The use of computer aided design (CAD) tools has catalyzed the growth of IC design techniques. The rapid growth in transistor count for synchronous digital circuits has increased circuit complexity. This growing complexity of synchronous circuits has exposed design issues such as clock skew, increased power consumption, increased electromagnetic interference and worst case performance. The increasing number of challenges posed by synchronous designs has encouraged researchers to explore asynchronous design techniques as an alternative methodology. Asynchronous circuits do not use a global clock signal that is the primary cause of many design challenges faced by synchronous designers. It has also been shown in some designs that asynchronous circuits consumes less power, and exhibits better average case performance than synchronous circuits. Asynchronous design techniques, even with their various advantages over synchronous systems, are not widely accepted by logic designers. This is due to the shortcomings of asynchronous design methodologies, primarily, the limited availability of CAD tool support and the use of proprietary specification languages. To overcome the shortcomings of current asynchronous design techniques, this research uses a methodology for designing asynchronous circuits starting from clocked RTL design. This research extends the concepts of Phased Logic (PL) and marked graphs to quasi-delay insensitive gates (QDI) gates to create an asynchronous PL-QDI methodology. The PL methodology is easy to use as it maps conventional RTL designs into delay insensitive PL circuits using commercial CAD tools. Caltech?s QDI gates exhibit fast forward latency, but the use of Caltech?s methodology requires a user skilled in the pecurialities of the Caltech design methodology. This research uses best of Caltech?s QDI circuit methodology and the PL methodology to come up with a new asynchronous PL-QDI methodology. It also presents a synthesis algorithm that uses commercially available synchronous CAD tools to map clocked designs to PL-QDI systems. Results of this research show that third-party clocked RTL codes including intellectual property (IP) cores can be converted to asynchronous PL-QDI systems using the PL-QDI CAD tools presented in this research. This work shows how mature synchronous CAD tools can be used to design clockless circuits.

Pre-Charge Half Buffers (PCHB) and NULL convention Logic (NCL) are two major commercially successful Quasi-Delay Insensitive (QDI) asynchronous paradigms, which are known for their low-power performance and inherent robustness. In industry, QDI circuits are synthesized from their synchronous counterparts using custom synthesis tools. Validation of the synthesized QDI implementation is a critical design prerequisite before fabrication. At present, validation schemes are mostly extensive simulation based that are good enough to detect shallow bugs, but may fail to detect corner-case bugs. Hence, development of formal verification methods for QDI circuits have been long desired. The very few formal verification methods that exist in the related field have major limiting factors. This dissertation presents different formal verification methodologies applicable to PCHB and NCL circuits, and aims at addressing the limitations of previous verification approaches. The developed methodologies can guarantee both safety (full functional correctness) and liveness (absence of deadlock), and are demonstrated using several increasingly larger sequential and combinational PCHB and NCL circuits, along with various ISCAS benchmarks. / National Science Foundation (Grant No. CCF-1717420)

Submitted by PPG Ci?ncia da Computa??o (ppgcc@pucrs.br) on 2018-06-15T14:23:09Z No. of bitstreams: 1 LEONARDO REZENDE JURACY_DIS.pdf: 2268947 bytes, checksum: bedc63f7c14296e039a798403cdeec80 (MD5) / Approved for entry into archive by Sheila Dias (sheila.dias@pucrs.br) on 2018-06-26T12:27:11Z (GMT) No. of bitstreams: 1 LEONARDO REZENDE JURACY_DIS.pdf: 2268947 bytes, checksum: bedc63f7c14296e039a798403cdeec80 (MD5) / Made available in DSpace on 2018-06-26T12:45:06Z (GMT). No. of bitstreams: 1 LEONARDO REZENDE JURACY_DIS.pdf: 2268947 bytes, checksum: bedc63f7c14296e039a798403cdeec80 (MD5) Previous issue date: 2018-03-21 / Atualmente, a abordagem s?ncrona ? a mais utilizada em projeto de circuitos integrados por ser altamente automatizado pelas ferramentas comerciais e por incorporar margens de tempo para garantir o funcionamento correto nos piores cen?rios de varia??es de processo e ambiente, limitando otimiza??es no per?odo do rel?gio e aumentando o consumo de pot?ncia. Por um lado, circuitos ass?ncronos apresentam algumas vantagens em potencial quando comparados com os circuitos s?ncronos, como menor consumo de pot?ncia e maior vaz?o de dados, mas tamb?m podem sofrer com varia??es de processo e ambiente. Por outro lado, circuitos resilientes s?o uma alternativa para manter o circuito funcionando na presen?a de efeitos de varia??o. Sendo assim, foi proposto o circuito Blade que combina as vantagens de circuitos ass?ncronos com circuitos resilientes. Blade utiliza latches em sua implementa??o e mant?m seu desempenho em cen?rios de caso m?dio. Independentemente do estilo de projeto (s?ncrono ou ass?ncrono), durante o processo de fabrica??o de circuitos integrados, algumas imperfei??es podem acontecer, causando defeitos que reduzem o rendimento de fabrica??o. Circuitos defeituosos podem apresentar um comportamento falho, gerando uma sa?da diferente da esperada, devendo ser identificados antes de sua comercializa??o. Metodologias de teste podem ajudar na identifica??o e diagn?stico desse comportamento falho. Projeto visando testabilidade (do ingl?s, Design for Testability - DfT) aumenta a testabilidade do circuito adicionando um grau de controlabilidade e observabilidade atrav?s de diferentes t?cnicas. Scan ? uma t?cnica de DfT que fornece para um equipamento de teste externo acesso aos elementos de mem?ria internos do circuito, permitindo inser??o de padr?es de teste e compara??o da resposta. O objetivo deste trabalho ? propor uma abordagem de DfT estrutural, completamente autom?tica e integrada com as ferramentas comerciais de projeto de circuitos, incluindo uma s?rie de m?todos para lidar com os desafios relacionados ao teste de circuitos ass?ncronos e resilientes, com foco no Blade. O fluxo de DfT proposto ? avaliado usando um m?dulo criptogr?fico e um microprocessador. Os resultados obtidos para o m?dulo criptogr?fico mostram uma cobertura de falha de 98,17% para falhas do tipo stuck-at e 89,37% para falhas do tipo path-delay, com um acr?scimo de ?rea de 112,16%. Os resultados obtidos para o microprocessador mostram uma cobertura de 96,04% para falhas do tipo stuck-at e 99,00% para falhas do tipo path-delay, com um acr?scimo de ?rea de 50,57%. / Nowadays, the synchronous circuits design approach is the most used design method since it is highly automated by commercial computer-aided design (CAD) tools. Synchronous designs incorporate timing margins to ensure the correct behavior under the worstcase scenario of process and environmental variations, limiting its clock period optimization and increasing power consumption. On one hand, asynchronous designs present some potential advantages when compared to synchronous ones, such as less power consumption and more data throughput, but they may also suffer with the process and environmental variations. On the other hand, resilient circuits techniques are an alternative to keep the design working in presence of effects of variability. Thus, Blade template has been proposed, combining the advantages of both asynchronous and resilient circuits. The Blade template employs latches in its implementation and supports average-case circuit performance. Independently of the design style (synchronous or asynchronous), during the fabrication process of integrated circuits, some imperfections can occur, causing defects that reduce the fabrication yield. These defective ICs can present a faulty behavior, which produces an output different from the expected, and it must be identified before the circuit commercialization. Test methodologies help to find and diagnose this faulty behavior. Design for Testability (DfT) increases circuit testability by adding a degree of controllability and observability through different test techniques. Scan design is a DfT technique that provides for an external test equipment the access to the internal memory elements of a circuit, allowing test pattern insertion and response comparison. The goal of this work is to propose a fully integrated and automated structural DfT approach using commercial EDA tools and to propose a series of design methods to address the challenges related to testing asynchronous and resilient designs, with focus on Blade template. The proposed DfT flow is evaluated with a criptocore module and a microprocessor. The obtained results for the criptocore module show a fault coverage of 98.17% for stuck-at fault model and 89.37% for path-delay fault model, with an area overhead of 112.16%. The obtained results for the microprocessor show a fault coverage of 96.04% for stuck-at fault model and 99.00% for path-delay fault model, with an area overhead of 50.57%.

Submitted by PPG Ci?ncia da Computa??o (ppgcc@pucrs.br) on 2018-05-21T13:19:36Z No. of bitstreams: 1 FELIPE_AUGUSTO_KUENTZER_TES.pdf: 3277301 bytes, checksum: 7e77c5eb72299302d091329bde56b953 (MD5) / Approved for entry into archive by Sheila Dias (sheila.dias@pucrs.br) on 2018-06-01T12:13:22Z (GMT) No. of bitstreams: 1 FELIPE_AUGUSTO_KUENTZER_TES.pdf: 3277301 bytes, checksum: 7e77c5eb72299302d091329bde56b953 (MD5) / Made available in DSpace on 2018-06-01T12:33:57Z (GMT). No. of bitstreams: 1 FELIPE_AUGUSTO_KUENTZER_TES.pdf: 3277301 bytes, checksum: 7e77c5eb72299302d091329bde56b953 (MD5) Previous issue date: 2018-03-28 / ? medida que o projeto de VLSI avan?a para tecnologias ultra submicron, as margens de atraso adicionadas para compensar variabilidades de processo de fabrica??o, temperatura de opera??o e tens?o de alimenta??o, tornam-se uma parte significativa do per?odo de rel?gio em circuitos s?ncronos tradicionais. As arquiteturas resilientes a varia??es de atraso surgiram como uma solu??o promissora para aliviar essas margens de tempo projetadas para o pior caso, melhorando o desempenho do sistema e reduzindo o consumo de energia. Essas arquiteturas incorporam circuitos adicionais para detec??o e recupera??o de viola??es de atraso que podem surgir ao projetar o circuito com margens de tempo menores. Os sistemas ass?ncronos apresentam potencial para melhorar a efici?ncia energ?tica e o desempenho devido ? aus?ncia de um sinal de rel?gio global. Al?m disso, os circuitos ass?ncronos s?o conhecidos por serem robustos a varia??es de processo, tens?o e temperatura. Blade ? um modelo que incorpora as vantagens de projeto ass?ncrono e resilientes a varia??es de atraso. No entanto, o Blade ainda apresenta desafios em rela??o ? sua testabilidade, o que dificulta sua aplica??o comercial ou em larga escala. Embora o projeto visando testabilidade com Scan seja amplamente utilizado na ind?stria, os altos custos de sil?cio associados com o seu uso no Blade podem ser proibitivos. Por outro lado, os circuitos ass?ncronos podem apresentar vantagens para testes funcionais, enquanto o circuito resiliente fornece feedback cont?nuo durante o funcionamento normal do circuito, uma caracter?stica que pode ser aplicada para testes concorrentes. Nesta Tese, a testabilidade do Blade ? avaliada sob uma perspectiva diferente, onde o circuito implementado com o Blade apresenta propriedades de confiabilidade que podem ser exploradas para testes. Inicialmente, um m?todo de classifica??o de falhas que relaciona padr?es comportamentais com falhas estruturais dentro da l?gica de detec??o de erro e uma nova implementa??o orientada para teste desse m?dulo de detec??o s?o propostos. A parte de controle ? analisada para falhas internas, e um novo projeto ? proposto, onde o teste ? melhorado e o circuito pode ser otimizado pelo fluxo de projeto. Um m?todo original de medi??o de tempo das linhas de atraso tamb?m ? abordado. Finalmente, o teste de falhas de atrasos em caminhos cr?ticos do caminho de dados ? explorado como uma consequ?ncia natural de um circuito implementado com Blade, onde o monitoramento cont?nuo para detec??o de viola??es de atraso fornece a informa??o necess?ria para a detec??o concorrente de viola??es que extrapolam a capacidade de recupera??o do circuito resiliente. A integra??o de todas as contribui??es fornece uma cobertura de falha satisfat?ria para um custo de ?rea que, para os circuitos avaliados nesta Tese, pode variar de 4,24% a 6,87%, enquanto que a abordagem Scan para os mesmos circuitos apresenta custo que varia de 50,19% a 112,70% em ?rea, respectivamente. As contribui??es desta Tese demonstraram que, com algumas melhorias na arquitetura do Blade, ? poss?vel expandir sua confiabilidade para al?m de um sistema de toler?ncia a viola??es de atraso no caminho de dados, e tamb?m um avan?o para teste de falhas (inclusive falhas online) de todo o circuito, bem como melhorar seu rendimento, e lidar com quest?es de envelhecimento. / As the VLSI design moves into ultra-deep-submicron technologies, timing margins added due to variabilities in the manufacturing process, operation temperature and supply voltage become a significant part of the clock period in traditional synchronous circuits. Timing resilient architectures emerged as a promising solution to alleviate these worst-case timing margins, improving system performance and/or reducing energy consumption. These architectures embed additional circuits for detecting and recovering from timing violations that may arise after designing the circuit with reduced time margins. Asynchronous systems, on the other hand, have a potential to improve energy efficiency and performance due to the absence of a global clock. Moreover, asynchronous circuits are known to be robust to process, voltage and temperature variations. Blade is an asynchronous timing resilient template that leverages the advantages of both asynchronous and timing resilient techniques. However, Blade still presents challenges regarding its testability, which hinders its commercial or large-scale application. Although the design for testability with scan chains is widely applied in the industry, the high silicon costs associated with its use in Blade can be prohibitive. Asynchronous circuits can also present advantages for functional testing, and the timing resilient characteristic provides continuous feedback during normal circuit operation, which can be applied for concurrent testing. In this Thesis, Blade?s testability is evaluated from a different perspective, where circuits implemented with Blade present reliability properties that can be explored for stuck-at and delay faults testing. Initially, a fault classification method that relates behavioral patterns with structural faults inside the error detection logic and a new test-driven implementation of this detection module are proposed. The control part is analyzed for internal faults, and a new design is proposed, where the test coverage is improved and the circuit can be further optimized by the design flow. An original method for time measuring delay lines is also addressed. Finally, delay fault testing of critical paths in the data path is explored as a natural consequence of a Blade circuit, where the continuous monitoring for detecting timing violations provide the necessary feedback for online detection of these delay faults. The integration of all the contributions provides a satisfactory fault coverage for an area overhead that, for the evaluated circuits in this thesis, can vary from 4.24% to 6.87%, while the scan approach for the same circuits implies an area overhead varying from 50.19% to 112.70%, respectively. The contributions of this Thesis demonstrated that with a few improvements in the Blade architecture it is possible to expand its reliability beyond a timing resilient system to delay violations in the data path, but also advances for fault testing (including online faults) of the entire circuit, yield, and aging.

After a brief review of machine learning techniques and applications, this Ph.D. thesis examines several approaches for implementing machine learning architectures and algorithms into hardware within our laboratory. From this interdisciplinary background support, we have motivations for novel approaches that we intend to follow as an objective of innovative hardware implementations of dynamically self-reconfigurable logic for enhanced self-adaptive, self-(re)organizing and eventually self-assembling machine learning systems, while developing this new particular area of research. And after reviewing some relevant background of robotic control methods followed by most recent advanced cognitive controllers, this Ph.D. thesis suggests that amongst many well-known ways of designing operational technologies, the design methodologies of those leading-edge high-tech devices such as cognitive chips that may well lead to intelligent machines exhibiting conscious phenomena should crucially be restricted to extremely well defined constraints. Roboticists also need those as specifications to help decide upfront on otherwise infinitely free hardware/software design details. In addition and most importantly, we propose these specifications as methodological guidelines tightly related to ethics and the nowadays well-identified workings of the human body and of its psyche.