Test cost is a critical component in the overall cost of the product. Test cost varies in direct proportion with test time. This thesis introduces a data driven feed forward adaptive technique for reducing test time at wafer sort while maintaining the product defect level. Test data from first insertion of wafer is statistically analyzed to make a decision about adaptive test flow at subsequent insertions. The data driven feed forward technique uses a statistical screen to analyze test data from first probe of wafer and provides recommendations for test elimination at second insertions. At the second insertion dies are subjected to only the optimum number of tests for a reduced test flow. This technique is applicable only for the products which are tested at two or more insertions. The statistical screen identifies the dies for reduced test flow based upon correlation of tests across insertions. The tests which are repeated at both the insertions and are highly correlated are the candidates of elimination at second insertion. The feed forward technique is applied to a mixed signal analog product and figures of merit are evaluated. Application of technique to the production data shows that there is an average 55% test time reduction when a single site is tested per touchdown and up to 10% when 16 sites are tested in parallel per touchdown.

At this time the synthesis of reversible circuits for quantum computing is an active area of research. In the most restrictive quantum computing models there are no ancilla lines and the quantum cost, or latency, of performing a reversible form of the AND gate, or Toffoli gate, increases exponentially with the number of input variables. In contrast, the quantum cost of performing any combination of reversible EXOR gates, or CNOT gates, on n input variables requires at most O(n2/log2n) gates. It was under these conditions that EXOR-AND-EXOR, or EPOE, synthesis was developed. In this work, the GF(2) logic theory used in EPOE is expanded and the concept of an EXOR-AND product transform is introduced. Because of the generality of this logic theory, it is adapted to EXOR-AND-OR, or SPOE, synthesis. Three heuristic spectral logic synthesis algorithms are introduced, implemented in a program called XAX, and compared with previous work in classical logic circuits of up to 26 inputs. Three linear reversible circuit methods are also introduced and compared with previous work in linear reversible logic circuits of up to 100 inputs.

Energetic particles from cosmic ray or terrestrial sources can strike sensitive areas of CMOS devices and cause soft errors. Understanding the effects of such interactions is crucial as the device technology advances, and chip reliability has become more important than ever. Particle accelerator testing has been the standard method to characterize the sensitivity of chips to single event upsets (SEUs). However, because of their costs and availability limitations, other techniques have been explored. Pulsed laser has been a successful tool for characterization of SEU behavior, but to this day, laser has not been recognized as a comparable method to beam testing. In this thesis, I propose a methodology of correlating laser soft error rate (SER) to particle beam gathered data. Additionally, results are presented showing a temperature dependence of SER and the "neighbor effect" phenomenon where due to the close proximity of devices a "weakening effect" in the ON state can be observed.

Through silicon via (TSV) based 3D integrated circuits have inspired a novel design paradigm which explores the vertical dimension, in order to alleviate the performance and power limitations associated with long interconnects in 2D circuits. TSVs enable vertical interconnects across stacked and thinned dies in 3D-IC designs, resulting in reduced wirelength, footprint, faster speed, improved bandwidth, and lesser routing congestion. However, the usage of TSVs itself gives rise to many critical design challenges towards the minimization of chip delay and power consumption. Therefore, realization of the benefits of 3D ICs necessitates an early and realistic prediction of circuit performance during the early layout design stage. The goal of this thesis is to meet the design challenges of 3D ICs by providing new capabilities to the existing floorplanning framework [87]. The additional capabilities included in the existing floorplanning tool is the co-placement of TSV islands with circuit blocks and performing non-deterministic assignment of signals to TSVs. We also replace the wirelength and number of TSVs in the floorplanning cost function with the total delay in the nets. The delay-aware cost function accounts for RC delay impact of TSVs on the delay of individual signal connection, and obviates the efforts required to balance the weight contributions of wirelength and TSVs in the wirelength-aware floorplanning. Our floorplanning tool results in 5% shorter wirelength and 21% lesser TSVs compared to recent approaches. The delay in the cost function improves total delay in the interconnects by 10% - 12% compared to wirelength-aware cost function. The influence of large coupling capacitance between TSVs on the delay, power and coupling noise in 3D interconnects also offers serious challenges to the performance of 3D-IC. Due to the degree of design complexity introduced by TSVs in 3D ICs, the importance of early stage evaluation and optimization of delay, power and signal integrity of 3D circuits cannot be ignored. The unique contribution of this work is to develop methods for accurate analysis of timing, power and coupling noise across multiple stacked device layers during the floorplanning stage. Incorporating the impact of TSV and the stacking of multiple device layers within floorplanning framework will help to achieve 3D layouts with superior performance. Therefore, we proposed an efficient TSV coupling noise model to evaluate the coupling noise in the 3D interconnects during floorplanning. The total coupling noise in 3D interconnects is included in the cost function to optimize positions of TSVs and blocks, as well as nets-to-TSVs assignment to obtain floorplans with minimized coupling noise. We also suggested diagonal TSV arrangement for larger TSV pitch and nonuniform pitch arrangement for reducing worst TSV-to-TSV coupling, thereby minimizing the coupling noise in the interconnects. This thesis also focuses on more realistic evaluation and optimization of delay and power in TSV based 3D integrated circuits considering the interconnect density on individual device layers. The floorplanning tool uses TSV locations and delay, non-uniform interconnect density across multiple stacked device layers to assess and optimize the buffer count, delay, and interconnect power dissipation in a design. It is shown that the impact of non-uniform interconnect density, across the stacked device layers, should not be ignored, as its contribution to the performance of the 3D interconnects is consequential. A wire capacitance-aware buffer insertion scheme is presented that determines the optimal distance between adjacent buffers on the individual device layers for nonuniform wire density between stacked device layers. The proposed approach also considers TSV location on a 3D wire to optimize the buffer insertion around TSVs. For 3D designs with uniform wire density across stacked device layers, we propose a TSV-aware buffer insertion approach that appropriately models the TSV RC delay impact on interconnect delay to determine the optimum interval between adjacent buffers for individual 3D nets. Moreover, our floorplanning tool help achieve 3D layouts with superior performance by incorporating the impact of nonuniform density on the delay, power and coupling noise in the interconnects during floorplanning.

As a major component of a computing system, memory has been a key performance and power consumption bottleneck in computer system design. While processor speeds have been kept rising dramatically, the overall computing performance improvement of the entire system is limited by how fast the memory can feed instructions/data to processing units (i.e. so-called memory wall problem). The increasing transistor density and surging access demands from a rapidly growing number of processing cores also significantly elevated the power consumption of the memory system. In addition, the interference of memory access from different applications and processing cores significantly degrade the computation predictability, which is essential to ensure timing specifications in real-time system design. The recent IC technologies (such as 3D-IC technology) and emerging data-intensive real-time applications (such as Virtual Reality/Augmented Reality, Artificial Intelligence, Internet of Things) further amplify these challenges. We believe that it is not simply desirable but necessary to adopt a joint CPU/Memory resource management framework to deal with these grave challenges. In this dissertation, we focus on studying how to schedule fixed-priority hard real-time tasks with memory impacts taken into considerations. We target on the fixed-priority real-time scheduling scheme since this is one of the most commonly used strategies for practical real-time applications. Specifically, we first develop an approach that takes into consideration not only the execution time variations with cache allocations but also the task period relationship, showing a significant improvement in the feasibility of the system. We further study the problem of how to guarantee timing constraints for hard real-time systems under CPU and memory thermal constraints. We first study the problem under an architecture model with a single core and its main memory individually packaged. We develop a thermal model that can capture the thermal interaction between the processor and memory, and incorporate the periodic resource sever model into our scheduling framework to guarantee both the timing and thermal constraints. We further extend our research to the multi-core architectures with processing cores and memory devices integrated into a single 3D platform. To our best knowledge, this is the first research that can guarantee hard deadline constraints for real-time tasks under temperature constraints for both processing cores and memory devices. Extensive simulation results demonstrate that our proposed scheduling can improve significantly the feasibility of hard real-time systems under thermal constraints.

Cette thèse présente la conception et la réalisation en VLSI d'un processeur programmable pour une nouvelle architecture MIMD massivement parallèle, intermédiaire entre la connection machine et les hypercubes de processeurs 32 bits. Elle est composée d'une grille 2d de cellules asynchrones communiquant par échanges de messages. Chaque cellule intégré une partie de traitement qui consiste en un petit microprocesseur 8 bits dote d'une mémoire (données et programme), et une partie de routage permettant l'acheminement des messages. A l'issue de l'étude des différents problèmes de communication dans les machines parallèles, nous proposons un routeur original utilisant le principe du Wormhole, et permettant d'acheminer jusqu'à cinq messages en parallèle. Nous décrivons ensuite l'architecture de la partie de traitement, en partant de la définition du jeu d'instructions, du chemin de données et de la partie contrôle jusqu'à la conception au bas niveau. Un premier prototype d'un circuit VLSI de ce processeur a été réalise sur silicium et a permis d'obtenir les mesures des surfaces et des performances

Dans le cadre de cette thèse nous avons étudie l'implantation des algorithmes de l'arithmétique en ligne. En particulier, la réalisation de deux circuits destines aux applications exigeant une précision infinie est exposée. En effet, dans de nombreux domaines tels que la génération de nombres aléatoires, cryptographie, calcul formel, arithmétique exacte, réduction de fraction en précision infinie, calcul modulaire, traitement d'images..., les opérateurs classiques manquent d'efficacité. Face a ce type de problèmes, un remède peut être apporte par le calcul en ligne selon lequel les calculs sont faits en introduisant les opérandes en série chiffre a chiffre en notation redondante. Nous obtenons ainsi un haut degré de parallélisme et une précision variable linéairement. Le premier circuit présenté implante un algorithme de pgcd nomme Euclide offrant, d'après les simulations, le meilleur compromis cout matériel/performance. Il donne également les coefficients de Bezout. Ce circuit est appelé a résoudre les problèmes lies au temps de calcul du pgcd par les méthodes classiques rencontrées dans beaucoup d'applications. Une deuxième application montre la possibilité de fusionner des opérateurs en ligne afin d'obtenir un opérateur complexe. L'exemple traite dans cette thèse est celui de la distance euclidienne: z=x#2+y#2 utilisée, entre autres, pour la resolution du moindre carre des systèmes linéaires

Le développement des techniques d'intégration permet de réaliser des circuits de 10**(5) a 10**(6) transistors et, dans un futur proche, des circuits encore plus complexes. Les problèmes de CAO deviennent donc de plus en plus ardus, comme la simulation logique ou le placement. Cette même évolution nous autorise a réaliser des machines parallèles très puissantes pour résoudre ces problèmes. Nous proposons l'architecture d'un réseau cellulaire asynchrone. Ce réseau, compose de nxn cellules dont chacune est physiquement connectée à ses quatre voisines, dispose d'un mécanisme de communication permettant l'acheminement de messages d'une cellule quelconque a n'importe quelle autre. Un circuit intégré, incluant un réseau de 2 x 2 cellules dédié à la simulation logique, a été réalisé. Utilisant cette architecture cellulaire nous avons développé un placeur, qui a partir d'une configuration initiale, minimise la longueur des connexions par échanges de paires. Nous avons aborde la manière dont le placement pourrait être amélioré par la méthode de recuit simule. Ces deux applications, différentes de l'architecture cellulaire, nous permettent de constater que beaucoup de parties sont communes aux deux circuits. Nous énonçons certaines règles de façon à rendre la conception plus rapide et plus sure

Étude pour des circuits VLSI sur substrat de silicium. Les modèles de pannes développés pour la technologie NMOS ne sont plus adaptes à la vérification des pannes en technologie CMOS. On examine les pannes de type déclenchement parasite, court-circuit, blocage sur et blocage ouvert. Pour chacune de ces pannes un modèle est défini et on détermine les méthodes de vérification correspondantes. Les principaux comportements étudies sont la transformation d'un circuit logique en analogique et la transformation d'un circuit combinatoire en un circuit séquentiel. On démontre un ensemble de lemmes et théorèmes de base pour la vérification des pannes en technologie CMOS. Ces théorèmes étendent à la vérification du blocage ouvert CMOS les résultats formules pour la vérification des collages logiques dans les réseaux. Certains de ces théorèmes impliquent une conception adaptée pour faciliter la vérification. Réduction des séquences de vérification et vérification simultanée.

Partant d'un jeu d'instructions spécifique à l'application (domaine des automatismes logiques), on propose une architecture permettant d'exécuter ledit jeu d'instructions et disposant de facilités de test en fin de conception et au cours de la vie du circuit. L'observabilité et la contrôlabilité du composant représentent une partie importante de l'étude. Après examen critique de plusieurs méthodes permettant de faciliter le test (en ligne et hors ligne) du circuit, un choix est réalisé afin d'intégrer dans l'architecture de ce dernier les dispositifs nécessaires à la mise en œuvre de certaines d'entre elles