Trends Of CPU, GPU and FPGA for high-performance computing

Floating-point computing with more than one TFLOP of peak performance is already a reality in recent Field-Programmable Gate Arrays (FPGA). General-Purpose Graphics Processing Units (GPGPU) and recent many-core CPUs have also taken advantage of the recent technological innovations in integrated circuit (IC) design and had also dramatically improved their peak performances. In this paper, we compare the trends of these computing architectures for high-performance computing and survey these platforms in the execution of algorithms belonging to different scientific application domains. Trends in peak performance, power consumption and sustained performances, for particular applications, show that FPGAs are increasing the gap to GPUs and many-core CPUs moving them away from high-performance computing with intensive floating-point calculations. FPGAs become competitive for custom floating-point or fixed-point representations, for smaller input sizes of certain algorithms, for combinational logic problems and parallel map-reduce problems. © 2014 Technical University of Munich (TUM).

Geração automática de controladores em FPGA integrando animação gráfica

Dissertação apresentada na Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa para obtenção do grau de Mestre em Engenharia Electrotécnica e de Computadores

A shared-disk parallel cluster file system

Dissertação apresentada para obtenção do Grau de Doutor em Informática Pela Universidade Nova de Lisboa, Faculdade de Ciências e Tecnologia

Using test infrastructures for (remote) online evaluation of the sensitivity to SEUs of FPGAs

This paper proposes an online mechanism that can evaluate the sensitivity of single event upsets (SEUs) of field programmable gate arrays (FPGAs). The online detection mechanism cyclically reads and compares the values form the external and internal configuration memories, taking into account the mask information. This remote detection method also signals any mismatch as a result of a SEU that affects both used and not-used FPGA parts, which maximizes the monitored area. By utilizing an external, Web-accessible controller that is connected to the test infrastructure, the possibility of running the same operation in a remote manner is enabled. Moreover, the need for a local memory to store the mask values is also eliminated.

Automated design of microwave discrete tuning differential capacitance circuits in Si-integrated technologies

A genetic algorithm used to design radio-frequency binary-weighted differential switched capacitor arrays (RFDSCAs) is presented in this article. The algorithm provides a set of circuits all having the same maximum performance. This article also describes the design, implementation, and measurements results of a 0.25 lm BiCMOS 3-bit RFDSCA. The experimental results show that the circuit presents the expected performance up to 40 GHz. The similarity between the evolutionary solutions, circuit simulations, and measured results indicates that the genetic synthesis method is a very useful tool for designing optimum performance RFDSCAs.

Reliability and availability in reconfigurable computing: a basis for a common solution

Dynamically reconfigurable SRAM-based field-programmable gate arrays (FPGAs) enable the implementation of reconfigurable computing systems where several applications may be run simultaneously, sharing the available resources according to their own immediate functional requirements. To exclude malfunctioning due to faulty elements, the reliability of all FPGA resources must be guaranteed. Since resource allocation takes place asynchronously, an online structural test scheme is the only way of ensuring reliable system operation. On the other hand, this test scheme should not disturb the operation of the circuit, otherwise availability would be compromised. System performance is also influenced by the efficiency of the management strategies that must be able to dynamically allocate enough resources when requested by each application. As those resources are allocated and later released, many small free resource blocks are created, which are left unused due to performance and routing restrictions. To avoid wasting logic resources, the FPGA logic space must be defragmented regularly. This paper presents a non-intrusive active replication procedure that supports the proposed test methodology and the implementation of defragmentation strategies, assuring both the availability of resources and their perfect working condition, without disturbing system operation.

Evidences for direct magnetic patterning via diffusive transformations using femtosecond laser interferometry

The application of femtosecond laser interferometry to direct patterning of thin-film magnetic alloys is demonstrated. The formation of stripe gratings with submicron periodicities is achieved in Fe1-xVx (x=18-34wt. %) layers, with a difference in magnetic moments up to Delta mu/mu similar to 20 between adjacent stripes but without any significant development of the topographical relief (<1% of the film thickness). The produced gratings exhibit a robust effect of their anisotropy shape on magnetization curves in the film plane. The obtained data witness ultrafast diffusive transformations associated with the process of spinodal decomposition and demonstrate an opportunity for producing magnetic nanostructures with engineered properties upon this basis.

Smart object for 3D interaction

This paper reports on the creation of an interface for 3D virtual environments, computer-aided design applications or computer games. Standard computer interfaces are bound to 2D surfaces, e.g., computer mouses, keyboards, touch pads or touch screens. The Smart Object is intended to provide the user with a 3D interface by using sensors that register movement (inertial measurement unit), touch (touch screen) and voice (microphone). The design and development process as well as the tests and results are presented in this paper. The Smart Object was developed by a team of four third-year engineering students from diverse scientific backgrounds and nationalities during one semester.

Run-time defragmentation for dynamically reconfigurable hardware

Reconfigurable computing experienced a considerable expansion in the last few years, due in part to the fast run-time partial reconfiguration features offered by recent SRAM-based Field Programmable Gate Arrays (FPGAs), which allowed the implementation in real-time of dynamic resource allocation strategies, with multiple independent functions from different applications sharing the same logic resources in the space and temporal domains. However, when the sequence of reconfigurations to be performed is not predictable, the efficient management of the logic space available becomes the greatest challenge posed to these systems. Resource allocation decisions have to be made concurrently with system operation, taking into account function priorities and optimizing the space currently available. As a consequence of the unpredictability of this allocation procedure, the logic space becomes fragmented, with many small areas of free resources failing to satisfy most requests and so remaining unused. A rearrangement of the currently running functions is therefore necessary, so as to obtain enough contiguous space to implement incoming functions, avoiding the spreading of their components and the resulting degradation of system performance. A novel active relocation procedure for Configurable Logic Blocks (CLBs) is herein presented, able to carry out online rearrangements, defragmenting the available FPGA resources without disturbing functions currently running.

Algoritmos de estimação da direção de chegada de sinal

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia de Eletrónica e Telecomunicações

A many-core co-processor for embedded parallel computing on FPGA

Single processor architectures are unable to provide the required performance of high performance embedded systems. Parallel processing based on general-purpose processors can achieve these performances with a considerable increase of required resources. However, in many cases, simplified optimized parallel cores can be used instead of general-purpose processors achieving better performance at lower resource utilization. In this paper, we propose a configurable many-core architecture to serve as a co-processor for high-performance embedded computing on Field-Programmable Gate Arrays. The architecture consists of an array of configurable simple cores with support for floating-point operations interconnected with a configurable interconnection network. For each core it is possible to configure the size of the internal memory, the supported operations and number of interfacing ports. The architecture was tested in a ZYNQ-7020 FPGA in the execution of several parallel algorithms. The results show that the proposed many-core architecture achieves better performance than that achieved with a parallel generalpurpose processor and that up to 32 floating-point cores can be implemented in a ZYNQ-7020 SoC FPGA.

FPGA-based architecture for hyperspectral unmixing

This paper proposes an FPGA-based architecture for onboard hyperspectral unmixing. This method based on the Vertex Component Analysis (VCA) has several advantages, namely it is unsupervised, fully automatic, and it works without dimensionality reduction (DR) pre-processing step. The architecture has been designed for a low cost Xilinx Zynq board with a Zynq-7020 SoC FPGA based on the Artix-7 FPGA programmable logic and tested using real hyperspectral datasets. Experimental results indicate that the proposed implementation can achieve real-time processing, while maintaining the methods accuracy, which indicate the potential of the proposed platform to implement high-performance, low cost embedded systems.

System-on-chip field-programmable gate array design for onboard real-time hyperspectral unmixing

Hyperspectral instruments have been incorporated in satellite missions, providing large amounts of data of high spectral resolution of the Earth surface. This data can be used in remote sensing applications that often require a real-time or near-real-time response. To avoid delays between hyperspectral image acquisition and its interpretation, the last usually done on a ground station, onboard systems have emerged to process data, reducing the volume of information to transfer from the satellite to the ground station. For this purpose, compact reconfigurable hardware modules, such as field-programmable gate arrays (FPGAs), are widely used. This paper proposes an FPGA-based architecture for hyperspectral unmixing. This method based on the vertex component analysis (VCA) and it works without a dimensionality reduction preprocessing step. The architecture has been designed for a low-cost Xilinx Zynq board with a Zynq-7020 system-on-chip FPGA-based on the Artix-7 FPGA programmable logic and tested using real hyperspectral data. Experimental results indicate that the proposed implementation can achieve real-time processing, while maintaining the methods accuracy, which indicate the potential of the proposed platform to implement high-performance, low-cost embedded systems, opening perspectives for onboard hyperspectral image processing.

Estruturas de dados para representação de um léxico bilingue

Trabalho apresentado no âmbito do Mestrado em Engenharia Informática, como requisito parcial para obtenção do grau de Mestre em Engenharia Informática

Terapêuticas antigénio-especificas no tratamento das doenças desmielinizantes: estudos sobre a vacinação com ADN e a descoberta de um novo alvo antigénico

RESUMO A Esclerose Múltipla (EM) é uma doença desmielinizante crónica do Sistema Nervoso Central (SNC), provocada, em grande parte, por um ataque imuno-mediado contra diversos elementos da bainha de mielina. Dentro dos alvos antigénicos desta resposta autoimune, vários componentes proteicos e lipídicos da mielina têm vindo a ser identificados ao longo dos anos, entre os quais se destacam a proteína básica de mielina(MBP), glicoproteína ligodendrocitária da mielina (MOG), proteína proteolipídica (PLP) e glicoproteína associada à mielina (MAG). Com o desenvolvimento do modelo animal de Encefalomielite Autoimune Experimental (EAE), diversas terapias antigénio-específicas foram desenhadas, baseadas na modificação benéfica da resposta autoimune contra a mielina, tais como a administração de mielina ou seus componentes, os copolímeros terapêuticos, os ligandos peptídeos alterados e, recentemente, a vacinação com ácido desoxirribonucleico (ADN) codificador de proteínas de mielina, integrado em plasmídeos e purificado para administração parentérica. Neste trabalho, apresentamos os resultados de um extenso conjunto de experiências, subordinadas a dois temas fundamentais: 1) avaliação do potencial terapêutico, e dos mecanismos de acção, da vacinação tolerizadora com ADN codificador de proteínas de mielina (MBP, MOG, PLP, MAG) na EAE, e da associação desta vacinação com a administração de ADN de citocinas Th2, ou de oligonucleótidos imunomoduladores; 2) identificação e caracterização da resposta imune contra um novo componente da mielina com potencial antigénico, a proteína inibidora do recrescimento axonal, Nogo-A. No que respeita à vacinação com ADN, os nossos resultados comprovam a eficácia desta terapêutica antigénio-específica na prevenção e tratamento da EAE. Os seus mecanismos de acção incluem, entre outros, a supressão anérgica da proliferação antigénioespecífica dos linfócitos T anti-mielina (no modo de prevenção da doença), o enviesamento Th2 da resposta imune (quando co-administrada com a vacina de ADN codificadora da citocina IL-4, funcionando como terapia génica local), e a redução da diversificação de epítopos da resposta humoral anti-mielina, avaliada através de myelin spotted arrays. A associação das vacinas de ADN com oligonucleótidos imunomoduladores GpG, desenvolvidos para contrariar as sequências CpG imunoestimuladoras presentes no vector de vacinação, levou à melhoria da sua eficácia terapêutica, devida, provavelmente, ao efeito estimulador preferencial dos oligonucleótidos GpG sobre linfócitos Th2 e sobre células reguladoras NK-T. Com base nestes resultados a vacinação com ADN foi desenvolvida para o tratamento da EM em humanos, com ensaios clínicos a decorrerem neste momento. Em relação à proteína Nogo-A, estudos de estrutura primária e de previsão de antigenicidade identificaram a região Nogo-66 como alvo antigénico potencial para a EAE. Nas estirpes de ratinho SJL/J e C57BL/6, fomos capazes de induzir sinais clínicos e histológicos de EAE após imunização com os epítopos encefalitogénicos Nogo1-22, Nogo23- 44 e Nogo45-66, utilizando protocolos de quebra de tolerância imune. Ao mesmo tempo, identificámos e caracterizámos uma resposta linfocitária T específica contra os antigénios contidos na região Nogo-66, e uma resposta linfocitária B com diversificação intra e intermolecular a vários determinantes presentes noutras proteínas da mielina. A transferência adoptiva de linhas celulares Th2 anti-Nogo45-66, levou à melhoria clínica e histológica da EAE em animais recipientes induzidos com outros antigénios de mielina, após migração destas células para o SNC. Estes dados comprovam a importância da Nogo-66 como antigénio na EAE, e a eficácia de terapias antigénio-específicas nela baseadas. No seu conjunto, os nossos resultados confirmam o potencial terapêutico das vacinas de ADN codificadoras de proteínas de mielina, bem como a importância dos encefalitogénios contidos na proteína Nogo-A para a fisiopatologia da EAE e da EM, com eventual relevância para o desenvolvimento de novas terapias antigénio-específicas. O aperfeiçoamento futuro destas terapias poderá levar, eventualmente, a uma capacidade de manipulação da resposta imune que permita o tratamento eficaz das doenças inflamatórias desmielinizantes, como a Esclerose Múltipla. ABSTRACT Multiple Sclerosis (MS) is a chronic demyelinating disease of the Central Nervous System (CNS), caused, mainly, by an immune-mediated attack against several elements of the myelin sheath. Among the antigenic targets for this autoimmune response, several proteic and lipidic myelin components have been identified throughout the years, of which myelin basic protein (MBP), myelin oligodendrocyte glycoprotein (MOG), proteolipidic protein (PLP), and myelin associated glycoprotein (MAG) are the best characterized. With the development of the animal model for MS, Experimental Autoimmune Encephalomyelitis (EAE), several antigen-specific therapies have been designed, based on beneficial modifications of the autoimmune response against myelin. These have included myelin and myelin component administration, therapeutic copolymers, altered peptide ligands and, more recently, vaccination with myelin-protein encoding deoxyribonucleic acid (DNA), integrated into plasmids and purified for parenteral administration. In this work we present the results of an extensive series of experiments, subordinate to two fundamental areas: 1) evaluating the therapeutic potential, and mechanisms of action, of tolerizing myelin protein (MBP, MOG, PLP, MAG) DNA vaccination in EAE, alone and in association with Th2 cytokine DNA administration, or immunomodulatory oligonucleotides; 2) identifying and characterizing the immuneresponse against a new myelin component with antigenic potential, the axonal regrowth inhibitor Nogo-A. Regarding DNA vaccination, our results prove the efficacy of this antigen-specific therapy for the prevention and treatment of EAE. Its mechanisms of action include, among others, anergic suppression of antigen-specific T-cell proliferation against myelin (in prevention mode), Th2 biasing of the immune response (when co-administered with the IL- 4 codifying DNA vaccine, acting as local gene therapy), and reduction of epitope spreading of the anti-myelin antibody response, assessed by myelin spotted arrays. The combination of myelin DNA vaccination with the administration of GpG immunomodulatory oligonucleotides, designed to counteract immunostimulatory CpG motifs present in the vaccination vector, led to an improvement in therapeutic efficacy, probably due to the preferential stimulatory effect of GpG oligonucleotides on Th2 lymphocytes and on regulatory NK-T cells. Based on these results, tolerizing DNA vaccination is being developed for human use, with ongoing clinical trials. As concerns the Nogo-A protein, based on studies of primary structure and prediction of antigenicity, we identified the Nogo-66 region (responsible for the most of the inhibitory capacity of this protein) as a potential antigenic target for EAE. In the SJL/Jand C57BL/6 mouse strains, we were able to induce clinical and histological signs of EAE,after immunization with the encefalitogenic epitopes Nogo1-22, Nogo23-44 and Nogo45-66,using a tolerance breakdown protocol. Concomitantly, we identified and characterized a specific T cell response against these antigens, together with a B cell response which showed extensive intra and intermolecular epitope spread to several determinants present in other myelin proteins. Adoptive transfer of nti-Nogo45-66 Th2 cell lines resulted in clinical and histological improvement of EAE in recipient animals induced with other myelin antigens, after intraparenchymal CNS migration of anti-Nogo cells. These data confirm the relevance of Nogo-66 as an antigen in EAE, as well as the efficacy of antigenspecific therapies based on the response against this protein.In conclusion, our results substantiate the therapeutic potential of myelin-encoding DNA vaccination, as well as the importance of encefalitogenic epitopes present in the Nogo-A protein for the pathophysiology of EAE and MS, with potential relevance for the creation of new antigen specific-therapies. The future development of these therapies may eventually lead to a degree of manipulation of the immune response that allows the effective treatment of autoimmune, inflammatory, demyelinating diseases, such as Multiple Sclerosis.