Un énorme progrès dans les performances des semiconducteurs a été accompli ces dernières années. Avec l’´émergence d’applications complexes, les systèmes embarqués doivent être à la fois performants et fiables. Une multitude de travaux ont été proposés pour améliorer l’efficacité des systèmes embarqués en réduisant le décalage entre la flexibilité des solutions logicielles et la haute performance des solutions matérielles. En vertu de leur nature reconfigurable, les FPGAs (Field Programmable Gate Arrays) représentent un pas considérable pour réduire ce décalage performance/flexibilité. Cependant, la reconfiguration dynamique a toujours souffert d’une limitation liée à la latence de reconfiguration.Dans cette thèse, une nouvelle technique de reconfiguration dynamiqueau niveau ”grain-moyen” pour les circuits à base de blocks DSP48E1 est proposée. L’idée est de profiter de la reprogrammabilité des blocks DSP48E1 couplée avec un circuit d’interconnection reconfigurable afin de changer la fonction implémentée par le circuit en un cycle horloge. D’autre part, comme les nouvelles technologies s’appuient sur la réduction des dimensions des transistors ainsi que les tensions d’alimentation, les circuits électroniques sont devenus de plus en plus susceptibles aux fautes transitoires. L’impact de ces erreurs au niveau système peut être catastrophique et les SETs (Single Event Transients) sont devenus une menace tangible à la fiabilité des systèmes embarqués, en l’occurrence pour les applications critiques comme les systèmes de transport. Les techniques de fiabilité qui se basent sur des taux d’erreurs (SERs) surestimés peuvent conduire à un gaspillage de ressources et par conséquent un cout en consommation de puissance électrique. Il est primordial de prendre en compte le phénomène de masquage d’erreur pour une estimation précise des SERs.Cette thèse propose une nouvelle modélisation inter-couches de la vulnérabilité des circuits qui combine les mécanismes de masquage au niveau transistor (TLM) et le masquage au niveau Système (SLM). Ce modèle est ensuite utilisé afin de construire une architecture adaptative tolérante aux fautes qui évalue la vulnérabilité effective du circuit en runtime. La stratégie d’amélioration de fiabilité est adaptée pour ne protéger que les parties vulnérables du système, ce qui engendre un circuit fiable avec un cout optimisé. Les expérimentations effectuées sur un système de détection d’obstacles à base de radar pour le transport ferroviaire montre que l’approche proposée permet d’´établir un compromis fiabilité/ressources utilisées. / During the last few decades, a tremendous progress in the performance of semiconductor devices has been accomplished. In this emerging era of high performance applications, machines need not only to be efficient but also need to be dependable at circuit and system levels. Several works have been proposed to increase embedded systems efficiency by reducing the gap between software flexibility and hardware high-performance. Due to their reconfigurable aspect, Field Programmable Gate Arrays (FPGAs) represented a relevant step towards bridging this performance/flexibility gap. Nevertheless, Dynamic Reconfiguration (DR) has been continuously suffering from a bottleneck corresponding to a long reconfiguration time.In this thesis, we propose a novel medium-grained high-speed dynamic reconfiguration technique for DSP48E1-based circuits. The idea is to take advantage of the DSP48E1 slices runtime reprogrammability coupled with a re-routable interconnection block to change the overall circuit functionality in one clock cycle. In addition to the embedded systems efficiency, this thesis deals with the reliability chanllenges in new sub-micron electronic systems. In fact, as new technologies rely on reduced transistor size and lower supply voltages to improve performance, electronic circuits are becoming remarkably sensitive and increasingly susceptible to transient errors. The system-level impact of these errors can be far-reaching and Single Event Transients (SETs) have become a serious threat to embedded systems reliability, especially for especially for safety critical applications such as transportation systems. The reliability enhancement techniques that are based on overestimated soft error rates (SERs) can lead to unnecessary resource overheads as well as high power consumption. Considering error masking phenomena is a fundamental element for an accurate estimation of SERs.This thesis proposes a new cross-layer model of circuits vulnerability based on a combined modeling of Transistor Level (TLM) and System Level Masking (SLM) mechanisms. We then use this model to build a self adaptive fault tolerant architecture that evaluates the circuit’s effective vulnerability at runtime. Accordingly, the reliability enhancement strategy is adapted to protect only vulnerable parts of the system leading to a reliable circuit with optimized overheads. Experimentations performed on a radar-based obstacle detection system for railway transportation show that the proposed approach allows relevant reliability/resource utilization tradeoffs.

Abstract The current Internet of Things solutions for simple measurement and monitoring tasks are evolving into ubiquitous sensor networks that are constantly observing both our well being and the conditions of our living environment. The oncoming omnipresent wireless infrastructure is expected to feature artificial intelligence capabilities that can interpret human actions, gestures and even needs. All of this will require processing power on a par with and energy efficiency far beyond that of the current mobile devices. The current Internet of Things devices rely mostly on commercial low power off-the-shelf micro-controllers. Optimized solely for low power, while paying little attention to computing performance, the present solutions are far from achieving the energy efficiency, let alone, the compute capability requirements of the future Internet of Things solutions. Since this domain is application specific by nature, the use of general purpose processors for signal processing tasks is counterintuitive. Instead, dedicated accelerator based solutions are more likely to be able to meet these strict demands. This thesis proposes one potential solution for achieving the necessary low energy, as well as the flexibility and performance requirements of the Internet of Things domain in a cost effective manner using reconfigurable heterogeneous processing solutions. A novel graphics processing unit-style accelerator for the Internet of Things application domain is presented. Since the accelerator can be reconfigured, it can be used for most applications of the Internet of Things domain, as well as other application domains. The solution is assessed using two computer vision applications, and is demonstrated to achieve an excellent combination of performance and energy efficiency. The accelerator is designed using an efficient and rapid co-design flow of software and hardware, featuring ease of development characteristics close to commercial off-the-shelf solutions, which also enables cost-efficient design flow. / Tiivistelmä Esineiden internet tulee muuttamaan tulevaisuudessa elinympäristömme täysin. Se tulee mahdollistamaan interaktiiviset ympäristöt nykyisten passiivisten ympäristöjen sijaan. Lisäksi elinympäristömme tulee reagoimaan tekoihimme ja puheeseemme sekä myös tunteisiimme. Tämä kaikkialla läsnä olevan langaton infrastruktuuri tulee vaatimaan ennennäkemätöntä laskentatehokkuutta yhdistettynä äärimmäiseen energiatehokkuuteen. Nykyiset esineiden internet ratkaisut nojaavat lähes täysin kaupallisiin "suoraan hyllyltä" saataviin yleiskäyttöisiin mikrokontrollereihin. Ne ovat kuitenkin optimoituja pelkästään matalan tehonkulutuksen näkökulmasta, eivätkä niinkään energiatehokkuuden, saati tulevaisuuden esineiden internetin vaatiman laskentatehon suhteen. Kuitenkin esineiden internet on lähtökohtaisesti sovelluskohtaista laskentaa vaativa, joten yleiskäyttöisten prosessoreiden käyttö signaalinkäsittelytehtäviin on epäloogista. Sen sijaan sovelluskohtaisten kiihdyttimien käyttö laskentaan, todennäköisesti mahdollistaisi tavoitellun vaatimustason saavuttamisen. Tämä väitöskirja esittelee yhden mahdollisen ratkaisun matalan energian kulutuksen, korkean suorituskyvyn ja joustavuuden yhdenaikaiseen saavuttamiseen kustannustehokkaalla tavalla, käyttäen uudelleenkonfiguroitavia heterogeenisiä prosessoriratkaisuja. Työssä esitellään uusi grafiikkaprosessori-tyylinen uudelleen konfiguroitava kiihdytin esineiden internet sovellusalueelle, jota pystytään hyödyntämään useimpien laskentatehoa vaativien sovellusten kanssa. Ehdotetun kiihdyttimen ominaisuuksia arvioidaan kahta konenäkösovellusta esimerkkinä käyttäen ja osoitetaan sen saavuttavan loistavan yhdistelmän energia tehokkuutta ja suorituskykyä. Kiihdytin suunnitellaan käyttäen tehokasta ja nopeaa ohjelmiston ja laitteiston yhteissuunnitteluketjua, jolla voidaan saavuttaa lähestulkoon kaupallisten "suoraan hyllyltä" saatavien prosessoreiden kehitystyön helppous, joka puolestaan mahdollistaa kustannustehokkaan kehitys- ja suunnittelutyön.

Les applications vidéo embarquées sont de plus en plus intégrées dans des systèmes de transport intelligents tels que les véhicules autonomes. De nombreux défis sont rencontrés par les concepteurs de ces applications, parmi lesquels : le développement des algorithmes complexes, la vérification et le test des différentes contraintes fonctionnelles et non-fonctionnelles, la nécessité d’automatiser le processus de conception pour augmenter la productivité, la conception d’une architecture matérielle adéquate pour exploiter le parallélisme inhérent et pour satisfaire la contrainte temps-réel, réduire la puissance consommée pour prolonger la durée de fonctionnement avant de recharger le véhicule, etc. Dans ce travail de thèse, nous avons utilisé les technologies FPGAs pour relever certains de ces défis et proposer des architectures matérielles reconfigurables dédiées pour des applications embarquées de traitement vidéo temps-réel. Premièrement, nous avons implémenté une architecture parallèle flexible avec deux contributions principales : (1) Nous avons proposé un modèle générique de distribution/collecte de pixels pour résoudre le problème de transfert de données à haut débit à travers le système. Les paramètres du modèle requis sont tout d’abord définis puis la génération de l’architecture a été automatisée pour minimiser le temps de développement. (2) Nous avons appliqué une technique d’ajustement de la fréquence pour réduire la consommation d’énergie. Nous avons dérivé les équations nécessaires pour calculer le niveau maximum de parallélisme ainsi que les équations utilisées pour calculer la taille des FIFO pour le passage d’un domaine de l’horloge à un autre. Au fur et à mesure que le nombre de cellules logiques sur une seule puce FPGAaugmente, passer à des niveaux d’abstraction plus élevés devient inévitable pour réduire la contrainte de « time-to-market » et augmenter la productivité des concepteurs. Pendant la phase de conception, l’espace de solutions architecturales présente un grand nombre d’alternatives avec des performances différentes en termes de temps d’exécution, ressources matérielles, consommation d’énergie, etc. Face à ce défi, nous avons développé l’outil ViPar avec deux contributions principales : (1) Un modèle empirique a été introduit pour estimer la consommation d’énergie basé sur l’utilisation du matériel (Slice et BRAM) et la fréquence de fonctionnement ; en plus de cela, nous avons dérivé les équations pour estimer les ressources matérielles et le temps d’exécution pour chaque alternative au cours de l’exploration de l’espace de conception. (2) En définissant les principales caractéristiques de l’architecture parallèle comme le niveau de parallélisme, le nombre de ports d’entrée/sortie, le modèle de distribution des pixels, ..., l’outil ViPar génère automatiquement l’architecture matérielle pour les solutions les plus pertinentes. Dans le cadre d’une collaboration industrielle avec NAVYA, nous avons utilisé l’outil ViPar pour implémenter une solution matérielle parallèle pour l’algorithme de stéréo matching « Multi-window Sum of Absolute Difference ». Dans cette implémentation, nous avons présenté un ensemble d’étapes pour modifier le code de description de haut niveau afin de l’adapter efficacement à l’implémentation matérielle. Nous avons également exploré l’espace de conception pour différentes alternatives en termes de performance, ressources matérielles, fréquence, et consommation d’énergie. Au cours de notre travail, les architectures matérielles ont été implémentées et testées expérimentalement sur la plateforme d’évaluation Xilinx Zynq ZC706. / Embedded video applications are now involved in sophisticated transportation systems like autonomous vehicles. Many challenges faced the designers to build those applications, among them: complex algorithms should be developed, verified and tested under restricted time-to-market constraints, the necessity for design automation tools to increase the design productivity, high computing rates are required to exploit the inherent parallelism to satisfy the real-time constraints, reducing the consumed power to extend the operating duration before recharging the vehicle, etc. In this thesis work, we used FPGA technologies to tackle some of these challenges to design parallel reconfigurable hardware architectures for embedded video streaming applications. First, we implemented a flexible parallel architecture with two main contributions: (1)We proposed a generic model for pixel distribution/collection to tackle the problem of the huge data transferring through the system. The required model parameters were defined then the architecture generation was automated to minimize the development time. (2) We applied frequency scaling as a technique for reducing power consumption. We derived the required equations for calculating the maximum level of parallelism as well as the ones used for calculating the depth of the inserted FIFOs for clock domain crossing. As the number of logic cells on a single FPGA chip increases, moving to higher abstraction design levels becomes inevitable to shorten the time-to-market constraint and to increase the design productivity. During the design phase, it is common to have a space of design alternatives that are different from each other regarding hardware utilization, power consumption and performance. We developed ViPar tool with two main contributions to tackle this problem: (1) An empirical model was introduced to estimate the power consumption based on the hardware utilization (Slice and BRAM) and the operating frequency. In addition to that, we derived the equations for estimating the hardware resources and the execution time for each point during the design space exploration. (2) By defining the main characteristics of the parallel architecture like parallelism level, the number of input/output ports, the pixel distribution pattern, etc. ViPar tool can automatically generate the parallel architecture for the selected designs for implementation. In the context of an industrial collaboration, we used high-level synthesis tools to implement a parallel hardware architecture for Multi-window Sum of Absolute Difference stereo matching algorithm. In this implementation, we presented a set of guiding steps to modify the high-level description code to fit efficiently for hardware implementation as well as we explored the design space for different alternatives in terms of hardware resources, performance, frequency and power consumption. During the thesis work, our designs were implemented and tested experimentally on Xilinx Zynq ZC706 (XC7Z045- FFG900) evaluation board.

A Coarse-Grained Reconfigurable Architecture (CGRA) is a processing platform which constitutes an interconnection of coarse-grained computation units (viz. Function Units (FUs), Arithmetic Logic Units (ALUs)). These units communicate directly, viz. send-receive like primitives, as opposed to the shared memory based communication used in multi-core processors. CGRAs are a well-researched topic and the design space of a CGRA is quite large. The design space can be represented as a 7-tuple (C, N, T, P, O, M, H) where each of the terms have the following meaning: C -choice of computation unit, N -choice of interconnection network, T -Choice of number of context frame (single or multiple), P -presence of partial reconfiguration, O choice of orchestration mechanism, M -design of memory hierarchy and H host-CGRA coupling. In this thesis, we develop an architectural framework for a Macro-Dataflow based CGRA where we make the following choice for each of these parameters: C -ALU, N -Network-on-Chip (NoC), T -Multiple contexts, P -support for partial reconfiguration, O -Macro Dataflow based orchestration, M -data memory banks placed at the periphery of the reconfigurable fabric (reconfigurable fabric is the name given to the interconnection of computation units), H -loose coupling between host processor and CGRA, enabling our CGRA to execute an application independent of the host-processor’s intervention. The motivations for developing such a CGRA are: To execute applications efficiently through reduction in reconfiguration time (i.e. the time needed to transfer instructions and data to the reconfigurable fabric) and reduction in execution time through better exploitation of all forms of parallelism: Instruction Level Parallelism (ILP), Data Level Parallelism (DLP) and Thread/Task Level Parallelism (TLP). We choose a macro-dataflow based orchestration framework in combination with partial reconfiguration so as to ease exploitation of TLP and DLP. Macro-dataflow serves as a light weight synchronization mechanism. We experiment with two variants of the macro-dataflow orchestration units, namely: hardware controlled orchestration unit and the compiler controlled orchestration unit. We employ a NoC as it helps reduce the reconfiguration overhead. To permit customization of the CGRA for a particular domain through the use of domain-specific custom-Intellectual Property (IP) blocks. This aids in improving both application performance and makes it energy efficient. To develop a CGRA which is completely programmable and accepts any program written using the C89 standard. The compiler and the architecture were co-developed to ensure that every feature of the architecture could be automatically programmed through an application by a compiler. In this CGRA framework, the orchestration mechanism (O) and the host-CGRA coupling (H) are kept fixed and we permit design space exploration of the other terms in the 7-tuple design space. The mode of compilation and execution remains invariant of these changes, hence referred to as a framework. We now elucidate the compilation and execution flow for this CGRA framework. An application written in C language is compiled and is transformed into a set of temporal partitions, referred to as HyperOps in this thesis. The macro-dataflow orchestration unit selects a HyperOp for execution when all its inputs are available. The instructions and operands for a ready HyperOp are transferred to the reconfigurable fabric for execution. Each ALU (in the computation unit) is capable of waiting for the availability of the input data, prior to issuing instructions. We permit the launch and execution of a temporal partition to progress in parallel, which reduces the reconfiguration overhead. We further cut launch delays by keeping loops persistent on fabric and thus eliminating the need to launch the instructions. The CGRA framework has been implemented using Bluespec System Verilog. We evaluate the performance of two of these CGRA instances: one for cryptographic applications and another instance for linear algebra kernels. We also run other general purpose integer and floating point applications to demonstrate the generic nature of these optimizations. We explore various microarchitectural optimizations viz. pipeline optimizations (i.e. changing value of T ), different forms of macro dataflow orchestration such as hardware controlled orchestration unit and compiler-controlled orchestration unit, different execution modes including resident loops, pipeline parallelism, changes to the router etc. As a result of these optimizations we observe 2.5x improvement in performance as compared to the base version. The reconfiguration overhead was hidden through overlapping launching of instructions with execution making. The perceived reconfiguration overhead is reduced drastically to about 9-11 cycles for each HyperOp, invariant of the size of the HyperOp. This can be mainly attributed to the data dependent instruction execution and use of the NoC. The overhead of the macro-dataflow execution unit was reduced to a minimum with the compiler controlled orchestration unit. To benchmark the performance of these CGRA instances, we compare the performance of these with an Intel Core 2 Quad running at 2.66GHz. On the cryptographic CGRA instance, running at 700MHz, we observe one to two orders of improvement in performance for cryptographic applications and up to one order of magnitude performance degradation for linear algebra CGRA instance. This relatively poor performance of linear algebra kernels can be attributed to the inability in exploiting ILP across computation units interconnected by the NoC, long latency in accessing data memory placed at the periphery of the reconfigurable fabric and unavailability of pipelined floating point units (which is critical to the performance of linear algebra kernels). The superior performance of the cryptographic kernels can be attributed to higher computation to load instruction ratio, careful choice of custom IP block, ability to construct large HyperOps which allows greater portion of the communication to be performed directly (as against communication through a register file in a general purpose processor) and the use of resident loops execution mode. The power consumption of a computation unit employed on the cryptography CGRA instance, along with its router is about 76mW, as estimated by Synopsys Design Vision using the Faraday 90nm technology library for an activity factor of 0.5. The power of other instances would be dependent on specific instantiation of the domain specific units. This implies that for a reconfigurable fabric of size 5 x 6 the total power consumption is about 2.3W. The area and power ( 84mW) dissipated by the macro dataflow orchestration unit, which is common to both instances, is comparable to a single computation unit, making it an effective and low overhead technique to exploit TLP.

Achieving high computation efficiency, in terms of Cycles per Instruction (CPI), for high-performance computing kernels is an interesting and challenging research area. Dense Linear Algebra (DLA) computation is a representative high-performance computing ap- plication, which is used, for example, in LU and QR factorizations. Unfortunately, mod- ern off-the-shelf microprocessors fall significantly short of achieving theoretical lower bound in CPI for high performance computing applications. In this thesis, we perform an in-depth analysis of the available parallelisms and propose suitable algorithmic and architectural variation to significantly improve the computation efficiency. There are two standard approaches for improving the computation effficiency, first, to perform application-specific architecture customization and second, to do algorithmic tuning. In the same manner, we first perform a graph-based analysis of selected DLA kernels. From the various forms of parallelism, thus identified, we design a custom processing element for improving the CPI. The processing elements are used as building blocks for a commercially available Coarse-Grained Reconfigurable Architecture (CGRA). By per- forming detailed experiments on a synthesized CGRA implementation, we demonstrate that our proposed algorithmic and architectural variations are able to achieve lower CPI compared to off-the-shelf microprocessors. We also benchmark against state-of-the-art custom implementations to report higher energy-performance-area product. DLA computations are encountered in many engineering and scientific computing ap- plications ranging from Computational Fluid Dynamics (CFD) to Eigenvalue problem. Traditionally, these applications are written in highly tuned High Performance Comput- ing (HPC) software packages like Linear Algebra Package (LAPACK), and/or Scalable Linear Algebra Package (ScaLAPACK). The basic building block for these packages is Ba- sic Linear Algebra Subprograms (BLAS). Algorithms pertaining LAPACK/ScaLAPACK are written in-terms of BLAS to achieve high throughput. Despite extensive intellectual efforts in development and tuning of these packages, there still exists a scope for fur- ther tuning in this packages. In this thesis, we revisit most prominent and widely used compute bound algorithms like GMM for further exploitation of Instruction Level Parallelism (ILP). We further look into LU and QR factorizations for generalizations and exhibit higher ILP in these algorithms. We first accelerate sequential performance of the algorithms in BLAS and LAPACK and then focus on the parallel realization of these algorithms. Major contributions in the algorithmic tuning in this thesis are as follows: Algorithms: We present graph based analysis of General Matrix Multiplication (GMM) and discuss different types of parallelisms available in GMM We present analysis of Givens Rotation based QR factorization where we improve GR and derive Column-wise GR (CGR) that can annihilate multiple elements of a column of a matrix simultaneously. We show that the multiplications in CGR are lower than GR We generalize CGR further and derive Generalized GR (GGR) that can annihilate multiple elements of the columns of a matrix simultaneously. We show that the parallelism exhibited by GGR is much higher than GR and Householder Transform (HT) We extend generalizations to Square root Free GR (also knows as Fast Givens Rotation) and Square root and Division Free GR (SDFG) and derive Column-wise Fast Givens, and Column-wise SDFG . We also extend generalization for complex matrices and derive Complex Column-wise Givens Rotation Coarse-grained Recon gurable Architectures (CGRAs) have gained popularity in the last decade due to their power and area efficiency. Furthermore, CGRAs like REDEFINE also exhibit support for domain customizations. REDEFINE is an array of Tiles where each Tile consists of a Compute Element and a Router. The Routers are responsible for on-chip communication, while Compute Elements in the REDEFINE can be domain customized to accelerate the applications pertaining to the domain of interest. In this thesis, we consider REDEFINE base architecture as a starting point and we design Processing Element (PE) that can execute algorithms in BLAS and LAPACK efficiently. We perform several architectural enhancements in the PE to approach lower bound of the CPI. For parallel realization of BLAS and LAPACK, we attach this PE to the Router of REDEFINE. We achieve better area and power performance compared to the yesteryear customized architecture for DLA. Major contributions in architecture in this thesis are as follows: Architecture: We present design of a PE for acceleration of GMM which is a Level-3 BLAS operation We methodically enhance the PE with different features for improvement in the performance of GMM For efficient realization of Linear Algebra Package (LAPACK), we use PE that can efficiently execute GMM and show better performance For further acceleration of LU and QR factorizations in LAPACK, we identify macro operations encountered in LU and QR factorizations, and realize them on a reconfigurable data-path resulting in 25-30% lower run-time