Algorithm-oriented design of efficient many-core architectures applied to dense matrix multiplication

Recent integrated circuit technologies have opened the possibility to design parallel architectures with hundreds of cores on a single chip. The design space of these parallel architectures is huge with many architectural options. Exploring the design space gets even more difficult if, beyond performance and area, we also consider extra metrics like performance and area efficiency, where the designer tries to design the architecture with the best performance per chip area and the best sustainable performance. In this paper we present an algorithm-oriented approach to design a many-core architecture. Instead of doing the design space exploration of the many core architecture based on the experimental execution results of a particular benchmark of algorithms, our approach is to make a formal analysis of the algorithms considering the main architectural aspects and to determine how each particular architectural aspect is related to the performance of the architecture when running an algorithm or set of algorithms. The architectural aspects considered include the number of cores, the local memory available in each core, the communication bandwidth between the many-core architecture and the external memory and the memory hierarchy. To exemplify the approach we did a theoretical analysis of a dense matrix multiplication algorithm and determined an equation that relates the number of execution cycles with the architectural parameters. Based on this equation a many-core architecture has been designed. The results obtained indicate that a 100 mm(2) integrated circuit design of the proposed architecture, using a 65 nm technology, is able to achieve 464 GFLOPs (double precision floating-point) for a memory bandwidth of 16 GB/s. This corresponds to a performance efficiency of 71 %. Considering a 45 nm technology, a 100 mm(2) chip attains 833 GFLOPs which corresponds to 84 % of peak performance These figures are better than those obtained by previous many-core architectures, except for the area efficiency which is limited by the lower memory bandwidth considered. The results achieved are also better than those of previous state-of-the-art many-cores architectures designed specifically to achieve high performance for matrix multiplication.

A 5GHz1.8V Low Power CMOS Low-Noise Amplifier

Wireless local-area networks (WLANs) have been deployed as office and home communications infrastructures worldwide. The diversification of the standards, such as IEEE 802.11 series demands the design of RF front-ends. Low power consumption is one of the most important design concerns in the application of those technologies. To maintain competitive hardware costs, CMOS has been used since it is the best solution for low cost and high integration processing, allowing analog circuits to be mixed with digital ones. In the receiver chain, the low noise amplifier (LNA) is one of the most critical blocks in a transceiver design. The sensitivity is mainly determined by the LNA noise figure and gain. It interfaces with the pre-select filter and the mixer. Furthermore, since it is the first gain stage, care must be taken to provide accurate input match, low-noise figure, good linearity and a sufficient gain over a wide band of operation. Several CMOS LNAs have been reported during the last decade, showing that the most research has been done at 802.11/b and GSM standards (900-2400MHz spectrum) and more recently at 802.11/a (5GHz band). One of the more significant disadvantages of 802.11/b is that the frequency band is crowded and subject to interference from other technologies, as is 2.4GHz cordless phones and Bluetooth. As the demand for radio-frequency integrated circuits, operating at higher frequency bands, increases, the IEEE 802.11/a standard becomes a very attractive option to wireless communication system developers. This paper presents the design and implementation of a low power, low noise amplifier aimed at IEEE 802.11a for WLAN applications. It was designed to be integrated with an active balun and mixer, representing the first step toward a fully integrated monolithic WLAN receiver. All the required circuits are integrated at the same die and are powered by 1.8V supply source. Preliminary experimental results (S-parameters) are shown and promise excellent results. The LNA circuit design details are illustrated in Section 2. Spectre simulation results focused at gain, noise figure (NF) and input/output matching are presented in Section 3. Finally, conclusions and comparison with other recently reported LNAs are made in Section 4, followed by future work.

Design of improved rail-to-rail low-distortion and low-stress switches in advanced CMOS technologies

This paper describes the efficient design of an improved and dedicated switched-capacitor (SC) circuit capable of linearizing CMOS switches to allow SC circuits to reach low distortion levels. The described circuit (SC linearization control circuit, SLC) has the advantage over conventional clock-bootstrapping circuits of exhibiting low-stress, since large gate voltages are avoided. This paper presents exhaustive corner simulation results of a SC sample-and-hold (S/H) circuit which employs the proposed and optimized circuits, together with the experimental evaluation of a complete 10-bit ADC utilizing the referred S/H circuit. These results show that the SLC circuits can reduce distortion and increase dynamic linearity above 12 bits for wide input signal bandwidths.

Projecto de um conversor redutor CC-CC comutado para SoC

A evolução da tecnologia CMOS tem possibilitado uma maior densidade de integração de circuitos tornando possível o aumento da complexidade dos sistemas. No entanto, a integração de circuitos de gestão de potência continua ainda em estudo devido à dificuldade de integrar todos os componentes. Esta solução apresenta elevadas vantagens, especialmente em aplicações electrónicas portáteis alimentadas a baterias, onde a autonomia é das principais características. No âmbito dos conversores redutores existem várias topologias de circuitos que são estudadas na área de integração. Na categoria dos conversores lineares utiliza-se o LDO (Low Dropout Regulator), apresentando no entanto baixa eficiência para relações de conversão elevadas. Os conversores comutados são elaborados através do recurso a circuitos de comutação abrupta, em que a eficiência deste tipo de conversores não depende do rácio de transformação entre a tensão de entrada e a de saída. A diminuição física dos processos CMOS tem como consequência a redução da tensão máxima que os transístores suportam, impondo o estudo de soluções tolerantes a “altatensão”, com o intuito de manter compatibilidade com tensões superiores que existam na placa onde o circuito é incluído. Os sistemas de gestão de energia são os primeiros a acompanhar esta evolução, tendo de estar aptos a fornecer a tensão que os restantes circuitos requerem. Neste trabalho é abordada uma metodologia de projecto para conversores redutores CCCC comutados em tecnologia CMOS, tendo-se maximizado a frequência com vista à integração dos componentes de filtragem em circuito integrado. A metodologia incide sobre a optimização das perdas totais inerentes à comutação e condução, dos transístores de potência e respectivos circuitos auxiliares. É apresentada uma nova metodologia para o desenvolvimento de conversores tolerantes a “alta-tensão”.

Codificador JPEG baseado em FPGA

O presente trabalho consiste na implementação em hardware de unidades funcionais dedicadas e optimizadas, para a realização das operações de codificação e descodificação, definidas na norma de codificação com perda Joint Photographic Experts Group (JPEG), ITU-T T.81 ISO/IEC 10918-1. Realiza-se um estudo sobre esta norma de forma a caracterizar os seus principais blocos funcionais. A finalidade deste estudo foca-se na pesquisa e na proposta de optimizações, de forma a minimizar o hardware necessário para a realização de cada bloco, de modo a que o sistema realizado obtenha taxas de compressão elevadas, minimizando a distorção obtida. A redução de hardware de cada sistema, codificador e descodificador, é conseguida à custa da manipulação das equações dos blocos Forward Discrete Cosine Transform (FDCT) e Quantificação (Q) e dos blocos Forward Discrete Cosine Transform (IDCT) e Quantificação Inversa (IQ). Com as conclusões retiradas do estudo e através da análise de estruturas conhecidas, descreveu-se cada bloco em Very-High-Speed Integrated Circuits (VHSIC) Hardware Description Language (VHDL) e fez-se a sua síntese em Field Programmable Gate Array (FPGA). Cada sistema implementado recorre à execução de cada bloco em paralelo de forma a optimizar a codificação/descodificação. Assim, para o sistema codificador, será realizada a operação da FDCT e Quantificação sobre duas matrizes diferentes e em simultâneo. O mesmo sucede para o sistema descodificador, composto pelos blocos Quantificação Inversa e IDCT. A validação de cada bloco sintetizado é executada com recurso a vectores de teste obtidos através do estudo efectuado. Após a integração de cada bloco, verificou-se que, para imagens greyscale de referência com resolução de 256 linhas por 256 colunas, é necessário 820,5 μs para a codificação de uma imagem e 830,5 μs para a descodificação da mesma. Considerando uma frequência de trabalho de 100 MHz, processam-se aproximadamente 1200 imagens por segundo.

A Voltage Limiter Circuit for Indoor Light Energy Harvesting Applications

A voltage limiter circuit for indoor light energy harvesting applications is presented. This circuit is a part of a bigger system, whose function is to harvest indoor light energy, process it and store it, so that it can be used at a later time. This processing consists on maximum power point tracking (MPPT) and stepping-up, of the voltage from the photovoltaic (PV) harvester cell. The circuit here described, ensures that even under strong illumination, the generated voltage will not exceed the limit allowed by the technology, avoiding the degradation, or destruction, of the integrated die. A prototype of the limiter circuit was designed in a 130 nm CMOS technology. The layout of the circuit has a total area of 23414 mu m(2). Simulation results, using Spectre, are presented.

A Compact and Scalable RNS Architecture

Conferência: IEEE 24th International Conference on Application-Specific Systems, Architectures and Processors (ASAP)- Jun 05-07, 2013

Sistema de reconhecimento de impressões digitais baseado em FPGA

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

Melhoria da eficiência de amplificadores de potência para sinais RF com envolvente variável

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

A CMOS micro power switched-capacitor DC-DC step-up converter for indoor light energy harvesting applications

This paper presents a micro power light energy harvesting system for indoor environments. Light energy is collected by amorphous silicon photovoltaic (a-Si:H PV) cells, processed by a switched capacitor (SC) voltage doubler circuit with maximum power point tracking (MPPT), and finally stored in a large capacitor. The MPPT fractional open circuit voltage (V-OC) technique is implemented by an asynchronous state machine (ASM) that creates and dynamically adjusts the clock frequency of the step-up SC circuit, matching the input impedance of the SC circuit to the maximum power point condition of the PV cells. The ASM has a separate local power supply to make it robust against load variations. In order to reduce the area occupied by the SC circuit, while maintaining an acceptable efficiency value, the SC circuit uses MOSFET capacitors with a charge sharing scheme for the bottom plate parasitic capacitors. The circuit occupies an area of 0.31 mm(2) in a 130 nm CMOS technology. The system was designed in order to work under realistic indoor light intensities. Experimental results show that the proposed system, using PV cells with an area of 14 cm(2), is capable of starting-up from a 0 V condition, with an irradiance of only 0.32 W/m(2). After starting-up, the system requires an irradiance of only 0.18 W/m(2) (18 mu W/cm(2)) to remain operating. The ASM circuit can operate correctly using a local power supply voltage of 453 mV, dissipating only 0.085 mu W. These values are, to the best of the authors' knowledge, the lowest reported in the literature. The maximum efficiency of the SC converter is 70.3 % for an input power of 48 mu W, which is comparable with reported values from circuits operating at similar power levels.

ROM-less RNS-to-binary converter moduli {22N − 1, 22N + 1, 2N − 3, 2N + 3}

In this paper, a novel ROM-less RNS-to-binary converter is proposed, using a new balanced moduli set {22n-1, 22n + 1, 2n-3, 2n + 3} for n even. The proposed converter is implemented with a two stage ROM-less approach, which computes the value of X based only in arithmetic operations, without using lookup tables. Experimental results for 24 to 120 bits of Dynamic Range, show that the proposed converter structure allows a balanced system with 20% faster arithmetic channels regarding the related state of the art, while requiring similar area resources. This improvement in the channel's performance is enough to offset the higher conversion costs of the proposed converter. Furthermore, up to 20% better Power-Delay-Product efficiency metric can be achieved for the full RNS architecture using the proposed moduli set. © 2014 IEEE.

CMOS indoor light energy harvesting system for wireless sensing applications

This book discusses in detail the CMOS implementation of energy harvesting. The authors describe an integrated, indoor light energy harvesting system, based on a controller circuit that dynamically and automatically adjusts its operation to meet the actual light circumstances of the environment where the system is placed. The system is intended to power a sensor node, enabling an autonomous wireless sensor network (WSN). Although designed to cope with indoor light levels, the system is also able to work with higher levels, making it an all-round light energy harvesting system. The discussion includes experimental data obtained from an integrated manufactured prototype, which in conjunction with a photovoltaic (PV) cell, serves as a proof of concept of the desired energy harvesting system. © 2016 Springer International Publishing. All rights are reserved.

A Service Integration Platform for Collaborative Networks

Integrated manufacturing constitutes a complex system made of heterogeneous information and control subsystems. Those subsystems are not designed to the cooperation. Typically each subsystem automates specific processes, and establishes closed application domains, therefore it is very difficult to integrate it with other subsystems in order to respond to the needed process dynamics. Furthermore, to cope with ever growing marketcompetition and demands, it is necessary for manufacturing/enterprise systems to increase their responsiveness based on up-to-date knowledge and in-time data gathered from the diverse information and control systems. These have created new challenges for manufacturing sector, and even bigger challenges for collaborative manufacturing. The growing complexity of the information and communication technologies when coping with innovative business services based on collaborative contributions from multiple stakeholders, requires novel and multidisciplinary approaches. Service orientation is a strategic approach to deal with such complexity, and various stakeholders' information systems. Services or more precisely the autonomous computational agents implementing the services, provide an architectural pattern able to cope with the needs of integrated and distributed collaborative solutions. This paper proposes a service-oriented framework, aiming to support a virtual organizations breeding environment that is the basis for establishing short or long term goal-oriented virtual organizations. The notion of integrated business services, where customers receive some value developed through the contribution from a network of companies is a key element.

Photodetector with integrated optical thin film filters

This paper reports on optical filters based on a-SiC:H tandem pi'n/pin heterostructures. The spectral sensitivity is analyzed. Steady state optical bias with different wavelengths are applied from each front and back sides and the photocurrent is measured. Results show that it is possible to control the sensitivity of the device and to tune a specific wavelength range by combining radiations with complementary light penetration depths. The transfer characteristics effects due to changes in the front and back optical bias wavelength are discussed. Depending on the wavelength of the external background and irradiation side, the device acts either as a short- or a long-pass band filter or as a band-stop filter. The output waveform presents a nonlinear amplitude-dependent response to the wavelengths of the input channels.

Scalable Unified Transform Architecture for Advanced Video Coding Embedded Systems

A novel high throughput and scalable unified architecture for the computation of the transform operations in video codecs for advanced standards is presented in this paper. This structure can be used as a hardware accelerator in modern embedded systems to efficiently compute all the two-dimensional 4 x 4 and 2 x 2 transforms of the H.264/AVC standard. Moreover, its highly flexible design and hardware efficiency allows it to be easily scaled in terms of performance and hardware cost to meet the specific requirements of any given video coding application. Experimental results obtained using a Xilinx Virtex-5 FPGA demonstrated the superior performance and hardware efficiency levels provided by the proposed structure, which presents a throughput per unit of area relatively higher than other similar recently published designs targeting the H.264/AVC standard. Such results also showed that, when integrated in a multi-core embedded system, this architecture provides speedup factors of about 120x concerning pure software implementations of the transform algorithms, therefore allowing the computation, in real-time, of all the above mentioned transforms for Ultra High Definition Video (UHDV) sequences (4,320 x 7,680 @ 30 fps).