As the number of cores per chip increases, maintaining cache coherence becomes prohibitive for both power and performance. Non Coherent Cache (NCC) architectures do away with hardware-based cache coherence, but they become difficult to program. Some existing architectures provide a middle ground by providing some shared memory in the hardware. Specifically, the 48-core Intel Single-chip Cloud Computer (SCC) provides some off-chip (DRAM) shared memory some on-chip (SRAM) shared memory. We call such architectures Hybrid Shared Memory, or HSM, manycore architectures. However, how to efficiently execute multi-threaded programs on HSM architectures is an open problem. To be able to execute a multi-threaded program correctly on HSM architectures, the compiler must: i) identify all the shared data and map it to the shared memory, and ii) map the frequently accessed shared data to the on-chip shared memory. This work presents a source-to-source translator written using CETUS that identifies a conservative superset of all the shared data in a multi-threaded application and maps it to the shared memory such that it enables execution on HSM architectures.

A benchmark suite that is representative of the programs a processor typically executes is necessary to understand a processor's performance or energy consumption characteristics. The first contribution of this work addresses this need for mobile platforms with MobileBench, a selection of representative smartphone applications. In smartphones, like any other portable computing systems, energy is a limited resource. Based on the energy characterization of a commercial widely-used smartphone, application cores are found to consume a significant part of the total energy consumption of the device. With this insight, the subsequent part of this thesis focuses on the portion of energy that is spent to move data from the memory system to the application core's internal registers. The primary motivation for this work comes from the relatively higher power consumption associated with a data movement instruction compared to that of an arithmetic instruction. The data movement energy cost is worsened esp. in a System on Chip (SoC) because the amount of data received and exchanged in a SoC based smartphone increases at an explosive rate. A detailed investigation is performed to quantify the impact of data movement

on the overall energy consumption of a smartphone device. To aid this study, microbenchmarks that generate desired data movement patterns between different levels of the memory hierarchy are designed. Energy costs of data movement are then computed by measuring the instantaneous power consumption of the device when the micro benchmarks are executed. This work makes an extensive use of hardware performance counters to validate the memory access behavior of microbenchmarks and to characterize the energy consumed in moving data. Finally, the calculated energy costs of data movement are used to characterize the portion of energy that MobileBench applications spend in moving data. The results of this study show that a significant 35% of the total device energy is spent in data movement alone. Energy is an increasingly important criteria in the context of designing architectures for future smartphones and this thesis offers insights into data movement energy consumption.

Software Managed Manycore (SMM) architectures - in which each core has only a scratch pad memory (instead of caches), - are a promising solution for scaling memory hierarchy to hundreds of cores. However, in these architectures, the code and data of the tasks mapped to the cores must be explicitly managed in the software by the compiler. State-of-the-art compiler techniques for SMM architectures require inter-procedural information and analysis. A call graph of the program does not have enough information, and Global CFG, i.e., combining all the control flow graphs of the program has too much information, and becomes too big. As a result, most new techniques have informally defined and used GCCFG (Global Call Control Flow Graph) - a whole program representation which captures the control-flow as well as function call information in a succinct way - to perform inter-procedural analysis. However, how to construct it has not been shown yet. We find that for several simple call and control flow graphs, constructing GCCFG is relatively straightforward, but there are several cases in common applications where unique graph transformation is needed in order to formally and correctly construct the GCCFG. This paper fills this gap, and develops graph transformations to allow the construction of GCCFG in (almost) all cases. Our experiments show that by using succinct representation (GCCFG) rather than elaborate representation (GlobalCFG), the compilation time of state-of-the-art code management technique [4] can be improved by an average of 5X, and that of stack management [20] can be improved by an average of 4X.

Android has been the dominant platform in which most of the mobile development is being done. By the end of the second quarter of 2014, 84.7 percent of the entire world mobile phones market share had been captured by Android. The Android library internally uses the modified Linux kernel as the part of its stack. The I/O scheduler, is a part of the Linux kernel, responsible for scheduling data requests to the internal and the external memory devices that are attached to the mobile systems.

The usage of solid state drives in the Android tablet has also seen a rise owing to its speed of operation and mechanical stability. The I/O schedulers that exist in the present Linux kernel are not better suited for handling solid state drives in particular to exploit the inherent parallelism offered by the solid state drives. The Android provides information to the Linux kernel about the processes running in the foreground and background. Based on this information the kernel decides the process scheduling and the memory management, but no such information exists for the I/O scheduling. Research shows that the resource management could be done better if the operating system is aware of the characteristics of the requester. Thus, there is a need for a better I/O scheduler that could schedule I/O operations based on the application and also exploit the parallelism in the solid state drives. The scheduler proposed through this research does that. It contains two algorithms working in unison one focusing on the solid state drives and the other on the application awareness.

The Android application context aware scheduler has the features of increasing the responsiveness of the time sensitive applications and also increases the throughput by parallel scheduling of request in the solid state drive. The suggested scheduler is tested using standard benchmarks and real-time scenarios, the results convey that our scheduler outperforms the existing default completely fair queuing scheduler of the Android.

Coarse Grain Reconfigurable Arrays (CGRAs) are promising accelerators capable of

achieving high performance at low power consumption. While CGRAs can efficiently

accelerate loop kernels, accelerating loops with control flow (loops with if-then-else

structures) is quite challenging. Techniques that handle control flow execution in

CGRAs generally use predication. Such techniques execute both branches of an

if-then-else structure and select outcome of either branch to commit based on the

result of the conditional. This results in poor utilization of CGRA s computational

resources. Dual-issue scheme which is the state of the art technique for control flow

fetches instructions from both paths of the branch and selects one to execute at

runtime based on the result of the conditional. This technique has an overhead in

instruction fetch bandwidth. In this thesis, to improve performance of control flow

execution in CGRAs, I propose a solution in which the result of the conditional

expression that decides the branch outcome is communicated to the instruction fetch

unit to selectively issue instructions from the path taken by the branch at run time.

Experimental results show that my solution can achieve 34.6% better performance

and 52.1% improvement in energy efficiency on an average compared to state of the

art dual issue scheme without imposing any overhead in instruction fetch bandwidth.

Stream processing has emerged as an important model of computation especially in the context of multimedia and communication sub-systems of embedded System-on-Chip (SoC) architectures. The dataflow nature of streaming applications allows them to be most naturally expressed as a set of kernels iteratively operating on continuous streams of data. The kernels are computationally intensive and are mainly characterized by real-time constraints that demand high throughput and data bandwidth with limited global data reuse. Conventional architectures fail to meet these demands due to their poorly matched execution models and the overheads associated with instruction and data movements.

This work presents StreamWorks, a multi-core embedded architecture for energy-efficient stream computing. The basic processing element in the StreamWorks architecture is the StreamEngine (SE) which is responsible for iteratively executing a stream kernel. SE introduces an instruction locking mechanism that exploits the iterative nature of the kernels and enables fine-grain instruction reuse. Each instruction in a SE is locked to a Reservation Station (RS) and revitalizes itself after execution; thus never retiring from the RS. The entire kernel is hosted in RS Banks (RSBs) close to functional units for energy-efficient instruction delivery. The dataflow semantics of stream kernels are captured by a context-aware dataflow execution mode that efficiently exploits the Instruction Level Parallelism (ILP) and Data-level parallelism (DLP) within stream kernels.

Multiple SEs are grouped together to form a StreamCluster (SC) that communicate via a local interconnect. A novel software FIFO virtualization technique with split-join functionality is proposed for efficient and scalable stream communication across SEs. The proposed communication mechanism exploits the Task-level parallelism (TLP) of the stream application. The performance and scalability of the communication mechanism is evaluated against the existing data movement schemes for scratchpad based multi-core architectures. Further, overlay schemes and architectural support are proposed that allow hosting any number of kernels on the StreamWorks architecture. The proposed oevrlay schemes for code management supports kernel(context) switching for the most common use cases and can be adapted for any multi-core architecture that use software managed local memories.

The performance and energy-efficiency of the StreamWorks architecture is evaluated for stream kernel and application benchmarks by implementing the architecture in 45nm TSMC and comparison with a low power RISC core and a contemporary accelerator.

Coarse-Grained Reconfigurable Architectures (CGRA) are a promising fabric for improving the performance and power-efficiency of computing devices. CGRAs are composed of components that are well-optimized to execute loops and rotating register file is an example of such a component present in CGRAs. Due to the rotating nature of register indexes in rotating register file, it is very challenging, if at all possible, to hold and properly index memory addresses (pointers) and static values. In this Thesis, different structures for CGRA register files are investigated. Those structures are experimentally compared in terms of performance of mapped applications, design frequency, and area. It is shown that a register file that can logically be partitioned into rotating and non-rotating regions is an excellent choice because it imposes the minimum restriction on underlying CGRA mapping algorithm while resulting in efficient resource utilization.

Software has a great impact on the energy efficiency of any computing system--it can manage the components of a system efficiently or inefficiently. The impact of software is amplified in the context of a wearable computing system used for activity recognition. The design space this platform opens up is immense and encompasses sensors, feature calculations, activity classification algorithms, sleep schedules, and transmission protocols. Design choices in each of these areas impact energy use, overall accuracy, and usefulness of the system. This thesis explores methods software can influence the trade-off between energy consumption and system accuracy. In general the more energy a system consumes the more accurate will be. We explore how finding the transitions between human activities is able to reduce the energy consumption of such systems without reducing much accuracy. We introduce the Log-likelihood Ratio Test as a method to detect transitions, and explore how choices of sensor, feature calculations, and parameters concerning time segmentation affect the accuracy of this method. We discovered an approximate 5X increase in energy efficiency could be achieved with only a 5% decrease in accuracy. We also address how a system's sleep mode, in which the processor enters a low-power state and sensors are turned off, affects a wearable computing platform that does activity recognition. We discuss the energy trade-offs in each stage of the activity recognition process. We find that careful analysis of these parameters can result in great increases in energy efficiency if small compromises in overall accuracy can be tolerated. We call this the ``Great Compromise.'' We found a 6X increase in efficiency with a 7% decrease in accuracy. We then consider how wireless transmission of data affects the overall energy efficiency of a wearable computing platform. We find that design decisions such as feature calculations and grouping size have a great impact on the energy consumption of the system because of the amount of data that is stored and transmitted. For example, storing and transmitting vector-based features such as FFT or DCT do not compress the signal and would use more energy than storing and transmitting the raw signal. The effect of grouping size on energy consumption depends on the feature. For scalar features energy consumption is proportional in the inverse of grouping size, so it's reduced as grouping size goes up. For features that depend on the grouping size, such as FFT, energy increases with the logarithm of grouping size, so energy consumption increases slowly as grouping size increases. We find that compressing data through activity classification and transition detection significantly reduces energy consumption and that the energy consumed for the classification overhead is negligible compared to the energy savings from data compression. We provide mathematical models of energy usage and data generation, and test our ideas using a mobile computing platform, the Texas Instruments Chronos watch.

We are expecting hundreds of cores per chip in the near future. However, scaling the memory architecture in manycore architectures becomes a major challenge. Cache coherence provides a single image of memory at any time in execution to all the cores, yet coherent cache architectures are believed will not scale to hundreds and thousands of cores. In addition, caches and coherence logic already take 20-50% of the total power consumption of the processor and 30-60% of die area. Therefore, a more scalable architecture is needed for manycore architectures. Software Managed Manycore (SMM) architectures emerge as a solution. They have scalable memory design in which each core has direct access to only its local scratchpad memory, and any data transfers to/from other memories must be done explicitly in the application using Direct Memory Access (DMA) commands. Lack of automatic memory management in the hardware makes such architectures extremely power-efficient, but they also become difficult to program. If the code/data of the task mapped onto a core cannot fit in the local scratchpad memory, then DMA calls must be added to bring in the code/data before it is required, and it may need to be evicted after its use. However, doing this adds a lot of complexity to the programmer's job. Now programmers must worry about data management, on top of worrying about the functional correctness of the program - which is already quite complex. This dissertation presents a comprehensive compiler and runtime integration to automatically manage the code and data of each task in the limited local memory of the core. We firstly developed a Complete Circular Stack Management. It manages stack frames between the local memory and the main memory, and addresses the stack pointer problem as well. Though it works, we found we could further optimize the management for most cases. Thus a Smart Stack Data Management (SSDM) is provided. In this work, we formulate the stack data management problem and propose a greedy algorithm for the same. Later on, we propose a general cost estimation algorithm, based on which CMSM heuristic for code mapping problem is developed. Finally, heap data is dynamic in nature and therefore it is hard to manage it. We provide two schemes to manage unlimited amount of heap data in constant sized region in the local memory. In addition to those separate schemes for different kinds of data, we also provide a memory partition methodology.

In recent years we have witnessed a shift towards multi-processor system-on-chips (MPSoCs) to address the demands of embedded devices (such as cell phones, GPS devices, luxury car features, etc.). Highly optimized MPSoCs are well-suited to tackle the complex application demands desired by the end user customer. These MPSoCs incorporate a constellation of heterogeneous processing elements (PEs) (general purpose PEs and application-specific integrated circuits (ASICS)). A typical MPSoC will be composed of a application processor, such as an ARM Coretex-A9 with cache coherent memory hierarchy, and several application sub-systems. Each of these sub-systems are composed of highly optimized instruction processors, graphics/DSP processors, and custom hardware accelerators. Typically, these sub-systems utilize scratchpad memories (SPM) rather than support cache coherency. The overall architecture is an integration of the various sub-systems through a high bandwidth system-level interconnect (such as a Network-on-Chip (NoC)). The shift to MPSoCs has been fueled by three major factors: demand for high performance, the use of component libraries, and short design turn around time. As customers continue to desire more and more complex applications on their embedded devices the performance demand for these devices continues to increase. Designers have turned to using MPSoCs to address this demand. By using pre-made IP libraries designers can quickly piece together a MPSoC that will meet the application demands of the end user with minimal time spent designing new hardware. Additionally, the use of MPSoCs allows designers to generate new devices very quickly and thus reducing the time to market. In this work, a complete MPSoC synthesis design flow is presented. We first present a technique \cite{leary1_intro} to address the synthesis of the interconnect architecture (particularly Network-on-Chip (NoC)). We then address the synthesis of the memory architecture of a MPSoC sub-system \cite{leary2_intro}. Lastly, we present a co-synthesis technique to generate the functional and memory architectures simultaneously. The validity and quality of each synthesis technique is demonstrated through extensive experimentation.