Os efeitos de radiação têm sido um dos problemas mais sérios em aplicações militares e espaciais. Mas eles também são uma preocupação crescente em tecnologias modernas, mesmo para aplicações comerciais no nível do solo. A proteção dos circuitos integrados contra os efeitos da radiação podem ser obtidos através do uso de processos de fabricação aprimorados e de estratégias em diferentes estágios do projeto do circuito. A técnica de TMR é bem conhecida e amplamente empregada para mascarar falhas únicas sem detectálas. No entanto, o projeto de circuitos TMR não é automatizado por ferramentas EDA comerciais e até mesmo eles podem remover parcial ou totalmente a lógica redundante. Por outro lado, existem várias ferramentas que podem ser usadas para implementar a técnica de TMR em circuitos integrados, embora a maioria delas sejam ferramentas comerciais licenciadas, convenientes apenas para dispositivos específicos, ou com uso restrito por causa do regime ITAR. O presente trabalho pretende superar esses incovenientes, para isso uma metodologia é proposta para automatizar o projeto de circuitos TMR utilizando um fluxo de projeto comercial. A abordagem proposta utiliza um netlist estruturado para implementar automaticamente os circuitos TMR em diferentes níveis de granularidade de redundância para projetos baseados em células e FPGA. A otimização do circuito TMR resultante também é aplicada com base na abordagem do dimensionamento de portas lógicas. Além disso, a verificação do circuito TMR implementado é baseada na verificação de equivalência e garante sua funcionalidade correta e sua capacidade de tolerancia a falhas simples. Experimentos com um circuito derivado de HLS e uma descrição ofuscada do soft-core ARM Cortex-M0 foram realizados para mostrar o uso e as vantagens do fluxo de projeto proposto. Diversas questões relacionadas à remoção da lógica redundante implementada foram encontradas, bem como o impacto no incremento de área causado pelos votadores de maioria. Além disso, a confiabilidade de diferentes implementações de TMR do soft core ARM sintetizado em FPGA foi avaliada usando campanhas de injeção de falhas emuladas. Como resultado, foi reforçado o nível de alta confiabilidade da implemntação com mais fina granularidade, mesmo na presença de até 10 falhas acumuladas, e a menor capacidade de mitigação correspondente à replicação de flip-flops apenas. / Radiation effects have been one of the most serious issues in military and space applications. But they are also an increasing concern in modern technologies, even for commercial applications at the ground level. Protection or hardening of integrated circuits against radiation effects can be obtained through the use of enhanced fabrication processes and strategies at different stages of the circuit design. The triple modular redundancy (TMR) technique is a widely and well-known technique employed to mask single faults without detecting them. However, the design of TMR circuits is not automated by commercial electronic design automation (EDA) tools and even they can remove partially or totally the redundant logic. On the other hand, there are several tools that can be used to implement the TMR technique in integrated circuits, although most of them are licensed commercial tools, convenient only for specific devices, or with restricted use because of the International Traffic in Arms Regulations (ITAR) regimen. The present work intends to overcome these issues so a methodology is proposed to automate the design of TMR circuits using a commercial design flow. The proposed approach uses a structured netlist to implement automatically TMR circuits at different granularity levels of redundancy for cell-based and field-programmable gate array (FPGA) designs. Optimization of the resulting TMR circuit is also applied based on the gate sizing approach. Moreover, verification of the implemented TMR circuit is based on equivalence checking, and guarantee its correct functionality and its fault-tolerant capability against soft errors. Experiments with an high-level synthesis (HLS)-derived circuit and an obfuscated description of the ARM Cortex-M0 soft-core are performed to show the use and the advantages of the proposed design flow. Several issues related to the removal of the implemented redundant logic were found as well as the impact in the increment of area caused by the majority voters. Furthermore, the reliability of different TMR implementations of the ARM soft-core synthesized in FPGA was evaluated using emulated-simulation fault injection campaigns. As a result, it was reinforced the high-reliability level of the finest granularity implementation even in the presence of up to 10 accumulated faults and the poorest mitigation capacity corresponding to the replication of flip-flops solely.

Internet plays a crucial part in today's world. Be it personal communication, business transactions or social networking, internet is used everywhere and hence the speed of the communication infrastructure plays an important role. As the number of users increase the network usage increases i.e., the network data rates ramped up from a few Mb/s to Gb/s in less than a decade. Hence the network infrastructure needed a major upgrade to be able to support such high data rates. Technological advancements have enabled the communication links like optical fibres to support these high bandwidths, but the processing speed at the nodes remained constant. This created a need for specialised devices for packet processing in order to match the increasing line rates which led to emergence of network processors. Network processors were both programmable and flexible. To support the growing number of internet applications, a single core network processor has transformed into a multi/many core network processor with multiple cores on a single chip rather than just one core. This improved the packet processing speeds and hence the performance of a network node. Multi-core network processors catered to the needs of a high bandwidth networks by exploiting the inherent packet-level parallelism in a network. But these processors still had intrinsic challenges like load balancing. In order to maximise throughput of these multi-core network processors, it is important to distribute the traffic evenly across all the cores. This thesis describes a multi-core network processor with dynamic workload management. A multi-core network processor, which performs multiple applications is designed to act as a test bed for an effective workload management algorithm. An effective workload management algorithm is designed in order to distribute the workload evenly across all the available cores and hence maximise the performance of the network processor. Runtime statistics of all the cores were collected and updated at run time to aid in deciding the application to be performed on a core to to enable even distribution of workload among the cores. Hence, when an overloading of a core is detected, the applications to be performed on the cores are re-assigned. For testing purposes, we built a flexible and a reusable platform on NetFPGA 10G board which uses a FPGA-based approach to prototyping network devices. The performance of the designed workload management algorithm is tested by measuring the throughput of the system for varying workloads.

Tato práce je zaměřena na zpracování a analýzu USB komunikačního protokolu a implementace jeho jednotlivých částí do FPGA obvodu s využitím programovacího jazyka VHDL. Ve finální podobě by měla práce představovat souhrnný a ucelený dokument popisující principy USB rozhraní a jeho komunikace doplněných praktickým návrhem v jazyce VHDL, který by byl schopen převést data do USB.

The CUBES (CUbesat x-ray Background Explorer using Scintillators) is a payload on the KTH student satellite MIST (MIniature STudent satellite) to evaluate Silicon Photo-multiplier technology and new scintillators such as GAGG (Gadolinium Aluminium Gallium Garnet, Gd3Al2Ga3O12) for future use in hard X-ray polarisation studies of Gamma-Ray Bursts. CUBES itself is designed to study the MIST in-orbit radiation environment by using a detector which is comprised of a silicon photomultiplier coupled to different scintillator materials. Three of these detectors will be mounted on the payload platform and then coupled to inputs of an Application Specific Integrated Circuit (ASIC) and connected to a Field-programmable Gate Array (FPGA) which will store and send data through the downlink on the MIST satellite to ground. This thesis covers the software development for the FPGA, together with two radiation tests of components and the preparation of these. / CUBES / MIST

abstract: While machine/deep learning algorithms have been successfully used in many practical applications including object detection and image/video classification, accurate, fast, and low-power hardware implementations of such algorithms are still a challenging task, especially for mobile systems such as Internet of Things, autonomous vehicles, and smart drones. This work presents an energy-efficient programmable application-specific integrated circuit (ASIC) accelerator for object detection. The proposed ASIC supports multi-class (face/traffic sign/car license plate/pedestrian), many-object (up to 50) in one image with different sizes (6 down-/11 up-scaling), and high accuracy (87% for face detection datasets). The proposed accelerator is composed of an integral channel detector with 2,000 classifiers for five rigid boosted templates to make a strong object detection. By jointly optimizing the algorithm and efficient hardware architecture, the prototype chip implemented in 65nm demonstrates real-time object detection of 20-50 frames/s with 22.5-181.7mW (0.54-1.75nJ/pixel) at 0.58-1.1V supply. In this work, to reduce computation without accuracy degradation, an energy-efficient deep convolutional neural network (DCNN) accelerator is proposed based on a novel conditional computing scheme and integrates convolution with subsequent max-pooling operations. This way, the total number of bit-wise convolutions could be reduced by ~2x, without affecting the output feature values. This work also has been developing an optimized dataflow that exploits sparsity, maximizes data re-use and minimizes off-chip memory access, which can improve upon existing hardware works. The total off-chip memory access can be saved by 2.12x. Preliminary results of the proposed DCNN accelerator achieved a peak 7.35 TOPS/W for VGG-16 by post-layout simulation results in 40nm. A number of recent efforts have attempted to design custom inference engine based on various approaches, including the systolic architecture, near memory processing, and in-meomry computing concept. This work evaluates a comprehensive comparison of these various approaches in a unified framework. This work also presents the proposed energy-efficient in-memory computing accelerator for deep neural networks (DNNs) by integrating many instances of in-memory computing macros with an ensemble of peripheral digital circuits, which supports configurable multibit activations and large-scale DNNs seamlessly while substantially improving the chip-level energy-efficiency. Proposed accelerator is fully designed in 65nm, demonstrating ultralow energy consumption for DNNs. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2019

This thesis presents a novel approach to operation-centric tracing for hardware debug with a retrospective analysis of traces which are distributed across a computer system. Therefore, these traces record entries about the operations at runtime, and a software tool correlates these entries after a problem occurred. This tool is based on a generic method using identifiers saved from operations. Because identifiers are changed along the path of an operation through the system and traces record different information, the entries are transformed to find matching entries in other traces. After the correlation, the method reconstructs the operation paths with help of an operation graph which describes for each type of operation the subtasks and their sequence. With these paths the designer gets a better overview about the chip or system activity, and can isolate the problem cause faster. The TRACE MATCHER implements the described method and it is evaluated with an example bridge chip. Therefore, the benefit for hardware debug, correctness of the reconstructed paths, the performance of their Implementation, and the configuration effort are evaluated. At the end guidelines for trace and system design describe how matching can be improved by carefully designed identifiers at operations.

Es wurden verschieden Kodierungsarten fuer FSMs untersucht, schwerpunktmaessig Gray Code und andere Arten der hazardfreien Kodierung. Ein spezieller Kodierungsalgorithmus zur hazardfreien Kodierung wurde entwickelt und in eine Entwurfsumgebung implementiert. Ein weitere Schwerpunkt der Arbeit sind Codegeneratoren, die eine Verhaltensbeschreibung der FSM in Verilog oder in VHDL erzeugen.

In this work, a coordinate wise descent algorithm is implemented which serves the purpose of estimating active users in a base station/client wireless communication setup. The implemented algorithm utilizes the sporadic nature of users, which is believed to be the norm with 5G Massive MIMO and Internet of Things, meaning that only a subset of all users are active simultaneously at any given time. This work attempts to estimate the viability of a direct algorithm implementation to test if the performance requirements can be satisfied or if a more sophisticated implementation, such as a parallelized version, needs to be created.The result is an isomorphic ASIC implementation made in a 28 nm FD-SOI process, with proper internal word lengths extracted through simulation. Some techniques to lessen the burden on hardware without losing performance is presented which helps reduce area and increase speed of the implementation. Finally, a parallelized version of the algorithm is proposed, if one should desire to explore an implementation with higher system throughput, at almost no furtherexpense of user estimation error.

Les systèmes de communication ont besoin d'accéder, stocker, manipuler, ou de communiquer des informations sensibles. Par conséquent, les primitives cryptographiques tels que les fonctions de hachage et le chiffrement par blocs sont déployés pour fournir le cryptage et l'authentification. Récemment, des techniques ont été inventés pour combiner cryptage et d'authentification en un seul algorithme qui est appelé authentifiés Encryption (AE). La combinaison de ces deux services de sécurité dans le matériel de meilleures performances par rapport aux deux algorithmes séparés puisque l'authentification et le cryptage peuvent partager une partie du calcul. En raison de la combinaison de la programmation de l'exécution d'matériel personnalisé, FPGA deviennent plus communs comme cible d'une mise en œuvre de ces algorithmes. La première partie de cette thèse est consacrée aux architectures d'algorithmes AE, AES-GCM et AEGIS-128 à base de FPGA efficaces et à grande vitesse, afin d'être utilisé dans la partie reconfigurable FPGA pour soutenir les services de sécurité des systèmes de communication. Notre focalisation sur l'état de l'art conduit à la mise en place d'architectures à haute vitesse pour les applications lentes touches changeantes comme les réseaux privés virtuels (VPN). En outre, nous présentons un procédé efficace pour mettre en oeuvre le GF($2^{128}$) multiplicateur, qui est responsable de la tâche d'authentification en AES-GCM, pour supporter les applications à grande vitesse. En outre, un système efficace AEGIS-128 est également mis en œuvre en utilisant seulement cinq tours AES. Nos réalisations matérielles ont été évaluées à l'aide Virtex-5 et Virtex-4 FPGA. La performance des architectures présentées (Thr. / Parts) surpasse ceux signalés précédemment.La deuxième partie de la thèse présente des techniques pour des solutions à faible coût afin de garantir la reconfiguration du FPGA. Nous présentons différentes gammes de mises en œuvre à faible coût de AES-GCM, AES-CCM, et AEGIS-128, qui sont utilisés dans la partie statique du FPGA afin de décrypter et authentifier le bitstream FPGA. Architectures ASIC présentées ont été évaluées à l'aide de 90 et 65 technologies nm et présentent de meilleures performances par rapport aux travaux antérieurs. / Communication systems need to access, store, manipulate, or communicate sensitive information. Therefore, cryptographic primitives such as hash functions and block ciphers are deployed to provide encryption and authentication. Recently, techniques have been invented to combine encryption and authentication into a single algorithm which is called Authenticated Encryption (AE). Combining these two security services in hardware produces better performance compared to two separated algorithms since authentication and encryption can share a part of the computation. Because of combining the programmability with the performance ofcustom hardware, FPGAs become more common as an implementation target for such algorithms. The first part of this thesis is devoted to efficient and high-speed FPGA-based architectures of AE algorithms, AES-GCM and AEGIS-128, in order to be used in the reconfigurable part of FPGAs to support security services of communication systems. Our focus on the state of the art leads to the introduction of high-speed architectures for slow changing keys applications like Virtual Private Networks (VPNs). Furthermore, we present an efficient method for implementing the GF($2^{128}$) multiplier, which is responsible for the authentication task in AES-GCM, to support high-speed applications. Additionally, an efficient AEGIS-128is also implemented using only five AES rounds. Our hardware implementations were evaluated using Virtex-5 and Virtex-4 FPGAs. The performance of the presented architectures (Thr./Slices) outperforms the previously reported ones.The second part of the thesis presents techniques for low cost solutions in order to secure the reconfiguration of FPGAs. We present different ranges of low cost implementations of AES-GCM, AES-CCM, and AEGIS-128, which are used in the static part of the FPGA in order to decrypt and authenticate the FPGA bitstream. Presented ASIC architectures were evaluated using 90 and 65 nm technologies and they present better performance compared to the previous work.

No description available.