Programming Abstractions and Synthesis-Aided Compilation for Emerging Computing Platforms

摘要: Author(s): Phothilimthana, Phitchaya | Advisor(s): Bodik, Rastislav; Yelick, Katherine Abstract: Today's cutting-edge applications, ranging from wearable devices and embedded medical sensors to high-performance data centers, put new demands on computer architectures. Those include more computation capability, a tight power budget, low latency, high throughput, many more. To meet these requirements, specialized architectures with energy consumption are becoming prevalent. Many of trade off programmability features for gains in efficiency performance. Hence, challenges inevitable as applications continue evolve make computing architectures.I propose key principles improving intended application writers well compiler developers language designers. First, I address issues by providing programming model that hides low-level details but sufficiently exposes essential control. Second, compile optimize programs, apply compilation methodology based synthesis. Unlike classical compiler's transformation, synthesis obtains correct optimal solution searching an candidate is semantically equivalent specification program. This search helps compilers generate efficient code without deriving program via sequence transformations, which challenging design unconventional architectures.In this thesis, demonstrate the three projects: Chlorophyll, low-power spatial architectures; Floem, system NIC-accelerated center applications; GreenThumb, framework building superoptimizer (an assembly optimizer synthesis).