressão tendo por base dados históricos de avaliações da condição da malha rodoviária do Estado do Paraná. A análise do comportamento dos pavimentos é efetuada utilizando-se uma programação fatorial que, através de análise de variância (ANOVA), permite a determinação do nível de significância de fatores pré-selecionados (variáveis independentes: tráfego, idade e estrutura do pavimento) e de suas interações, bem como a modelagem do desempenho dos pavimentos (variáveis dependentes: irregularidade longitudinal e condição estrutural). Para cada uma das células da matriz fatorial, que correspondem às combinações dos fatores considerados, também são desenvolvidos modelos probabilísticos para previsão do desempenho de pavimentos, a partir de avaliações realizadas por especialista (engenheiros do DER-PR) e mediante o estabelecimento de matrizes de probabilidade de transição de Markov. Este trabalho mostra que é possível o desenvolvimento de modelos de desempenho sem dados históricos de avaliação da condição dos pavimentos ou tendo-se apenas dados coletados por um curto período de tempo. Observa-se, também, boa concordância entre os modelos estatísticos e probabilísticos, particularmente para previsão do desempenho funcional dos pavimentos. Os modelos de desempenhos desenvolvidos neste trabalho, quando comparados com equações desenvolvidas por pesquisadores e órgãos rodoviários brasileiros e estrangeiros, apresentaram melhores resultados, evidenciando as limitações de modelos de desempenho desenvolvidos e calibrados sob condições específicas. / The decision-making in pavement management systems depends, among other factors, of the prediction of the pavement condition during the service life. This prediction is obtained through a relation between causes and effects of pavement deterioration, called performance prediction model. This work develops statistic models for the predicion of pavement performance, based on regression equations from data of pavement evaluation performed in the highway network of the State of Paraná-Brazil. The pavement behavior is evaluated from an Analysis of Variance (ANOVA) of a factorial array, which calculates the level of significance of preselected factors (independent variables: traffic, age, and pavement structure) and their interactions and gives the performance models(dependent variables: roughness and structural condition). For each cell of the factorial array, that corresponds to combinations of the considered factors, it is also developed probabilistic models for the prediction of pavement performance, based on evaluations of pavement condition performed by specialists (State of Parana DOT engineers) and the definition of Markov transition matrices. This work shows that it is possible to develop performance prediction models without historic data of pavement evaluation or having just data colected in a short period of time. It is observed good correspondence between both models, statistic and probabilistic, particularly for the prediction of thefunctional behavior. The performance prediction models developed in this work show better results than equations developed by Brazilian and foreign researches and highway agencies, in a clear evidence of the limitation of models developed and calibrated under specific conditions.

abstract: Teams are increasingly indispensable to achievements in any organizations. Despite the organizations' substantial dependency on teams, fundamental knowledge about the conduct of team-enabled operations is lacking, especially at the {\it social, cognitive} and {\it information} level in relation to team performance and network dynamics. The goal of this dissertation is to create new instruments to {\it predict}, {\it optimize} and {\it explain} teams' performance in the context of composite networks (i.e., social-cognitive-information networks). Understanding the dynamic mechanisms that drive the success of high-performing teams can provide the key insights into building the best teams and hence lift the productivity and profitability of the organizations. For this purpose, novel predictive models to forecast the long-term performance of teams ({\it point prediction}) as well as the pathway to impact ({\it trajectory prediction}) have been developed. A joint predictive model by exploring the relationship between team level and individual level performances has also been proposed. For an existing team, it is often desirable to optimize its performance through expanding the team by bringing a new team member with certain expertise, or finding a new candidate to replace an existing under-performing member. I have developed graph kernel based performance optimization algorithms by considering both the structural matching and skill matching to solve the above enhancement scenarios. I have also worked towards real time team optimization by leveraging reinforcement learning techniques. With the increased complexity of the machine learning models for predicting and optimizing teams, it is critical to acquire a deeper understanding of model behavior. For this purpose, I have investigated {\em explainable prediction} -- to provide explanation behind a performance prediction and {\em explainable optimization} -- to give reasons why the model recommendations are good candidates for certain enhancement scenarios. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2018

Efficient execution of scientific applications requires an understanding of how system features impact the performance of the application. Performance models provide significant insight into the performance relationships between an application and the system used for execution. In particular, models can be used to predict the relative performance of different systems used to execute an application. Recently, a significant effort has been devoted to gaining a more detailed understanding of the performance characteristics of a fusion reaction application, GYRO. GYRO is a plasma-physics application used to gain a better understanding of the interaction of ions and electrons in fusion reactions. In this thesis, we use the well-known Prophesy system to analyze and model the performance of GYRO across various supercomputer platforms. Using processor partitioning, we determine that utilizing the smallest number of processors per node is the most effective processor configuration for executing the application. Further, we explore trends in kernel coupling values across platforms to understand how kernels of GYRO interact. In this work, experiments are conducted on the supercomputers Seaborg and Jacquard at the DOE National Energy Research Scientific Computing Center and the supercomputers DataStar P655 and P690 at the San Diego Supercomputing Center. Across all four platforms, our results show that utilizing one processor per node (ppn) yields better performance than full or half ppn usage. Our experimental results also show that using kernel coupling to model and predict the performance of GYRO is more accurate than summation. On average, kernel coupling provides for prediction estimates that have less than a 7% error. The performance relationship between kernel coupling values and the sharing of information throughout the GYRO application is explored by understanding the global communication within the application and data locality.

Predicting the condition of a degrading dynamic system is critical for implementing successful control and designing the optimal operation and maintenance strategies throughout the lifetime of the system. In many situations, especially in manufacturing, systems experience multiple degradation cycles, failures, and maintenance events throughout their lifetimes. In such cases, historical records of sensor readings observed during the lifecycle of a machine can yield vital information about degradation patterns of the monitored machine, which can be used to formulate dynamic models for predicting its future performance. Besides the ability to predict equipment failures, another major component of cost effective and high-throughput manufacturing is tight control of product quality. Quality control is assured by taking periodic measurements of the products at various stages of production. Nevertheless, quality measurements of the product require time and are often executed on costly measurement equipment, which increases the cost of manufacturing and slows down production. One possible way to remedy this situation is to utilize the inherent link between the manufacturing equipment condition, mirrored in the readings of sensors mounted on that machine, and the quality of products coming out of it. The concept of Virtual Metrology (VM) addresses the quality control problem by using data-driven models that relate the product quality to the equipment sensors, enabling continuous estimation of the quality characteristics of the product, even when physical measurements of product quality are not available. VM can thus bring significant production benefits, including improved process control, reduced quality losses and higher productivity. In this dissertation, new methods are formulated that will combine long-term performance prediction of sensory signatures from a degrading manufacturing machine with VM quality estimation, which enables integration of predictive condition monitoring (prediction of sensory signatures) with predictive manufacturing process control (predictive VM model). The recently developed algorithm for prediction of sensory signatures is capable of predicting the system condition by comparing the similarity of the most recent performance signatures with the known degradation patterns available in the historical records. The method accomplishes the prediction of non-Gaussian and non-stationary time-series of relevant performance signatures with analytical tractability, which enables calculations of predicted signature distributions with significantly greater speeds than what can be found in literature. VM quality estimation is implemented using the recently introduced growing structure multiple model system paradigm (GSMMS), based on the use of local linear dynamic models. The concept of local models enables representation of complex, non-linear dependencies with non-Gaussian and non-stationary noise characteristics, using a locally tractable model representation. Localized modeling enables a VM that can detect situations when the VM model is not adequate and needs to be improved, which is one of the main challenges in VM. Finally, uncertainty propagation with Monte Carlo simulation is pursued in order to propagate the predicted distributions of equipment signatures through the VM model to enable prediction of distributions of the quality variables using the readily available sensor readings streaming from the monitored manufacturing machine. The newly developed methods are applied to long-term production data coming from an industrial plasma-enhanced chemical vapor deposition (PECVD) tool operating in a major semiconductor manufacturing fab. / text

In a country with continental dimensions like Brazil, road infrastructure that provides a displacement with comfort and safety is extremely important, once it is from these roads that the majority of supplies and people moves daily to distant regions. Is through appropriate management of corrective floors and measurements taken at the correct times that these highways provides safety and comfort to users in their displacement. One way to perform this management properly is making use of performance prediction models that makes the manager can predict the appearance of defects and the necessity of performing maintenance, providing financial resources required for repairs. Thus, the objective of this research is to monitor three highway stretches implemented in the region of Santa Maria - RS, checking their functional and structural performance in order to assist in the development of performance prediction models. To achieving the study, tests were performed at predetermined periods: Sand Patch, British Pendulum, Roughness, Rutting analysis, Distress identification and Deflections Basin Survey by Benkelman Beam and Falling Weight Deflectometer (FWD) at Roraima Avenue (restoration stretch), Hélvio Basso Avenue and Quartéis Intersection (new highway stretches). In these tests, quantitative and classifying counts of vehicles were made in order to find the number of equivalent demand of standard axis through the FEC calculation by AASHTO and USACE methodology. With that was obtained in all 34 performance prediction models for Avenues Roraima and Hélvio Basso (17 - AASHTO and 17 - USACE). No models have been developed for Quartéis Intersection once it is at the beginning of monitoring. In addition to performance prediction models, performance evaluations were carried out of three sections monitored the mediated that it were requested by the traffic, and found that the Roraima Avenue had the highest deflection values, IGG and cracked area. The Quartéis Intersection had the highest values of rutting. Back Analysis were performed by BAKFAA software to obtain resilient modulus in all layers of pavements studied and, in general, met values coherent with those studied in the literature. It was also made structural analysis by AEMC/SisPAv (2009) software in order to find the estimated durability in the highway stretches. The Roraima Avenue despite being an old stretch and present the highest amount of early distress, was the one that had the highest prediction of durability between the three analysed sections. Furthermore, structural analysis was performed for Roraima Avenue, by standard DNER PRO 011/1979, once it is a restored pavement, it was found that this stretch is expected five years and ten months of durability. Therefore, monitoring and obtaining performance prediction models are extremely important for the proper management of pavements. / Em um país com dimensões continentais como o Brasil, infraestrutura de rodovias que proporcione um deslocamento com conforto e segurança é extremamente importante, uma vez que é a partir destas rodovias que a grande maioria dos insumos e das pessoas se desloca diariamente para regiões distantes. É por meio da gerência adequada dos pavimentos e medidas corretivas realizadas nos tempos corretos, que estas rodovias fornecem aos usuários segurança e conforto nos seus deslocamentos. Uma forma de realizar esta gerência de maneira adequada é fazendo uso de modelos de previsão de desempenho, que fazem com que o gestor possa prever de forma antecipada o aparecimento de defeitos e a necessidade da realização de manutenções, disponibilizando, assim, recursos financeiros necessários para as obras de intervenções. Com isso, o objetivo desta pesquisa é realizar o monitoramento de três trechos implantados na região de Santa Maria RS, verificando seus desempenhos funcionais e estruturais de modo a subsidiar a construção de modelos de previsão de desempenho. Para a concretização do estudo foi realizado, em períodos pré-determinados, ensaios de Mancha de Areia, Pêndulo Britânico, Irregularidade Longitudinal, Afundamento em Trilha de Roda, Levantamento Visual de Defeitos e Levantamento de Bacia de Deflexões através da Viga Benkelman e Falling Weight Deflectometer (FWD) na Avenida Roraima (trecho de restauração), Avenida Hélvio Basso e no Trevo dos Quartéis (trechos novos). Além destes ensaios, foram feitas contagens quantitativas e classificatórias dos veículos que solicitavam os trechos, com a intenção de encontrar o número de solicitações equivalentes do eixo padrão através dos cálculos de FEC pela metodologia AASHTO e USACE. Com isso, obteve-se no total 34 modelos de previsão de desempenho para as Avenidas Roraima e Hélvio Basso (17 AASHTO e 17 USACE). Não foram desenvolvidos modelos para o Trevo dos Quartéis, uma vez que o mesmo está em início de monitoramento. Além dos modelos de previsão de desempenho, foram realizadas avaliações do desempenho dos três trechos monitorados à mediada que o mesmo era solicitado pelo tráfego, sendo verificado que a Avenida Roraima apresentou os maiores valores de deflexão, IGG e Área Trincada. Já o Trevo dos Quartéis apresentou os maiores valores de ATR. Foram realizadas retroanálises pelo software BAKFAA para obtenção de módulo de resiliência em todas as camadas dos pavimentos estudados e, de forma geral, encontrou-se valores coerentes com os estudados na literatura. Fez-se, também, análises estruturais através do software AEMC/SisPAv (2009), a fim de encontrar a estimativa de durabilidade dos trechos. A Avenida Roraima apesar de ser um trecho antigo e apresentar a maior quantidade de defeitos prematuramente, foi a que teve a maior previsão de durabilidade entre os três trechos. Além disso, realizou-se a análise estrutural da Avenida Roraima pela norma DNER PRO 011/1979, uma vez que se trata de uma restauração de um pavimento e constatou-se que este trecho tem previsão de durabilidade de cinco anos e dez meses. Assim, o monitoramento e obtenção de modelos de previsão de desempenho são de suma importância para a boa gerência dos pavimentos.

abstract: With the massive multithreading execution feature, graphics processing units (GPUs) have been widely deployed to accelerate general-purpose parallel workloads (GPGPUs). However, using GPUs to accelerate computation does not always gain good performance improvement. This is mainly due to three inefficiencies in modern GPU and system architectures. First, not all parallel threads have a uniform amount of workload to fully utilize GPU’s computation ability, leading to a sub-optimal performance problem, called warp criticality. To mitigate the degree of warp criticality, I propose a Criticality-Aware Warp Acceleration mechanism, called CAWA. CAWA predicts and accelerates the critical warp execution by allocating larger execution time slices and additional cache resources to the critical warp. The evaluation result shows that with CAWA, GPUs can achieve an average of 1.23x speedup. Second, the shared cache storage in GPUs is often insufficient to accommodate demands of the large number of concurrent threads. As a result, cache thrashing is commonly experienced in GPU’s cache memories, particularly in the L1 data caches. To alleviate the cache contention and thrashing problem, I develop an instruction aware Control Loop Based Adaptive Bypassing algorithm, called Ctrl-C. Ctrl-C learns the cache reuse behavior and bypasses a portion of memory requests with the help of feedback control loops. The evaluation result shows that Ctrl-C can effectively improve cache utilization in GPUs and achieve an average of 1.42x speedup for cache sensitive GPGPU workloads. Finally, GPU workloads and the co-located processes running on the host chip multiprocessor (CMP) in a heterogeneous system setup can contend for memory resources in multiple levels, resulting in significant performance degradation. To maximize the system throughput and balance the performance degradation of all co-located applications, I design a scalable performance degradation predictor specifically for heterogeneous systems, called HeteroPDP. HeteroPDP predicts the application execution time and schedules OpenCL workloads to run on different devices based on the optimization goal. The evaluation result shows HeteroPDP can improve the system fairness from 24% to 65% when an OpenCL application is co-located with other processes, and gain an additional 50% speedup compared with always offloading the OpenCL workload to GPUs. In summary, this dissertation aims to provide insights for the future microarchitecture and system architecture designs by identifying, analyzing, and addressing three critical performance problems in modern GPUs. / Dissertation/Thesis / Doctoral Dissertation Computer Engineering 2017