Superscalar processors contain large, complex structures to hold data and instructions as they wait to be executed. However, many of these structures consume large amounts of energy, making them hotspots requiring sophisticated cooling systems. With the trend towards larger, more complex processors, this will become more of a problem, having important implications for future technology. This thesis uses compiler-based optimisation schemes to target the issue queue and register file. These are two of the most energy consuming structures in the processor. The algorithms and hardware techniques developed in this work dynamically adapt the processor's resources to the changing program phases, turning off parts of each structure when they are unused to save dynamic and static energy. To optimise the issue queue, the compiler analysis tracks data dependences through each program procedure. It identifies the critical path through each program region and informs the hardware of the minimum number of queue entries required to prevent it slowing down. This reduces the occupancy of the queue and increases the opportunities to save energy. With just a 1.3% performance loss, 26% dynamic and 32% static energy savings are achieved. Registers can be idle for many cycles after they are last read, before they are released and put back on the free-list to be reused by another instruction. Alternatively, they can be turned off for energy savings. Early register releasing can be used to perform this operation sooner than usual, but hardware schemes must wait for the instruction redefining the relevant logical register to enter the pipeline. This thesis presents an exploration of compiler-directed early register releasing. The compiler can exactly identify the last use of each register and pass the information to the hardware, based on simple data-flow and liveness analysis. The best scheme achieves 15% dynamic and 19% static energy savings. Finally, the issue queue limiting and early register releasing schemes are combined for energy savings in both processor structures. Four different configurations are evaluated bringing 25% to 31% dynamic and 19% to 34% static issue queue energy savings and reductions of 18% to 25% dynamic and 20% to 21% static energy in the register file.

Designing a modern processor is a very complex task. Writing the entire design using a hardware description language (like Verilog) is time consuming and difficult to verify. There exists a split architecture/microarchitecture description technique, in which, the description of any hardware can be divided into two orthogonal descriptions: (a) an architectural contract between the user and the implementation, and (b) a microarchitecture which describes the implementation of the architecture. The main aim of this thesis is to build realistic processors using this technique. We have designed an in-order and an out-of-order superscalar processor using the split-description compiler. The backend of this compiler is another contribution of this thesis. / text

As emerging portable multimedia applications demand more and more computational throughput with limited energy consumption, the need for high-efficiency, high-throughput embedded processing is becoming an important challenge in computer architecture. In this regard, this dissertation addresses application-, architecture-, and technology-level issues in existing processing systems to provide efficient processing of multimedia in many, or ideally all, of its form. In particular, this dissertation explores color imaging in multimedia while focusing on two architectural enhancements for memory- and performance-hungry embedded applications: (1) a pixel-truncation technique and (2) a color-aware instruction set (CAX) for embedded multimedia systems. The pixel-truncation technique differs from previous techniques (e.g., 4:2:2 and 4:2:0 subsampling) used in image and video compression applications (e.g., JPEG and MPEG) in that it reduces the information content in individual pixel word sizes rather than in each dimension. Thus, this technique drastically reduces the bandwidth and memory required to transport and store color images without perceivable distortion in color. At the same time, it maintains the pixel storage format of color image processing in which each pixel computation is performed simultaneously on 3-D YCbCr components, which are widely used in the image and video processing community. CAX supports parallel operations on two-packed 16-bit (6:5:5) YCbCr data in a 32-bit datapath processor, providing greater concurrency and efficiency for processing color image sequences. This dissertation presents the impact of CAX on processing performance and on both area and energy efficiency for color imaging applications in three major processor architectures: dynamically scheduled (superscalar), statically scheduled (very long instruction word, VLIW), and embedded single instruction multiple data (SIMD) array processors. Unlike typical multimedia extensions, CAX obtains substantial performance and code density improvements through direct support for color data processing rather than depending solely on generic subword parallelism. In addition, the ability to reduce data format size reduces system cost. The reduction in data bandwidth also simplifies system design. In summary, CAX, coupled with the pixel-truncation technique, provides an efficient mechanism that meets the computational requirements and cost goals for future embedded multimedia products.

Η σύγχρονη αρχιτεκτονική στρέφεται σε λύσεις που έχουν ως στόχο την εξοικονόμηση ενέργειας, χωρίς όμως να επιβαρύνεται σε μεγάλο βαθμό η απόδοση του επεξεργαστή. Ιδιαίτερα οι υπερβαθμωτοί (superscalar) επεξεργαστές που επιτρέπουν εκτέλεση εκτός σειράς (out-of-order execution) διακρίνονται από υψηλή κατανάλωση ενέργειας, εξαιτίας των πολύπλοκων δομών που χρησιμοποιούν για την αύξηση της απόδοσης. Η δυναμική ρύθμιση τάσης – συχνότητας (DVFS) αποτελεί μία ευρέως χρησιμοποιούμενη τεχνική για την επίτευξη εξοικονόμησης ενέργειας. Μειώνοντας τη συχνότητα λειτουργίας ενός κυκλώματος, είναι δυνατόν να μειωθεί και η τάση τροφοδοσίας του κυκλώματος. Με τον τρόπο αυτό ελαττώνεται και η ενέργεια που καταναλώνει το κύκλωμα. Σκοπός της εργασίας είναι η ανάπτυξη ενός μηχανισμού πραγματικού χρόνου που θα ρυθμίζει τη συχνότητα και την τάση λειτουργίας ενός superscalar, out-of-order επεξεργαστή ώστε να επιτυγχάνεται εξοικονόμηση ενέργειας χωρίς μεγάλη μείωση της απόδοσης του επεξεργαστή. Αυτό μπορεί να επιτευχθεί ελαττώνοντας τη συχνότητα και την τάση κατά τις περιόδους που ο επεξεργαστής εκτελεί πολλές λειτουργίες μνήμης. Η εξομοίωση του μηχανισμού μας για μία σειρά από μετροπρογράμματα δείχνει ότι μπορούμε να επιτύχουμε μεγάλη εξοικονόμηση ενέργειας χωρίς σημαντική αύξηση του χρόνου εκτέλεσης των προγραμμάτων. / Modern research in computer architecture focuses on techniques whose purpose is to save energy, without much loss in processor's performance. Especially superscalar processors that allow out of order execution are characterized by high energy consumption, because of the complex structures the use in order to increase performance. Dynamic Voltage - Frequency Scaling (DVFS) is a widely used technique for energy saving. Reducing the frequency of the processor's clock, it is possible to reduce the supply voltage. In this way the consumed energy is also reduced. The purpose of this diploma thesis is to create a real time mechanism that will scale the frequency and the voltage of a superscalar, out of order processor so that the processor saves energy without much loss in processor's performance. This can be made by reducing the frequency and the voltage during the periods that the processor executes many memory functions. The simulation of our mechanism for a variety of benchmarks proved that we can save much energy without much increase in the benchmark's execution time.

A Large instruction window is a key requirement to exploit greater Instruction Level Parallelism in out-of-order superscalar processors. Along with the instruction window size, the size of various other structures including the issue queue, store queue and register file need to increase as well. However, the cycle time and energy consumption of conventional large monolithic Content Addressable Memories (CAMs), the underlying structure of most conventional issue queue and store queue designs, worsen rapidly with an increase in size. This results in a three way trade-off involving ILP, clock frequency and energy consumption. In this thesis, we propose efficient designs for the issue queue and the store queue that improve the circuit latency and energy consumption while minimizing the loss in IPC. We propose the Scalable Low power Issue Queue (SLIQ) design which segments the issue queue structure to reduce the latency. This is complemented with a fast Wakeup index to a consumer in the issue queue for every instruction. As this consumer instruction can be woken up directly, without any delay, this mitigates the IPC loss faced by the pipelined issue queue. Also, as the scheme incorporates a pipelined broadcast, the indices are not required for correctness and can simply be gang invalidated on branch mispredictions. The IPC loss of an 8 segment SLIQ is Within 2.3% for the entire SPEC CPU2000 benchmark suite while achieving a 39.3% reduction in issue latency. Further, in the SLIQ design unnecessary broadcasts to the higher segments are avoided most of the time as in a large majority of the cases, an instruction has a single consumer. This consumer is woken up either by direct indexing or by broadcast in the first segment of the SLIQ. This enables the 8 segment SLIQ to significantly reduce the energy consumption and the energy-delay product by 48.3% and 67.4% respectively on an average. SLIQ also allows the architects to segment the issue queue carefully so that the latency of the issue logic is just within the per pipeline stage latency goals of the design. We also propose the Scalable Low power Store Queue (SLSQ) to address similar problems associated with the store queue data forwarding logic. We extend the state- of-the-art Store Vector based Disambiguator to also predict the index of the store that will forward to a given load. SLSQ marginally adds to the hardware budget, but predicts the store queue index of the store which will forward with an accuracy of 99.5% on an average. SLSQ, thus, eliminates unnecessary address broadcasts and Compares and reduces energy consumption of the store-to-load forwarding logic by 78.4% and 91.6% for the SPEC Int and FP suites respectively. Another variant of SLSQ, eliminates the need for a CAM in the forwarding logic and achieves a 49.9% reduction in store to load data forwarding latency while incurring a minimal IPC loss less than 0.1% on average for the entire SPEC CPU2000 benchmark suite.

Processor architectures are becoming increasingly complex and hence architects have to evaluate a large design space consisting of several parameters, each with a number of potential settings. In order to assist in guiding design decisions we develop simple and accurate models of the superscalar processor design space using a detailed and validated superscalar processor simulator. Firstly, we obtain precise estimates of all significant micro-architectural parameters and their interactions by building linear regression models using simulation based experiments. We obtain good approximate models at low simulation costs using an iterative process in which Akaike’s Information Criteria is used to extract a good linear model from a small set of simulations, and limited further simulation is guided by the model using D-optimal experimental designs. The iterative process is repeated until desired error bounds are achieved. We use this procedure for model construction and show that it provides a cost effective scheme to experiment with all relevant parameters. We also obtain accurate predictors of the processors performance response across the entire design-space, by constructing radial basis function networks from sampled simulation experiments. We construct these models, by simulating at limited design points selected by latin hypercube sampling, and then deriving the radial neural networks from the results. We show that these predictors provide accurate approximations to the simulator’s performance response, and hence provide a cheap alternative to simulation while searching for optimal processor design points.

Bien que les processeurs actuels possèdent plus de 10 cœurs, de nombreux programmes restent purement séquentiels. Cela peut être dû à l'algorithme que le programme met en œuvre, au programme étant vieux et ayant été écrit durant l'ère des uni-processeurs, ou simplement à des contraintes temporelles, car écrire du code parallèle est notoirement long et difficile. De plus, même pour les programmes parallèles, la performance de la partie séquentielle de ces programmes devient rapidement le facteur limitant l'augmentation de la performance apportée par l'augmentation du nombre de cœurs disponibles, ce qui est exprimé par la loi d'Amdahl. Conséquemment, augmenter la performance séquentielle reste une approche valide même à l'ère des multi-cœurs.Malheureusement, la façon conventionnelle d'améliorer la performance (augmenter la taille de la fenêtre d'instructions) contribue à l'augmentation de la complexité et de la consommation du processeur. Dans ces travaux, nous revisitons une technique visant à améliorer la performance de façon orthogonale : La prédiction de valeurs. Au lieu d'augmenter les capacités du moteur d'exécution, la prédiction de valeurs améliore l'utilisation des ressources existantes en augmentant le parallélisme d'instructions disponible.En particulier, nous nous attaquons aux trois problèmes majeurs empêchant la prédiction de valeurs d'être mise en œuvre dans les processeurs modernes. Premièrement, nous proposons de déplacer la validation des prédictions depuis le moteur d'exécution vers l'étage de retirement des instructions. Deuxièmement, nous proposons un nouveau modèle d'exécution qui exécute certaines instructions dans l'ordre soit avant soit après le moteur d'exécution dans le désordre. Cela réduit la pression exercée sur ledit moteur et permet de réduire ses capacités. De cette manière, le nombre de ports requis sur le fichier de registre et la complexité générale diminuent. Troisièmement, nous présentons un mécanisme de prédiction imitant le mécanisme de récupération des instructions : La prédiction par blocs. Cela permet de prédire plusieurs instructions par cycle tout en effectuant une unique lecture dans le prédicteur. Ces trois propositions forment une mise en œuvre possible de la prédiction de valeurs qui est réaliste mais néanmoins performante. / Although currently available general purpose microprocessors feature more than 10 cores, many programs remain mostly sequential. This can either be due to an inherent property of the algorithm used by the program, to the program being old and written during the uni-processor era, or simply to time to market constraints, as writing and validating parallel code is known to be hard. Moreover, even for parallel programs, the performance of the sequential part quickly becomes the limiting improvement factor as more cores are made available to the application, as expressed by Amdahl's Law. Consequently, increasing sequential performance remains a valid approach in the multi-core era. Unfortunately, conventional means to do so - increasing the out-of-order window size and issue width - are major contributors to the complexity and power consumption of the chip. In this thesis, we revisit a previously proposed technique that aimed to improve performance in an orthogonal fashion: Value Prediction (VP). Instead of increasing the execution engine aggressiveness, VP improves the utilization of existing resources by increasing the available Instruction Level Parallelism. In particular, we address the three main issues preventing VP from being implemented. First, we propose to remove validation and recovery from the execution engine, and do it in-order at Commit. Second, we propose a new execution model that executes some instructions in-order either before or after the out-of-order engine. This reduces pressure on said engine and allows to reduce its aggressiveness. As a result, port requirement on the Physical Register File and overall complexity decrease. Third, we propose a prediction scheme that mimics the instruction fetch scheme: Block Based Prediction. This allows predicting several instructions per cycle with a single read, hence a single port on the predictor array. This three propositions form a possible implementation of Value Prediction that is both realistic and efficient.