Kernel density correlation based non-rigid point set matching for statistical model-based 2D/3D reconstruction

This paper presents a kernel density correlation based nonrigid point set matching method and shows its application in statistical model based 2D/3D reconstruction of a scaled, patient-specific model from an un-calibrated x-ray radiograph. In this method, both the reference point set and the floating point set are first represented using kernel density estimates. A correlation measure between these two kernel density estimates is then optimized to find a displacement field such that the floating point set is moved to the reference point set. Regularizations based on the overall deformation energy and the motion smoothness energy are used to constraint the displacement field for a robust point set matching. Incorporating this non-rigid point set matching method into a statistical model based 2D/3D reconstruction framework, we can reconstruct a scaled, patient-specific model from noisy edge points that are extracted directly from the x-ray radiograph by an edge detector. Our experiment conducted on datasets of two patients and six cadavers demonstrates a mean reconstruction error of 1.9 mm

Design and Validation of an Optimized MB-OFDM Ultra Wideband Transceiver System

Esta tesis está incluida dentro del campo del campo de Multiband Orthogonal Frequency Division Multiplexing Ultra Wideband (MB-OFDM UWB), el cual ha adquirido una gran importancia en las comunicaciones inalámbricas de alta tasa de datos en la última década. UWB surgió con el objetivo de satisfacer la creciente demanda de conexiones inalámbricas en interiores y de uso doméstico, con bajo coste y alta velocidad. La disponibilidad de un ancho de banda grande, el potencial para alta velocidad de transmisión, baja complejidad y bajo consumo de energía, unido al bajo coste de implementación, representa una oportunidad única para que UWB se convierta en una solución ampliamente utilizada en aplicaciones de Wireless Personal Area Network (WPAN). UWB está definido como cualquier transmisión que ocupa un ancho de banda de más de 20% de su frecuencia central, o más de 500 MHz. En 2002, la Comisión Federal de Comunicaciones (FCC) definió que el rango de frecuencias de transmisión de UWB legal es de 3.1 a 10.6 GHz, con una energía de transmisión de -41.3 dBm/Hz. Bajo las directrices de FCC, el uso de la tecnología UWB puede aportar una enorme capacidad en las comunicaciones de corto alcance. Considerando las ecuaciones de capacidad de Shannon, incrementar la capacidad del canal requiere un incremento lineal en el ancho de banda, mientras que un aumento similar de la capacidad de canal requiere un aumento exponencial en la energía de transmisión. En los últimos años, s diferentes desarrollos del UWB han sido extensamente estudiados en diferentes áreas, entre los cuales, el protocolo de comunicaciones inalámbricas MB-OFDM UWB está considerado como la mejor elección y ha sido adoptado como estándar ISO/IEC para los WPANs. Combinando la modulación OFDM y la transmisión de datos utilizando las técnicas de salto de frecuencia, el sistema MB-OFDM UWB es capaz de soportar tasas de datos con que pueden variar de los 55 a los 480 Mbps, alcanzando una distancia máxima de hasta 10 metros. Se esperara que la tecnología MB-OFDM tenga un consumo energético muy bajo copando un are muy reducida en silicio, proporcionando soluciones de bajo coste que satisfagan las demandas del mercado. Para cumplir con todas estas expectativas, el desarrollo y la investigación del MBOFDM UWB deben enfrentarse a varios retos, como son la sincronización de alta sensibilidad, las restricciones de baja complejidad, las estrictas limitaciones energéticas, la escalabilidad y la flexibilidad. Tales retos requieren un procesamiento digital de la señal de última generación, capaz de desarrollar sistemas que puedan aprovechar por completo las ventajas del espectro UWB y proporcionar futuras aplicaciones inalámbricas en interiores. Esta tesis se centra en la completa optimización de un sistema de transceptor de banda base MB-OFDM UWB digital, cuyo objetivo es investigar y diseñar un subsistema de comunicación inalámbrica para la aplicación de las Redes de Sensores Inalámbricas Visuales. La complejidad inherente de los procesadores FFT/IFFT y el sistema de sincronización así como la alta frecuencia de operación para todos los elementos de procesamiento, se convierten en el cuello de la botella para el diseño y la implementación del sistema de UWB digital en base de banda basado en MB-OFDM de baja energía. El objetivo del transceptor propuesto es conseguir baja energía y baja complejidad bajo la premisa de un alto rendimiento. Las optimizaciones están realizadas tanto a nivel algorítmico como a nivel arquitectural para todos los elementos del sistema. Una arquitectura hardware eficiente en consumo se propone en primer lugar para aquellos módulos correspondientes a núcleos de computación. Para el procesado de la Transformada Rápida de Fourier (FFT/IFFT), se propone un algoritmo mixed-radix, basado en una arquitectura con pipeline y se ha desarrollado un módulo de Decodificador de Viterbi (VD) equilibrado en coste-velocidad con el objetivo de reducir el consumo energético e incrementar la velocidad de procesamiento. También se ha implementado un correlador signo-bit simple basado en la sincronización del tiempo de símbolo es presentado. Este correlador es usado para detectar y sincronizar los paquetes de OFDM de forma robusta y precisa. Para el desarrollo de los subsitemas de procesamiento y realizar la integración del sistema completo se han empleado tecnologías de última generación. El dispositivo utilizado para el sistema propuesto es una FPGA Virtex 5 XC5VLX110T del fabricante Xilinx. La validación el propuesta para el sistema transceptor se ha implementado en dicha placa de FPGA. En este trabajo se presenta un algoritmo, y una arquitectura, diseñado con filosofía de co-diseño hardware/software para el desarrollo de sistemas de FPGA complejos. El objetivo principal de la estrategia propuesta es de encontrar una metodología eficiente para el diseño de un sistema de FPGA configurable optimizado con el empleo del mínimo esfuerzo posible en el sistema de procedimiento de verificación, por tanto acelerar el periodo de desarrollo del sistema. La metodología de co-diseño presentada tiene la ventaja de ser fácil de usar, contiene todos los pasos desde la propuesta del algoritmo hasta la verificación del hardware, y puede ser ampliamente extendida para casi todos los tipos de desarrollos de FPGAs. En este trabajo se ha desarrollado sólo el sistema de transceptor digital de banda base por lo que la comprobación de señales transmitidas a través del canal inalámbrico en los entornos reales de comunicación sigue requiriendo componentes RF y un front-end analógico. No obstante, utilizando la metodología de co-simulación hardware/software citada anteriormente, es posible comunicar el sistema de transmisor y el receptor digital utilizando los modelos de canales propuestos por IEEE 802.15.3a, implementados en MATLAB. Por tanto, simplemente ajustando las características de cada modelo de canal, por ejemplo, un incremento del retraso y de la frecuencia central, podemos estimar el comportamiento del sistema propuesto en diferentes escenarios y entornos. Las mayores contribuciones de esta tesis son: • Se ha propuesto un nuevo algoritmo 128-puntos base mixto FFT usando la arquitectura pipeline multi-ruta. Los complejos multiplicadores para cada etapa de procesamiento son diseñados usando la arquitectura modificada shiftadd. Los sistemas word length y twiddle word length son comparados y seleccionados basándose en la señal para cuantización del SQNR y el análisis de energías. • El desempeño del procesador IFFT es analizado bajo diferentes situaciones aritméticas de bloques de punto flotante (BFP) para el control de desbordamiento, por tanto, para encontrar la arquitectura perfecta del algoritmo IFFT basado en el procesador FFT propuesto. • Para el sistema de receptor MB-OFDM UWB se ha empleado una sincronización del tiempo innovadora, de baja complejidad y esquema de compensación, que consiste en funciones de Detector de Paquetes (PD) y Estimación del Offset del tiempo. Simplificando el cross-correlation y maximizar las funciones probables solo a sign-bit, la complejidad computacional se ve reducida significativamente. • Se ha propuesto un sistema de decodificadores Viterbi de 64 estados de decisión-débil usando velocidad base-4 de arquitectura suma-comparaselecciona. El algoritmo Two-pointer Even también es introducido en la unidad de rastreador de origen con el objetivo de conseguir la eficiencia en el hardware. • Se han integrado varias tecnologías de última generación en el completo sistema transceptor basebanda , con el objetivo de implementar un sistema de comunicación UWB altamente optimizado. • Un diseño de flujo mejorado es propuesto para el complejo sistema de implementación, el cual puede ser usado para diseños de Cadena de puertas de campo programable general (FPGA). El diseño mencionado no sólo reduce dramáticamente el tiempo para la verificación funcional, sino también provee un análisis automático como los errores del retraso del output para el sistema de hardware implementado. • Un ambiente de comunicación virtual es establecido para la validación del propuesto sistema de transceptores MB-OFDM. Este método es provisto para facilitar el uso y la conveniencia de analizar el sistema digital de basebanda sin parte frontera analógica bajo diferentes ambientes de comunicación. Esta tesis doctoral está organizada en seis capítulos. En el primer capítulo se encuentra una breve introducción al campo del UWB, tanto relacionado con el proyecto como la motivación del desarrollo del sistema de MB-OFDM. En el capítulo 2, se presenta la información general y los requisitos del protocolo de comunicación inalámbrica MBOFDM UWB. En el capítulo 3 se habla de la arquitectura del sistema de transceptor digital MB-OFDM de banda base . El diseño del algoritmo propuesto y la arquitectura para cada elemento del procesamiento está detallado en este capítulo. Los retos de diseño del sistema que involucra un compromiso de discusión entre la complejidad de diseño, el consumo de energía, el coste de hardware, el desempeño del sistema, y otros aspectos. En el capítulo 4, se ha descrito la co-diseñada metodología de hardware/software. Cada parte del flujo del diseño será detallado con algunos ejemplos que se ha hecho durante el desarrollo del sistema. Aprovechando esta estrategia de diseño, el procedimiento de comunicación virtual es llevado a cabo para probar y analizar la arquitectura del transceptor propuesto. Los resultados experimentales de la co-simulación y el informe sintético de la implementación del sistema FPGA son reflejados en el capítulo 5. Finalmente, en el capítulo 6 se incluye las conclusiones y los futuros proyectos, y también los resultados derivados de este proyecto de doctorado. ABSTRACT In recent years, the Wireless Visual Sensor Network (WVSN) has drawn great interest in wireless communication research area. They enable a wealth of new applications such as building security control, image sensing, and target localization. However, nowadays wireless communication protocols (ZigBee, Wi-Fi, and Bluetooth for example) cannot fully satisfy the demands of high data rate, low power consumption, short range, and high robustness requirements. New communication protocol is highly desired for such kind of applications. The Ultra Wideband (UWB) wireless communication protocol, which has increased in importance for high data rate wireless communication field, are emerging as an important topic for WVSN research. UWB has emerged as a technology that offers great promise to satisfy the growing demand for low-cost, high-speed digital wireless indoor and home networks. The large bandwidth available, the potential for high data rate transmission, and the potential for low complexity and low power consumption, along with low implementation cost, all present a unique opportunity for UWB to become a widely adopted radio solution for future Wireless Personal Area Network (WPAN) applications. UWB is defined as any transmission that occupies a bandwidth of more than 20% of its center frequency, or more than 500 MHz. In 2002, the Federal Communications Commission (FCC) has mandated that UWB radio transmission can legally operate in the range from 3.1 to 10.6 GHz at a transmitter power of －41.3 dBm/Hz. Under the FCC guidelines, the use of UWB technology can provide enormous capacity over short communication ranges. Considering Shannon’s capacity equations, increasing the channel capacity requires linear increasing in bandwidth, whereas similar channel capacity increases would require exponential increases in transmission power. In recent years, several different UWB developments has been widely studied in different area, among which, the MB-OFDM UWB wireless communication protocol is considered to be the leading choice and has recently been adopted in the ISO/IEC standard for WPANs. By combing the OFDM modulation and data transmission using frequency hopping techniques, the MB-OFDM UWB system is able to support various data rates, ranging from 55 to 480 Mbps, over distances up to 10 meters. The MB-OFDM technology is expected to consume very little power and silicon area, as well as provide low-cost solutions that can satisfy consumer market demands. To fulfill these expectations, MB-OFDM UWB research and development have to cope with several challenges, which consist of high-sensitivity synchronization, low- complexity constraints, strict power limitations, scalability, and flexibility. Such challenges require state-of-the-art digital signal processing expertise to develop systems that could fully take advantages of the UWB spectrum and support future indoor wireless applications. This thesis focuses on fully optimization for the MB-OFDM UWB digital baseband transceiver system, aiming at researching and designing a wireless communication subsystem for the Wireless Visual Sensor Networks (WVSNs) application. The inherent high complexity of the FFT/IFFT processor and synchronization system, and high operation frequency for all processing elements, becomes the bottleneck for low power MB-OFDM based UWB digital baseband system hardware design and implementation. The proposed transceiver system targets low power and low complexity under the premise of high performance. Optimizations are made at both algorithm and architecture level for each element of the transceiver system. The low-power hardwareefficient structures are firstly proposed for those core computation modules, i.e., the mixed-radix algorithm based pipelined architecture is proposed for the Fast Fourier Transform (FFT/IFFT) processor, and the cost-speed balanced Viterbi Decoder (VD) module is developed, in the aim of lowering the power consumption and increasing the processing speed. In addition, a low complexity sign-bit correlation based symbol timing synchronization scheme is presented so as to detect and synchronize the OFDM packets robustly and accurately. Moreover, several state-of-the-art technologies are used for developing other processing subsystems and an entire MB-OFDM digital baseband transceiver system is integrated. The target device for the proposed transceiver system is Xilinx Virtex 5 XC5VLX110T FPGA board. In order to validate the proposed transceiver system in the FPGA board, a unified algorithm-architecture-circuit hardware/software co-design environment for complex FPGA system development is presented in this work. The main objective of the proposed strategy is to find an efficient methodology for designing a configurable optimized FPGA system by using as few efforts as possible in system verification procedure, so as to speed up the system development period. The presented co-design methodology has the advantages of easy to use, covering all steps from algorithm proposal to hardware verification, and widely spread for almost all kinds of FPGA developments. Because only the digital baseband transceiver system is developed in this thesis, the validation of transmitting signals through wireless channel in real communication environments still requires the analog front-end and RF components. However, by using the aforementioned hardware/software co-simulation methodology, the transmitter and receiver digital baseband systems get the opportunity to communicate with each other through the channel models, which are proposed from the IEEE 802.15.3a research group, established in MATLAB. Thus, by simply adjust the characteristics of each channel model, e.g. mean excess delay and center frequency, we can estimate the transmission performance of the proposed transceiver system through different communication situations. The main contributions of this thesis are: • A novel mixed radix 128-point FFT algorithm by using multipath pipelined architecture is proposed. The complex multipliers for each processing stage are designed by using modified shift-add architectures. The system wordlength and twiddle word-length are compared and selected based on Signal to Quantization Noise Ratio (SQNR) and power analysis. • IFFT processor performance is analyzed under different Block Floating Point (BFP) arithmetic situations for overflow control, so as to find out the perfect architecture of IFFT algorithm based on the proposed FFT processor. • An innovative low complex timing synchronization and compensation scheme, which consists of Packet Detector (PD) and Timing Offset Estimation (TOE) functions, for MB-OFDM UWB receiver system is employed. By simplifying the cross-correlation and maximum likelihood functions to signbit only, the computational complexity is significantly reduced. • A 64 state soft-decision Viterbi Decoder system by using high speed radix-4 Add-Compare-Select architecture is proposed. Two-pointer Even algorithm is also introduced into the Trace Back unit in the aim of hardware-efficiency. • Several state-of-the-art technologies are integrated into the complete baseband transceiver system, in the aim of implementing a highly-optimized UWB communication system. • An improved design flow is proposed for complex system implementation which can be used for general Field-Programmable Gate Array (FPGA) designs. The design method not only dramatically reduces the time for functional verification, but also provides automatic analysis such as errors and output delays for the implemented hardware systems. • A virtual communication environment is established for validating the proposed MB-OFDM transceiver system. This methodology is proved to be easy for usage and convenient for analyzing the digital baseband system without analog frontend under different communication environments. This PhD thesis is organized in six chapters. In the chapter 1 a brief introduction to the UWB field, as well as the related work, is done, along with the motivation of MBOFDM system development. In the chapter 2, the general information and requirement of MB-OFDM UWB wireless communication protocol is presented. In the chapter 3, the architecture of the MB-OFDM digital baseband transceiver system is presented. The design of the proposed algorithm and architecture for each processing element is detailed in this chapter. Design challenges of such system involve trade-off discussions among design complexity, power consumption, hardware cost, system performance, and some other aspects. All these factors are analyzed and discussed. In the chapter 4, the hardware/software co-design methodology is proposed. Each step of this design flow will be detailed by taking some examples that we met during system development. Then, taking advantages of this design strategy, the Virtual Communication procedure is carried out so as to test and analyze the proposed transceiver architecture. Experimental results from the co-simulation and synthesis report of the implemented FPGA system are given in the chapter 5. The chapter 6 includes conclusions and future work, as well as the results derived from this PhD work.

Time complexity and convergence analysis of domain theoretic Picard method

We present an implementation of the domain-theoretic Picard method for solving initial value problems (IVPs) introduced by Edalat and Pattinson [1]. Compared to Edalat and Pattinson's implementation, our algorithm uses a more efficient arithmetic based on an arbitrary precision floating-point library. Despite the additional overestimations due to floating-point rounding, we obtain a similar bound on the convergence rate of the produced approximations. Moreover, our convergence analysis is detailed enough to allow a static optimisation in the growth of the precision used in successive Picard iterations. Such optimisation greatly improves the efficiency of the solving process. Although a similar optimisation could be performed dynamically without our analysis, a static one gives us a significant advantage: we are able to predict the time it will take the solver to obtain an approximation of a certain (arbitrarily high) quality.

Digit recurence division under HUB format

Half-Unit-Biased format is based on shifting the representation line of the binary numbers by half Unit in the Last Place. The main feature of this format is that the round to nearest is carried out by a simple truncation, preventing any carry propagation and saving time and area. Algorithms and architectures have been defined for addition/substraction and multiplication operations under this format. Nevertheless, the division operation has not been confronted yet. In this paper we deal with the floating-point division under HUB format, studying the architecture for the digit recurrence method, including the on-the-fly conversion of the signed digit quotient.

Optimal design of piezolaminated structures using B-spline strip finite element models

A package of B-spline finite strip models is developed for the linear analysis of piezolaminated plates and shells. This package is associated to a global optimization technique in order to enhance the performance of these types of structures, subjected to various types of objective functions and/or constraints, with discrete and continuous design variables. The models considered are based on a higher-order displacement field and one can apply them to the static, free vibration and buckling analyses of laminated adaptive structures with arbitrary lay-ups, loading and boundary conditions. Genetic algorithms, with either binary or floating point encoding of design variables, were considered to find optimal locations of piezoelectric actuators as well as to determine the best voltages applied to them in order to obtain a desired structure shape. These models provide an overall economy of computing effort for static and vibration problems.

Receptor MIMO em FPGA baseado no esquema de Alamouti

A oferta de serviços baseados em comunicações sem fios tem vindo a crescer exponencialmente na última década. Cada vez mais são exigidas maiores taxas de transmissão assim como uma melhor QoS, sem comprometer a potência de transmissão ou argura de banda disponível. A tecnologia MIMO consegue oferecer um aumento da capacidade destes sistemas sem requerer aumento da largura de banda ou da potência transmitida. O trabalho desenvolvido nesta dissertação consistiu no estudo dos sistemas MIMO, caracterizados pela utilização de múltiplas antenas para transmitir e receber a informação. Com um sistema deste tipo consegue-se obter um ganho de diversidade espacial utilizando códigos espaço-temporais, que exploram simultaneamente o domínio espacial e o domínio do tempo. Nesta dissertação é dado especial ênfase à codificação por blocos no espaço-tempo de Alamouti, a qual será implementada em FPGA, nomeadamente a parte de recepção. Esta implementação é efectuada para uma configuração de antenas 2x1, utilizando vírgula flutuante e para três tipos de modulação: BPSK, QPSK e 16-QAM. Por fim será analisada a relação entre a precisão alcançada na representação numérica dos resultados e os recursos consumidos pela FPGA. Com a arquitectura adoptada conseguem se obter taxas de transferência na ordem dos 29,141 Msimb/s (sem pipelines) a 262,674 Msimb/s (com pipelines), para a modulação BPSK.

An Efficient Long Distance Echo Canceller

This paper describes an implementation of a long distance echo canceller, operating on full-duplex with hands-free and in real-time with a single Digital Signal Processor (DSP). The proposed solution is based on short length adaptive filters centered on the positions of the most significant echoes, which are tracked by time delay estimators, for which we use a new approach. To deal with double talking situations a speech detector is employed. The floating-point DSP TMS320C6713 from Texas Instruments is used with software written in C++, with compiler optimizations for fast execution. The resulting algorithm enables long distance echo cancellation with low computational requirements, suited for embbeded systems. It reaches greater echo return loss enhancement and shows faster convergence speed when compared to the conventional approach. The experimental results approach the CCITT G.165 recommendation levels.

Implementação em hardware reconfigurável de método de separação de dados hiperespetrais

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

Sistema de reconhecimento de impressões digitais baseado em FPGA

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

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).

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 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.

Parallel GPU architecture for hyperspectral unmixing based on augmented Lagrangian method

Hyperspectral imaging has become one of the main topics in remote sensing applications, which comprise hundreds of spectral bands at different (almost contiguous) wavelength channels over the same area generating large data volumes comprising several GBs per flight. This high spectral resolution can be used for object detection and for discriminate between different objects based on their spectral characteristics. One of the main problems involved in hyperspectral analysis is the presence of mixed pixels, which arise when the spacial resolution of the sensor is not able to separate spectrally distinct materials. Spectral unmixing is one of the most important task for hyperspectral data exploitation. However, the unmixing algorithms can be computationally very expensive, and even high power consuming, which compromises the use in applications under on-board constraints. In recent years, graphics processing units (GPUs) have evolved into highly parallel and programmable systems. Specifically, several hyperspectral imaging algorithms have shown to be able to benefit from this hardware taking advantage of the extremely high floating-point processing performance, compact size, huge memory bandwidth, and relatively low cost of these units, which make them appealing for onboard data processing. In this paper, we propose a parallel implementation of an augmented Lagragian based method for unsupervised hyperspectral linear unmixing on GPUs using CUDA. The method called simplex identification via split augmented Lagrangian (SISAL) aims to identify the endmembers of a scene, i.e., is able to unmix hyperspectral data sets in which the pure pixel assumption is violated. The efficient implementation of SISAL method presented in this work exploits the GPU architecture at low level, using shared memory and coalesced accesses to memory.

A Grid Supercomputing Enviroment for High Demand Computational Applications

Despite the huge increase in processor and interprocessor network performace, many computational problems remain unsolved due to lack of some critical resources such as floating point sustained performance, memory bandwidth, etc... Examples of these problems are found in areas of climate research, biology, astrophysics, high energy physics (montecarlo simulations) and artificial intelligence, among others. For some of these problems, computing resources of a single supercomputing facility can be 1 or 2 orders of magnitude apart from the resources needed to solve some them. Supercomputer centers have to face an increasing demand on processing performance, with the direct consequence of an increasing number of processors and systems, resulting in a more difficult administration of HPC resources and the need for more physical space, higher electrical power consumption and improved air conditioning, among other problems. Some of the previous problems can´t be easily solved, so grid computing, intended as a technology enabling the addition and consolidation of computing power, can help in solving large scale supercomputing problems. In this document, we describe how 2 supercomputing facilities in Spain joined their resources to solve a problem of this kind. The objectives of this experience were, among others, to demonstrate that such a cooperation can enable the solution of bigger dimension problems and to measure the efficiency that could be achieved. In this document we show some preliminary results of this experience and to what extend these objectives were achieved.

Estudi de comparació de rendiment de Benchmark multicore i multithreading amb OpenMP 2.5 i OpenMP 3.0

Estudi comparatiu amb benchmark del rendiment en dues plataformes multicore multithreading de diferents modalitats de paral·lelització de multiplicacions de matrius de nombres enters i de nombres en coma flotant mitjançant el model de memòria compartida OpenMP versió 2.5 i OpenMP versió 3.0.