'Modern' radar systems have come in for much criticism in recent years, particularly in the aftermath of the Falklands campaign. There have also been notable failures in commercial designs, including the well-publicised 'Nimrod' project which was abandoned due to persistent inability to meet signal processing requirements. There is clearly a need for improvement in radar signal processing techniques as many designs rely on technology dating from the late 1970's, much of which is obsolete by today’s standards. The Durham Radar Automatic Track Extraction System (RATES) is a practical implementation of current microprocessor technology, applied to plot extraction of surveillance radar data. In addition to suggestions for the design of such a system, results are quoted for the predicted performance when compared with a similar product using 1970's design methodology. Suggestions are given for the use of other VLSI techniques in plot extraction, including logic arrays and digital signal processors. In conclusion, there is an illustrated discussion concerning the use of systolic arrays in RATES and a prediction that this will represent the optimum architecture for future high-speed radar signal processors.

Thesis (Ph. D.)--State University of New York at Binghamton, Dept. of Computer Science, 2007. / Includes bibliographical references.

Thesis (Ph. D.)--College of Computing, Georgia Institute of Technology, 2005. Directed by Karsten Schwan. / Peter Steenkiste, Committee Member ; George Riley, Committee Member ; Kenneth Mackenzie, Committee Member ; Calton Pu, Committee Member ; Karsten Schwan, Committee Member. Vita. Includes bibliographical references.

Thesis (M.S.)--Oregon State University, 2006. / Printout. Includes bibliographical references (leaves 29-30). Also available on the World Wide Web.

This thesis examines systolic array architectures and their methods of control and communication synchronisation. Systolic array processors suffer from synchronisation problems associated with the clocking mechanism that causally restricts their scalability. To overcome this problem both return-to-zero (RTZ) and non-return-to zero (NRTZ) delay-insensitive self-timed (ST) techniques can be used to realise architectures that operate correctly in the presence of arbitrary delays at all levels in their design. As a consequence, RTZ and NRTZ versions of an existing systolic array architecture, namely the Single instruction Systolic Array (SISA), have been developed in order to investigate the potential for realising architecturally scaleable systolic arrays. The new architectures, called the RTZ and NRTZ ST-SISAs, have been compared with each other and against their synchronous counterpart to establish their relative trade-offs. The new designs exhibit several novel features including: variable length bit-serial data words, average case processing speeds dependent on data word length as well as computational complexity, a novel autonomous inter-processor data communication mechanism and architectural scalability independent of fabrication technology. This thesis introduces an implementation of the RTZ and NRTZ ST-SISA architectures, along with their performance and area characteristics. Guidelines have been developed from the resulting RTZ and NRTZ architectures allowing novel self-timed systolic architectures to be derived.

Most embedded devices are based on heterogeneous Multiprocessor System on Chips (MPSoCs). These contain a variety of processors like CPUs, micro-controllers, DSPs, GPUs and specialised accelerators. The heterogeneity of these systems helps in achieving good performance and energy efficiency but makes programming inherently difficult. There is no single programming language or runtime to program such platforms. This thesis makes three contributions to these problems. First, it presents a framework that allows code in Single Program Multiple Data (SPMD) form to be mapped to a heterogeneous platform. The mapping space is explored, and it is shown that the best mapping depends on the metric used. Next, a compiler framework is presented which bridges the gap between the high -level programming model of OpenMP and the heterogeneous resources of MPSoCs. It takes OpenMP programs and generates code which runs on all processors. It delivers programming ease while exploiting heterogeneous resources. Finally, a compiler-based approach to runtime power management for heterogeneous cores is presented. Given an externally provided budget, the approach generates heterogeneous, partitioned code that attempts to give the best performance within that budget.

This thesis describes the design and implementation of a positional-control scheme for a limited angle torque motor, using a digital signal processor (DSP).

There is a growing demand for high-speed arithmetic co-processors for use in applications with computationally intensive tasks. For instance, Fast Fourier Transform (FFT) co-processors are used in real-time multimedia services and financial applications use decimal co-processors to perform large amounts of decimal computations. Using redundant number systems to eliminate word-wide carry propagation within interim operations is a well-known technique to increase the speed of arithmetic hardware units. Redundant number systems are mostly useful in applications where many consecutive arithmetic operations are performed prior to the final result, making it advantageous for arithmetic co-processors. This thesis discusses the implementation of two popular arithmetic co-processors based on redundant number systems: namely, the binary FFT co-processor and the decimal arithmetic co-processor. FFT co-processors consist of several consecutive multipliers and adders over complex numbers. FFT architectures are implemented based on fixed-point and floating-point arithmetic. The main advantage of floating-point over fixed-point arithmetic is the wide dynamic range it introduces. Moreover, it avoids numerical issues such as scaling and overflow/underflow concerns at the expense of higher cost. Furthermore, floating-point implementation allows for an FFT co-processor to collaborate with general purpose processors. This offloads computationally intensive tasks from the primary processor. The first part of this thesis, which is devoted to FFT co-processors, proposes a new FFT architecture that uses a new Binary-Signed Digit (BSD) carry-limited adder, a new floating-point BSD multiplier and a new floating-point BSD three-operand adder. Finally, a new unit labeled as Fused-Dot-Product-Add (FDPA) is designed to compute AB+CD+E over floating-point BSD operands. The second part of the thesis discusses decimal arithmetic operations implemented in hardware using redundant number systems. These operations are popularly used in decimal floating-point co-processors. A new signed-digit decimal adder is proposed along with a sequential decimal multiplier that uses redundant number systems to increase the operational frequency of the multiplier. New redundant decimal division and square-root units are also proposed. The architectures proposed in this thesis were all implemented using Hardware-Description-Language (Verilog) and synthesized using Synopsys Design Compiler. The evaluation results prove the speed improvement of the new arithmetic units over previous pertinent works. Consequently, the FFT and decimal co-processors designed in this thesis work with at least 10% higher speed than that of previous works. These architectures are meant to fulfill the demand for the high-speed co-processors required in various applications such as multimedia services and financial computations.

Thesis (M.S.)--Rochester Institute of Technology, 2007. / Typescript. Includes bibliographical references (leaves 58-60).

The continued growth of the computational capability of throughput processors has made throughput processors the platform of choice for a wide variety of high performance computing applications. Graphics Processing Units (GPUs) are a prime example of throughput processors that can deliver high performance for applications ranging from typical graphics applications to general-purpose data parallel (GPGPU) applications. However, this success has been accompa- nied by new performance bottlenecks throughout the memory hierarchy of GPU-based systems. This dissertation identifies and eliminates performance bottlenecks caused by major sources of interference throughout the memory hierarchy. Specifically, we provide an in-depth analysis of inter- and intra-application as well as inter- address-space interference that significantly degrade the performance and efficiency of GPU-based systems. To minimize such interference, we introduce changes to the memory hierarchy for systems with GPUs that allow the memory hierarchy to be aware of both CPU and GPU applications’ charac- teristics. We introduce mechanisms to dynamically analyze different applications’ characteristics and propose four major changes throughout the memory hierarchy. First, we introduce Memory Divergence Correction (MeDiC), a cache management mecha- nism that mitigates intra-application interference in GPGPU applications by allowing the shared L2 cache and the memory controller to be aware of the GPU’s warp-level memory divergence characteristics. MeDiC uses this warp-level memory divergence information to give more cache space and more memory bandwidth to warps that benefit most from utilizing such resources. Our evaluations show that MeDiC significantly outperforms multiple state-of-the-art caching policies proposed for GPUs. Second, we introduce the Staged Memory Scheduler (SMS), an application-aware CPU-GPU memory request scheduler that mitigates inter-application interference in heterogeneous CPU-GPU systems. SMS creates a fundamentally new approach to memory controller design that decouples the memory controller into three significantly simpler structures, each of which has a separate task, These structures operate together to greatly improve both system performance and fairness. Our three-stage memory controller first groups requests based on row-buffer locality. This grouping allows the second stage to focus on inter-application scheduling decisions. These two stages en- force high-level policies regarding performance and fairness. As a result, the last stage is simple logic that deals only with the low-level DRAM commands and timing. SMS is also configurable: it allows the system software to trade off between the quality of service provided to the CPU versus GPU applications. Our evaluations show that SMS not only reduces inter-application interference caused by the GPU, thereby improving heterogeneous system performance, but also provides better scalability and power efficiency compared to multiple state-of-the-art memory schedulers. Third, we redesign the GPU memory management unit to efficiently handle new problems caused by the massive address translation parallelism present in GPU computation units in multi- GPU-application environments. Running multiple GPGPU applications concurrently induces significant inter-core thrashing on the shared address translation/protection units; e.g., the shared Translation Lookaside Buffer (TLB), a new phenomenon that we call inter-address-space interference. To reduce this interference, we introduce Multi Address Space Concurrent Kernels (MASK). MASK introduces TLB-awareness throughout the GPU memory hierarchy and introduces TLBand cache-bypassing techniques to increase the effectiveness of a shared TLB. Finally, we introduce Mosaic, a hardware-software cooperative technique that further increases the effectiveness of TLB by modifying the memory allocation policy in the system software. Mosaic introduces a high-throughput method to support large pages in multi-GPU-application environments. The key idea is to ensure memory allocation preserve address space contiguity to allow pages to be coalesced without any data movements. Our evaluations show that the MASK-Mosaic combination provides a simple mechanism that eliminates the performance overhead of address translation in GPUs without significant changes to GPU hardware, thereby greatly improving GPU system performance. The key conclusion of this dissertation is that a combination of GPU-aware cache and memory management techniques can effectively mitigate the memory interference on current and future GPU-based systems as well as other types of throughput processors.