Search

The slowing of CMOS technology scaling mismatches the ever-increasing demand for computational power, leading to a rise in the use of heterogeneous systems, which pair scalar processors such as CPUs with specialized accelerators like FPGAs and GPUs. These systems enable continued performance and efficiency scaling for specialized tasks while retaining limited generality. This restricted generality inherent in heterogeneous platforms requires specialized knowledge of hardware architectures and low-level programming models, posing a substantial barrier to software developers.This dissertation addresses the challenges software developers face in leveraging heterogeneous computing resources, particularly FPGA acceleration. We identify three major limitations: limited programmability support in domain-specific resources, difficulty in achieving high performance and efficiency, and time-consuming porting across diverse computational architectures. We present novel approaches and tools to bridge the gap between high-level software development and efficient hardware implementation, making heterogeneous computing more accessible to a broader range of developers.In this dissertation, we introduce Heterosys, an end-to-end optimization framework simplifying heterogeneous hardware development. It decouples algorithmic descriptions from underlying fabrics and offers layout-driven and architecture-driven design generation, bridging the gap between high-level designs and hardware details.The frontend of Heterosys is HeteroRefactor, which combines dynamic invariant analysis, automated refactoring, and selective offloading. HeteroRefactor optimizes software kernels onto accelerators for common-case inputs while maintaining correctness through CPU fallback mechanisms. HeteroRefactor automatically refactors software code to make it FPGA-compatible and hardware-friendly, reducing chip resource usage through bitwidth optimization and floating-point precision tuning.From the individual synthesizable hardware kernels, Adroit optimizes them using a static approach to identify data and control broadcasts. It analyzes data and control dependencies in the source code and reports, trading off clock-cycle latency for higher frequency. By optimizing the FPGA architecture generated by high-level synthesis tools, Adroit relieves software developers from needing to understand the underlying fabric.As the backend, Heterosys composes multiple kernels into an optimized FPGA system using RapidIR, a comprehensive infrastructure for high-level physical synthesis optimizations. RapidIR integrates coarse-grained floorplanning with high-level pipelining, supporting hierarchical composition of heterogeneous designs from diverse sources. It automates the exploration of various physical optimization strategies, freeing programmers from designing device-specific hardware layouts for each target device.Our research demonstrates substantial performance improvements across diverse applications and benchmarks, including genomic sequencing and large language model accelerations. Our FPGA optimization techniques achieve operating frequency improvements of 30% to over 100% compared to state-of-the-art EDA tools, resource requirement reductions of 21% to over 90%, and 51% code reduction in porting between platforms.This dissertation contributes a comprehensive set of methodologies and tools that significantly lower the barriers to entry for heterogeneous computing, particularly FPGA acceleration. By abstracting away much of the hardware complexity, our work paves the way for broader adoption of heterogeneous acceleration in software development practices, potentially driving research innovation and performance improvements across a wide range of applications and industries.

Dense matrix multiply (MM) serves as one of the most heavily used kernels in deep learning applications. To cope with the high computation demands of these applications, heterogeneous architectures featuring both FPGA and dedicated ASIC accelerators have emerged as promising platforms. For example, the AMD/Xilinx Versal ACAP architecture combines general-purpose CPU cores and programmable logic with AI Engine processors optimized for AI/ML. With 400 AIEs, it provides up to 6.4 TFLOPs performance for 32-bit floating-point data. However, machine learning models often contain both large and small MM operations. While large MM operations can be parallelized efficiently across many cores, small MM operations typically cannot. We observe that executing some small MM layers from the BERT natural language processing model on a large, monolithic MM accelerator in Versal ACAP achieved less than 5% of the theoretical peak performance. Therefore, one key question arises: How can we design accelerators to fully use the abundant computation resources under limited communication bandwidth for applications with multiple MM layers of diverse sizes? We identify the biggest system throughput bottleneck resulting from the mismatch of massive computation resources of one monolithic accelerator and the various MM layers of small sizes in the application. To resolve this problem, we propose the CHARM framework to compose multiple diverse MM accelerator architectures working concurrently towards different layers in one application. We deploy the CHARM framework for four different applications, including BERT, ViT, NCF, MLP, on the AMD Versal ACAP VCK190 evaluation board. Our experiments show that we achieve 1.46 TFLOPs, 1.61 TFLOPs, 1.74 TFLOPs, and 2.94 TFLOPs inference throughput for BERT, ViT, NCF and MLP, which obtain 5.40x, 32.51x, 1.00x and 1.00x throughput gains compared to one monolithic accelerator.

This thesis explores Inuvialuit cultural heritage through the lens of Inuvialuit Pitqusiat Inuusimitkun or living art, a term coined by Iñupiaq/Inuvialuk Elder Pauline Saturgina Tardiff and translated to Sallirmiutun by Inuvialuit Elders Albert and Shirley Elias. Using semi-structured interviews, focus groups, and ethnography, it brings together the knowledge of 11 Inuvialuit artists to discuss Inuvialuit living art through: its ability to tell stories through time and space; its role in surviving and thriving on the land; and its connection to inner “heartwork”. Using a community-based participatory research (CBPR) framework, it outlines the 2019 Inuvialuit Living History Culture Camp at Ivvavik National Park, where living art, photovoice, and participatory design were employed in action with nine Inuvialuit research partners to co-create visual products and documentation of Inuvialuit cultural heritage. This thesis concludes that living art is a vital aspect of Inuvialuit cultural heritage and advocates for continued engagements with living art, musicality, and a heart-centred approach in the future of archaeological research.

Presentation on the Secondary Analysis to Generate Evidence (SAGE) collaborative platform and how it relates to the Accessibility aspect of the FAIR Principles. For the National Data Services Framework Summit 2019.

This resource is a single blog post created as part of the Day of Archaeology initiative. The Day of Archaeology project aimed to provide a window into the daily lives of archaeologists from all over the world. The project asked people working, studying or volunteering in the archaeological world to participate in a 'Day of Archaeology' each year by recording their day and sharing it through text, images or video on the Day of Archaeology blog.The project asked anyone with a personal, professional or voluntary interest in archaeology to get involved, and help highlight the reasons why archaeology is vital to protect the past and inform our futures. The resulting Day of Archaeology project archive demonstrates the wide variety of work the archaeological profession undertakes day-to-day across the globe, and helps to raise public awareness of the relevance and importance of archaeology to the modern world.The first ever Day of Archaeology was held in 2011 and, since then, the project had thousands of archaeologists take part in the event, from those working in the field through to specialists working in laboratories and behind computers. The last Day of Archaeology was on Friday 28th July 2017.

Many animals on earth need to navigate in their own environments for breeding and foraging. The ability to determine a correct heading and the current location for a migratory bird is critical for its survival. Animals that are sensitive to the Earth’s magnetic field can use it to obtain their direction of travel. In 2000, a paper suggested that radical pair reaction could form the basis for magnetoreception in migratory birds and a flavoprotein, cryptochrome, was proposed as the candidate for the radical pair precursor. Recent in vivo experimental results strongly support the hypothesis that radical pairs formed in the eyes of migratory birds are responsible for their magnetic compass sense. Cryptochrome has also been located in the UV-cones in the retinas of two different species of bird. Radical pairs in living cells are influenced mainly by Zeeman interaction, hyperfine interaction, rotational modulation, etc., and together they influence the recombination reactions of the radical pairs. This thesis considers the possible role of radical pairs in avian magnetoreception, using computer simulations of the quantum mechanical evolution of a radical pair under a variety of conditions. Chapter 1 contains the introductions to spin chemistry, avian magnetoreception, and the mathematical description of the quantum evolution of a radical pair. Chapter 2 describes the four different theoretical models for a general non-diffusion-controlled radical pair reaction and the product yields of a radical pair reaction predicted by these four models are analysed and compared. Chapter 3 introduces a model for avian magnetoreception that integrates photoselection with the radical pair reaction and the model is used to predict the retinal patterns that a bird may be able to use for magnetoreception. The anisotropic singlet product yields of a radical pair comprises the flavin chromophore and the tryptophan of a cryptochrome are also presented in this chapter. A paper based on some parts of this chapter is published [1] in the Journal of The Royal Society Interface. Chapter 4 describes a modified version of an algorithm that is used to calculate the product yields detected in a reaction yield detected magnetic resonance (RYDMR) experiment. The new algorithm is used to analyse the results of two sets of RYDMR experiments in which two radical pair systems, pyrene/1,3-dicyanobenzene and chrysene/1,4-dicyanobenzene, were used. The modulated detection technique used in the RYDMR experiments is also discussed in this chapter. A section of this chapter is published [2] in a paper in the journal Physical Chemistry Chemical Physics. [1] Lau, J.C.S.; Rodgers, C.T.; Hore, P.J., J. R. Soc. Interface, 9 (77): 3329–3337, 2012 [2] Wedge, C.J.; Lau, J.C.S.; Ferguson, K.-A.; Norman, S.A.; Hore, P.J.; Timmel, C.R. Phys. Chem. Chem. Phys., 15 (38): 16043–16053, 2013

A technique to investigate the properties and performance of new multi-electron metal/air battery systems is proposed and presented. A method for synthesizing nanoscopic VB2 is presented as well as step-by-step procedure for applying a zirconium oxide coating to the VB2 particles for stabilization upon discharge. The process for disassembling existing zinc/air cells is shown, in addition construction of the new working electrode to replace the conventional zinc/air cell anode with a the nanoscopic VB2 anode. Finally, discharge of the completed VB2/air battery is reported. We show that using the zinc/air cell as a test bed is useful to provide a consistent configuration to study the performance of the high-energy high capacity nanoscopic VB2 anode.