Manycore Research Innovation and Opportunities Network (MaRIONet)

众核研究创新和机会网络 (MaRIONet)

• 批准号: EP/P006434/1
• 负责人: Jeremy Singer
• 金额: $ 11.67万
• 依托单位: University of Glasgow
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2016
• 资助国家: 英国
• 起止时间: 2016 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FP006434%2F1
• 关键词: Manycore; Research; Innovation; Opportunities; Network

The transition to manycore systems is arguably the greatest challengefacing the ICT industry, Computational Science, and Computing Scienceresearch. The challenge is globally recognised, and is an EPSRC ICT priority:Manycore Architectures in Distributed and Embedded Systems (MACDES).The proposed network will identify and bring together a critical massof UK manycore researchers. The manycore revolution is fundamentally changing multiple levels ofthe computation execution stack: from processor architecture, throughsystems software, to applications. The UK has world leadingresearchers and industrialists working on specific technologies ateach level, but currently researchers and engineers do not collaborateadequately on the complex manycore challenges. Furthermore,meaningful industry engagement is limited, despite the large number ofUK-based industrial research labs with strategic interest in manycoresystems.We will foster effective communication between academic and industrialresearch teams both at different levels in the execution stack(e.g. compiler developers targeting an emergent architecture), andwithin each level (e.g. program analysis applied to improve memoryaccess latency in the systems layer). The teams will benefit by beingable to identify and exploit synergies and opportunities forcross-pollination.We will establish well defined, highly available channels forcommunication and low-overhead mechanisms for collaboration. Giventhat many world-class UK researchers are isolated in relatively smalldepartments and research groups, the network will enable nationalsharing of ideas, information and infrastructure. Community supportand resource pooling will enable UK researchers to achieve greateracademic impact globally, competing effectively with larger,well-resourced groups in the USA (e.g. Aspire Lab at Berkeley), China(e.g. State Key Lab of Computer System and Architecture, ChineseAcademy of Sciences), and elsewhere.Both architectural innovation and systems software development aremassive investments, often involving person-decades ofeffort. MaRIONet will enable sharing of expertise and infrastructureto lower the barrier to entry for UK systems researchers working inthis area. Industrial network partners will gain exposure tocutting-edge research, and ideas to address current problems. Academicpartners gain real-world use cases and grand challenges.At least three of the UK Policy Exchange `Eight great technologies'are directly enabled by the increased scale and speed of future manycore systems and software, namely:* Big Data revolution: enabled by manycore systems for energy-efficient collection, analysis and storage of big data.* Robotics and autonomous systems: enabled by manycore systems for artificial intelligence, planning, image recognition, humanoid motion control. * Synthetic biology: enabled by manycore processing of biological data, in-silico simulation of bio-mechanisms. Hence we anticipate that future UK economic growth and societal well-beingwill be strongly influenced by manycore technology, which is the focus ofour research network.

向多核心系统的过渡可以说是ICT行业、计算科学和计算科学研究面临的最大挑战。这一挑战是全球公认的，是EPSRC ICT的优先事项：分布式和嵌入式系统中的多核架构(MACDES)。拟议的网络将识别并汇聚一大批关键的英国多核研究人员。多核革命正在从根本上改变计算执行堆栈的多个层次：从处理器体系结构到系统软件，再到应用程序。英国拥有世界领先的研究人员和实业家，致力于各个层面的特定技术，但目前研究人员和工程师在复杂的许多核心挑战上没有充分合作。此外，尽管大量英国工业研究实验室对许多核心系统具有战略兴趣，但有意义的行业参与是有限的。我们将促进学术和工业研究团队之间在执行堆栈的不同级别(例如，针对紧急体系结构的编译器开发人员)和每个级别(例如，应用程序分析来改善系统层中的存储器访问延迟)之间的有效沟通。这些团队将受益于能够识别和利用协同效应和异花授粉的机会。我们将建立定义良好、高度可用的沟通渠道和低开销的合作机制。鉴于许多世界级的英国研究人员被隔离在相对较小的部门和研究小组中，该网络将使全国共享想法、信息和基础设施。社区支持和资源共享将使英国研究人员在全球范围内获得更大的学术影响力，有效地与美国(例如伯克利的Aspire实验室)、中国(例如中国科学院计算机系统与体系结构国家重点实验室)等更大、资源丰富的团队竞争。架构创新和系统软件开发都是巨大的投资，往往涉及数十年的人。Marionet将使专业知识和基础设施共享成为可能，从而降低在这一领域工作的英国系统研究人员的准入门槛。工业网络合作伙伴将接触到尖端研究和解决当前问题的想法。学术合作伙伴获得了现实世界中的用例和巨大的挑战。英国政策交易所至少有三项伟大的技术直接得益于未来多核心系统和软件的规模和速度的增加，即：*大数据革命：由多核心系统实现对大数据的高能效收集、分析和存储。*机器人和自主系统：由多核心系统实现，用于人工智能、规划、图像识别和人形运动控制。*合成生物学：通过对生物数据的多核处理，实现对生物机制的电子模拟。因此，我们预计英国未来的经济增长和社会福祉将受到许多核心技术的强烈影响，这也是我们研究网络的重点。

项目成果

Improving GHC Haskell NUMA profiling

改进 GHC Haskell NUMA 分析

• DOI: 10.1145/3471873.3472974
• 发表时间: 2021
• 作者: MacGregor R

Generic Exact Combinatorial Search at HPC Scale

HPC 规模的通用精确组合搜索

• DOI: 10.1007/s10766-022-00744-3
• 发表时间: 2022
• 期刊: International Journal of Parallel Programming
• 影响因子: 1.5
• 作者: MacGregor R

Could Tierless Languages Reduce IoT Development Grief?

无分层语言可以减少物联网开发的痛苦吗？

• DOI: 10.1145/3572901
• 发表时间: 2023
• 期刊: ACM Transactions on Internet of Things
• 作者: Lubbers M

Towards Generic Scalable Parallel Combinatorial Search

• DOI: 10.1145/3115936.3115942
• 发表时间: 2017-07
• 期刊: Proceedings of the International Workshop on Parallel Symbolic Computation
• 作者: B. Archibald;Patrick Maier;Robert J. Stewart;P. Trinder;J. Beule

Replicable parallel branch and bound search

• DOI: 10.1016/j.jpdc.2017.10.010
• 发表时间: 2017-03
• 期刊: J. Parallel Distributed Comput.
• 作者: B. Archibald;Patrick Maier;Ciaran McCreesh;Robert J. Stewart;P. Trinder