Energise: Refactorings and Skeletons for Energy-Aware Applications on High-Performance Embedded Systems

Energise：高性能嵌入式系统上能源感知应用的重构和骨架

• 批准号: EP/V006290/1
• 负责人: Christopher Brown
• 金额: $ 48.78万
• 依托单位: University of St Andrews
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2021
• 资助国家: 英国
• 起止时间: 2021 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FV006290%2F1
• 关键词: Energise; Refactorings; Skeletons; Energy; Aware

Non-functional properties, including, for example, energy, are becoming increasingly and critically important for programming embedded devices. These devices, typically rely heavily on battery power, and have restrictive constraints (or energy budgets) within which to operate. Devices such as tablets, phones, drones, medical devices and the Internet of Things, all of which are becoming ever more commonplace in society, are driving the pressing demand for optimisation of energy usage in applications that execute on them.Furthermore, as the global climate change crisis becomes critical, our carbon footprints are accelerating at an alarming rate, partly due to increased energy consumption from computing devices in the modern world. New green-computing techniques are needed in order to reduce the overall energy consumption of computing devices, without decreasing their overall performance. Despite this, dealing with energy consumption is often treated as a secondary concern, and when addressed, something of a black-art to the average developer, where energy budgets are often met by developers randomly changing source code or applying optimisations in the hope that energy budgets will be achieved indirectly. Furthermore, these embedded devices are becoming increasingly parallel, with multi-core hardware now commonplace in e.g. smart phones and tablets. However, dealing with energy properties at the language level usually requires developers to have highly specialised skills, and use low-level tools and techniques.Even small devices, such as the Raspberry Pi and the Jetson Nano, offer high-performance low-energy models of hardware, at very affordable prices. Parallel hardware can offer hardware manufacturers a route to low-energy consumption, as multi-core typically increases the performance of the software, while also decreasing its energy usage.Consequently, there is a very clear and timely need to provide the typical non-specialist programmer with the necessary software development tools and programming abstractions required to develop applications that are able to conform to specified energy requirements, thereby making software developers more lean, agile and productive; and software, and the devices they run on, more greener. To date, there has been very little effort on the fundamental and essential problems of:- providing suitable abstractions to the programmer, allowing them to program with energy as a target goal; and,- determining how to correctly transform code structure so that it yields optimal energy results, or targets the software to stay within a pre-defined energy budget.The Energise project will directly tackle these fundamental problems by building on and complementing current research into both program transformation and algorithmic skeletons. Unusually, and very importantly, \TheProject{} \emph{tackles the energy crisis from a software-engineering perspective}, employing software-development techniques such as refactoring, and high-level programming abstractions using novel energy-optimised skeletons.This is a critical and fundamental goal to solving the energy ---and, therefore, the green-computing--- crisis in general.

非功能属性（例如能量）对于嵌入式设备编程变得越来越重要。这些设备通常严重依赖电池电源，并且运行时具有严格的约束（或能量预算）。平板电脑、手机、无人机、医疗设备和物联网等设备在社会中变得越来越普遍，推动了对在其上执行的应用程序中优化能源使用的迫切需求。此外，随着全球气候变化危机变得严峻，我们的碳足迹正在以惊人的速度加速，部分原因是现代世界计算设备的能源消耗增加。需要新的绿色计算技术来降低计算设备的整体能耗，而不降低其整体性能。尽管如此，处理能源消耗通常被视为次要问题，一旦解决，对普通开发人员来说就像是魔法，开发人员经常随机更改源代码或应用优化来满足能源预算，以期间接实现能源预算。此外，这些嵌入式设备变得越来越并行，多核硬件现在在计算机等领域很常见。智能手机和平板电脑。然而，在语言级别处理能源属性通常需要开发人员具备高度专业化的技能，并使用低级工具和技术。即使是小型设备，例如 Raspberry Pi 和 Jetson Nano，也能以非常实惠的价格提供高性能低能耗硬件模型。并行硬件可以为硬件制造商提供一条实现低能耗的途径，因为多核通常可以提高软件的性能，同时还可以降低其能耗。因此，非常明确且及时地需要为典型的非专业程序员提供必要的软件开发工具和编程抽象，以开发能够符合指定能源要求的应用程序，从而使软件开发人员更加精简、敏捷和高效；软件及其运行的设备更加环保。迄今为止，在以下基本和本质问题上几乎没有做出任何努力： - 向程序员提供合适的抽象，使他们能够以精力为目标进行编程； - 确定如何正确地转换代码结构，以产生最佳的能源结果，或使软件保持在预先定义的能源预算内。Energize 项目将通过建立和补充当前对程序转换和算法框架的研究来直接解决这些基本问题。不同寻常但非常重要的是，\TheProject{} \emph{从软件工程的角度解决能源危机}，采用重构等软件开发技术，以及使用新型能源优化框架的高级编程抽象。这是解决能源危机（进而解决绿色计算危机）的关键和基本目标。

项目成果

Towards a refactoring tool for dependently-typed programs

面向依赖类型程序的重构工具

• 发表时间: 2022
• 作者: Brown C

COMPROF and COMPLACE: shared-memory communication profiling and automated thread placement via dynamic binary instrumentation

COMPROF 和 COMPLACE：通过动态二进制检测进行共享内存通信分析和自动线程放置

• 发表时间: 2023
• 作者: Kirckpatrick R

Semi-automatic ladderisation: improving code security through rewriting and dependent types

• DOI: 10.1145/3498886.3502202
• 发表时间: 2022-01
• 期刊: Proceedings of the 2022 ACM SIGPLAN International Workshop on Partial Evaluation and Program Manipulation
• 作者: Christopher Brown;Adam D. Barwell;Yoann Marquer;Olivier Zendra;Tania Richmond;Chen Gu

Proving Renaming for Haskell via Dependent Types: A Case-Study in Refactoring Soundness Extended Abstract

通过依赖类型证明 Haskell 的重命名：重构稳健性的案例研究扩展摘要

• 发表时间: 2021
• 作者: A D Barwell