MRI: Acquisition of STEMS: A Laboratory for End-to-End Development of Software and Tools for Emerging Multigrain Supercomputers

该项目为开发多粒度并行体系结构的编程模型和方法提供了试验台，支持在操作系统、运行时系统和支持并行网格生成、可视化和生物信息学应用的网络层开发创新软件技术和服务的领域的多学科研究。所请求的检测基于一个处理器，该处理器使用两个超标量执行核心，将芯片多处理和同步多线程结合在一起，每个执行核心在同一芯片上具有先进的SMT功能。IBM Power5处理器为算法模型、并行编程系统和低级系统软件的真正开发提供了机会，以支持多粒度体系结构上的多级并发。这项研究旨在提高高端系统上的程序员生产率。此外，该基础设施还支持ITR项目，该项目调查使用围绕多线程处理器构建的集群和SMP的并行网格生成算法的多级别并发性的利用，以及一位职业获奖者为多粒度系统构建通用并行编程系统，并使用连续监控基础设施调查计算和布局的运行时适配。此次收购还将使大规模科学可视化和大数据集的并行实时可视化取得进展。此外，它还将为最终用户部署透明的性能监控和虚拟化服务，这些用户可以利用高端集群进行地理位置遥远的计算。广泛影响：这项工作加强了多项协作和协同研究的影响，这些研究与来自医学院(专注于脑成像的网格生成)和计算生物学家(专注于为生命科学分析管道透明访问远程数据集)的合作伙伴托管跨学科研究。该设施还丰富了这个以本科生为主的服务机构的教学环境。科学可视化和医学图像计算方面的新产品将被启用。

This project, forming a test bed for the development of programming models and methodologies for multigrain parallel architectures, supports multidisciplinary research in areas that develop innovative software technology and services at the operating system, runtime system, and network layers supporting parallel mesh generation, visualization, and bioinformatics applications. The requested instrumentation is based on a processor that combines chip multiprocessing and simultaneous multithreading using two superscalar execution cores, each with advanced SMT capabilities on the same die. The IBM Power5 processor offers an opportunity for real development of algorithmic models, parallel programming systems, and low-level system software support for multilevel concurrency on multigrain architectures. The research is geared towards increasing programmer productivity on high-end systems. Moreover, the infrastructure supports an ITR project investigating the exploitation of multiple levels of concurrency from parallel mesh generation algorithms using clusters and SMP build around multithreaded processors and a CAREER awardee building generic parallel programming system for multigrain systems and investigating runtime adaptation of computation and layouts using continuous monitoring infrastructure. The acquisition will also enable advances in large-scale scientific visualization and parallel real-time visualization for large datasets. Furthermore, it will enable the deployment of transparent performance monitoring and virtualization services for end users who can utilize high-end clusters for computations distributed over geographically distant locations.Broader Impact: The work strengthens the impact of multiple collaborative and synergistic research hosting interdisciplinary studies with partners from medical schools (focusing on mesh generation for brain imaging) and computational biologists (focusing on transparent access to remote data sets for life science analysis pipelines). The facility also enriches the teaching environment in this mainly undergraduate serving institution. New offerings in scientific visualization and medical image computing will be enabled.