Este trabalho trata de arquiteturas e algoritmos para o desenvolvimento de sistemas computacionais. Tais sistemas são constituídos de um microprocessador (específico ou comercialmente disponível), de seu conjunto de programas e de um HW dedicado que será utilizado para otimização do sistema. O objetivo principal desta tese é demonstrar que, presentemente, a linha divisória entre HW e SW e cada vez mais tênue, e a transição entre um e outro pode ser feita de maneira suave pelo projetista de sistemas, na busca de um ponto ótimo no balanço entre custo e desempenho. Apresenta-se em seqüência o ambiente de CAD, a classificação de rotinas e os métodos de otimização tendo em vista esta classificação para o aumento de desempenho de sistemas computacionais. A seguir são apresentadas técnicas para processadores dedicados de arquitetura Risc, visando a otimização de certos tipos de programas. Os resultados de aceleração são apresentados para um conjunto de exemplos. Tendo em vista o mercado nacional de eletrônica, fortemente baseado em microcontroladores, estudam-se e mostram-se possibilidades de otimização e integração de sistemas baseados em tais processadores, assim como a aplicabilidade das mesmas técnicas para processadores dedicados. A viabilidade técnica desta realização é discutida através de exemplos baseados em aplicações reais. Finalmente, a validação de sistemas computacionais, em especial aqueles trabalhados nesta tese, é discutida. / This work discusses architectures and algorithms for the development of computational systems, which are based on a microprocessor (custom or off-the-shelf), the set of application programs and a dedicated HW, used to increase the performance of the whole system. The goal of this work is to show that, nowadays, the division line between SW and HW is smooth, and the transition from one to the other can be achieved by the system designer using a specific CAD in order to obtain a trade-off between cost and performance. The CAD environment is presented, followed by routine classification and optimization methods based on the former classification to increase the performance of the system. Techniques devoted to systems based on dedicated Risc processors are showed next, to optimize certain type of programs. Positive results are shown for a set of examples. Since the Brazilian electronics market is strongly based on microcontrollers, the study and results of optimization techniques regarding this type of systems are also presented. The same techniques can be applied to dedicated processors as well. Results of this proposal are obtained for a set of real world examples. The last topic of this work regards the validation of computational systems, mainly those presented throughout this work.

Embedded systems are becoming ubiquitous, primarily due to the fast evolution of digital electronic devices. The design of modern embedded systems requires systems to exhibit, high performance and reliability, yet have short design time and low cost. Application Specific Instruction set processors (ASIPs) are widely used in embedded system since they are economical to use, flexible, and reusable (thus saves design time). During the last decade research work on ASIPs have been carried out in mainly for single pipelined processors. Improving performance in processors is possible by exploring the available parallelism in the program. Designing of multiple parallel execution paths for parallel execution of the processor naturally incurs additional cost. The methodology presented in this dissertation has addressed the problem of improving performance in ASIPs, at minimal additional cost. The devised methodology explores the available parallelism of an application to generate a multi-pipeline heterogeneous ASIP. The processor design is application specific. No pre-defined IPs are used in the design. The generated processor contains multiple standalone pipelined data paths, which are not necessarily identical, and are connected by the necessary bypass paths and control signals. Control unit are separate for each pipeline (though with the same clock) resulting in a simple and cost effective design. By using separate instruction and data memories (Harvard architecture) and by allowing memory access by two separate pipes, the complexity of the controller and buses are reduced. The impact of higher memory latencies is nullified by utilizing parallel pipes during memory access. Efficient bypass network selection and encoding techniques provide a better implementation. The initial design approach with only two pipelines without bypass paths show speed improvements of up to 36% and switching activity reductions of up to 11%. The additional area costs around 16%. An improved design with different number of pipelines (more than two) based on applications show on average of 77% performance improvement with overheads of: 49% on area; 51% on leakage power; 17% on switching activity; and 69% on code size. The design was further trimmed, with bypass path selection and encoding techniques, which show a saving of up to 32% of area and 34% of leakage power with 6% performance improvement and 69% of code size reduction compared to the design approach without these techniques in the multi pipeline design.

Les processeurs à jeux d'instructions spécifiques (ASIP) sont des processeurs spécialisés qui combinent la flexibilité d'un processeur programmable avec la performance d'un processeur dédié. L'une des approches de conception de tels processeurs consiste à spécialiser un cœur de processeur existant en y ajoutant des instructions spécialisées, mises en œuvre dans un module matériel fortement couplé au chemin de données du processeur. C'est l'extension de jeu d'instructions. La conception d'un ASIP nécessite des méthodologies et des outils logiciels appropriés garantissant une maîtrise des contraintes de conception et de la complexité grandissante des applications. Dans ce contexte, cette thèse vise à proposer une méthodologie de génération automatique d'extensions de jeux d'instructions. Celle-ci consiste à tout d'abord identifier l'ensemble des instructions candidates qui satisfont les contraintes architecturales et technologiques, afin de garantir leurs mises en œuvre. Ensuite, les instructions candidates qui minimisent le temps d'exécution séquentielle de l'application sont sélectionnées. Les ressources matérielles de l'extension, telles que les registres et les multiplexeurs, sont optimisées. Enfin, la dernière étape génère la description matérielle et le modèle de simulation de l'extension. Le code applicatif est adapté pour tenir compte des nouvelles instructions. Cette thèse propose des techniques basées sur la programmation par contraintes pour résoudre les problèmes difficiles (voir intraitables) que sont l'identification d'instructions, la sélection d'instructions et l'allocation de registres.

The number of application specific instruction set processors is rapidly increasing, because of increased demand for low power and small area designs. A lot of new instruction sets are born, but they are usually confidential. University of California in Berkeley took an opposite approach. The RISC-V instruction set is completely free. This master's thesis focuses on analysis of RISC-V instruction set and two programming languages used to model instruction sets and microarchitectures, CodAL and Chisel. Implementation of RISC-V base instruction set along with multiplication, division and 64-bit address space extensions and implementation of cycle accurate model of Rocket Core-like microarchitecture in CodAL are main goals of this master's thesis. The instruction set model is used to generate the C compiler and the cycle accurate model is used to generate RTL representation, all thanks to Codasip Studio. Generated compiler is compared against the one implemented manually and results are used for instruction set optimizations. RTL is synthesized to Artix 7 FPGA and compared to the Rocket Core synthesis.

Fatigue is a well-recognized issue in lightweight and high-performance aircraft structures. As fatigue failures have led to serious accidents and caused significant economic impact in the past, design against fatigue is crucial. Fatigue testing of full-scale aircraft as well as components is an important tool for the advance identification of potential fatigue issues in both new and operational aircraft. Furthermore, coupon testing is used extensively to obtain allowables for materials and structural details to be used in the design process. To obtain accurate results from fatigue testing, not only the test object but also the used load spectrum must accurately represent reality. If the aircraft is operational, an accurate load spectrum can be obtained by measuring the loads in-flight during a sufficiently long period of normal operation of the aircraft. However, the in-flight measured loads data contains an extraordinarily large number of cycles, resulting in long and uneconomical test durations. This thesis aims to propose a method for the selection of an optimal filtering level for fatigue test spectra developed from in-flight measured loads. The thesis also discusses and recommends methods for in-flight measurement of loads, cycle counting as well as damage evaluation using a crack-growth approach. Furthermore, ways to validate the proposed method and its practical application are discussed. An example filtering study is conducted using four different specimens chosen to represent typical structural details of aircraft. The study uses real in-flight measured loads of a light aircraft and also discusses temperature compensation of the loads data. The effect of filtering on fatigue damage is evaluated using crack-growth simulations conducted at a range of filtering and stress levels.  The results show that a remarkable reduction of testing time is possible and as many as 99 % of all cycles in the studied flight load history can be discarded without significantly reducing fatigue damage. The allowable filtering level is shown to differ between the specimens and the different stages of fatigue crack growth. In addition, the applied stress level is found to have a consistent effect on the allowable filtering level.

L'implantation efficace des algorithmes de traitement numérique du signal (TNS) dans les systèmes embarqués requiert l'utilisation de l'arithmétique virgule fixe afin de satisfaire les contraintes de coût, de consommation et d'encombrement exigées par ces applications. Le codage manuel des données en virgule fixe est une tâche fastidieuse et source d'erreurs. De plus, la réduction du temps de mise sur le marché des applications exige l'utilisation d'outils de développement de haut niveau, permettant d'automatiser certaines tâches. Ainsi, le développement de méthodologies de codage automatique des données en virgule fixe est nécessaire. Dans le cadre des processeurs programmables de traitement du signal, la méthodologie doit déterminer le codage optimal, permettant de maximiser la précision et de minimiser le temps d'exécution et la taille du code. L'objectif de ce travail de recherche est de définir une nouvelle méthodologie de compilation d'algorithmes spécifiés en virgule flottante au sein d'architectures programmables en virgule fixe sous contrainte de respect des critères de qualité associés à l'application. Ce travail de recherche s'articule autour de trois points principaux. Le premier aspect de notre travail a consisté à définir la structure de la méthodologie. L'analyse de l'influence de l'architecture sur la précision des calculs montre la nécessité de tenir compte de l'architecture cible pour obtenir une implantation optimisée d'un point de vue du temps d'exécution et de la précision. De plus, l'étude de l'interaction entre les phases de compilation et de codage des données permet de définir le couplage nécessaire entre les phases de conversion en virgule fixe et le processus de génération de code. Le second aspect de ce travail de recherche concerne l'évaluation de la précision au sein d'un système en virgule fixe à travers la détermination du Rapport Signal à Bruit de Quantification (RSBQ). Une méthodologie permettant de déterminer automatiquement l'expression analytique du RSBQ en fonction du format des données en virgule fixe est proposée. Dans un premier temps, un nouveau modèle de bruit est présenté. Ensuite, les concepts théoriques pour déterminer la puissance du bruit de quantification en sortie des systèmes linéaires et des systèmes non-linéaires et non-récursifs sont détaillés. Finalement, la méthodologie mise en oeuvre pour obtenir automatiquement l'expression du RSBQ dans le cadre des systèmes linéaires est exposée. Le troisième aspect de ce travail de recherche correspond à la mise en oeuvre de la méthodologie de codage des données en virgule fixe. Dans un premier temps, la dynamique des données est déterminée à l'aide d'une approche analytique combinant deux techniques différentes. Ces informations sur la dynamique permettent de déterminer la position de la virgule de chaque donnée en tenant compte de la présence éventuelle de bits de garde au sein de l'architecture. Pour obtenir un format des données en virgule fixe complet, la largeur de chaque donnée est déterminée en prenant en compte l'ensemble des types des données manipulées au sein du DSP. La méthode sélectionne la séquence d'instructions permettant de fournir une précision suffisante en sortie de l'algorithme et de minimiser le temps d'exécution du code. La dernière phase du processus de codage correspond à l'optimisation du format des données en vue d'obtenir une implantation plus efficace. Les différentes opérations de recadrage sont déplacées afin de minimiser le temps d'exécution global tant que la précision en sortie de l'algorithme est supérieure à la contrainte. Deux types de méthode ont été mis en {\oe}uvre en fonction des capacités de parallélisme au niveau instruction de l'architecture ciblée. Cette méthodologie a été testée sur différents algorithmes de traitement numérique du signal présents au sein des systèmes de radio-communications de troisième génération. Les résultats obtenus montrent l'intérêt de notre méthodologie pour réduire le temps de développement des systèmes en virgule fixe.

This thesis describes the design of the universal assembler that represents a part of the Lissom project. You will be provided with the description of the assembler architectures and their usual tasks. Special attention is paid to GNU assembler. Designed assembler consists of the fixed and the generated part. The generated part is created automatically from the description of instruction set, that is defined using architecture and instructions set description language ISAC. Using this approach, it is possible to change assembler target architecture automatically. The second part of thesis describes the Parserlib2 library implementation that is a part of the Lissom project and provides the information about the target instruction set for an assembler generator.