CAREER: Exploiting Parallel Heterogeneous Architectures to Enable Time-domain Astronomy in the LSST era

职业：利用并行异构架构实现 LSST 时代的时域天文学

• 批准号: 2042155
• 负责人: Michael Gowanlock
• 金额: $ 41.2万
• 依托单位: Northern Arizona University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-07-15 至 2026-06-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2042155&HistoricalAwards=false
• 关键词: CAREER; Exploiting; Parallel; Heterogeneous; Architectures

Recent and near future scientific instruments will generate large amounts of data. One example of such an instrument is the Vera C. Rubin Observatory that will carry out the Legacy Survey of Space and Time (LSST) over a ten year period. This astronomical survey has the potential to advance many fields of astronomy, and may even lead to the development of new fields of scientific inquiry. However, the large data volume implies that many processors will need to be used to process the data within a reasonable amount of time. This project creates new technologies and algorithms that can utilize a large number of processors. In particular, the project harnesses the power of both standard central processing units (CPUs) and graphics processing units (GPUs) that are good at processing many data items simultaneously. The developed technologies are designed to use the data from LSST and find interesting events in the Solar System. Once an interesting event is detected on a given astronomical object, alerts are sent to the astronomy community so that they can use additional telescopes to further study these objects. Without the technologies developed in this project, astronomers will miss out on opportunities to study transient phenomena. The project integrates several teaching activities that ensure both computer scientists and astronomers receive the necessary training to exploit future generation computer systems. The project includes mentoring undergraduate and graduate students. In addition, the local community will be engaged through outreach activities that promote science, technology, engineering, and mathematical fields, particularly through activities targeting K-12 students. The project serves the national interest, as stated by NSF's mission, by promoting the progress of science, and to advance the national health, prosperity, and welfare. The Vera C. Rubin Observatory will have unprecedented time domain capabilities. However, LSST will generate large volumes of data that need to be examined in order to realize many scientific goals. This project focuses on LSST supporting cyberinfrastructure (CI) in the context of Solar System science. Fast outlier detection is needed to enable rapid follow up by other facilities to ensure that transient events in the Solar System and objects with intrinsically unusual properties are discovered. To ensure rapid detection capabilities, the outlier detection algorithms will exploit heterogeneous CPU and GPU architectures. Furthermore, heterogeneous computing will be employed where the work is distributed between the CPU and GPU. Also, the project examines using application specific integrated circuits on modern GPU hardware, such as tensor and ray tracing cores as applied to a broader range of applications than matrix multiplication and ray tracing. Algorithmic transformations are needed to exploit these heterogeneous processors; consequently, a unifying framework is developed that models the performance of these algorithms as executed on these architectures. This framework and novel parallel and scalable algorithms provide foundational CI that will enable the LSST to successfully explore the Solar System, understand its origins, and identify potentially hazardous asteroids, among other scientific objectives.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

最近和不久的将来，科学仪器将产生大量的数据。这种仪器的一个例子是Vera C。鲁宾天文台将在十年内进行时空遗产调查（LSST）。这项天文调查有可能推动天文学的许多领域，甚至可能导致新的科学探索领域的发展。然而，大数据量意味着需要使用许多处理器来在合理的时间内处理数据。该项目创造了可以利用大量处理器的新技术和算法。特别是，该项目利用了标准中央处理器（CPU）和图形处理器（GPU）的能力，这些处理器擅长同时处理许多数据项。开发的技术旨在使用LSST的数据并发现太阳系中有趣的事件。一旦在某个天体上探测到令人感兴趣的事件，就会向天文学界发出警报，以便他们能够使用更多的望远镜进一步研究这些天体。如果没有在这个项目中开发的技术，天文学家将错过研究瞬态现象的机会。该项目整合了多项教学活动，确保计算机科学家和天文学家都能接受必要的培训，以开发下一代计算机系统。该项目包括指导本科生和研究生。此外，当地社区将通过推广科学、技术、工程和数学领域的外联活动，特别是通过针对K-12学生的活动参与进来。 正如NSF的使命所述，该项目通过促进科学进步，促进国家健康，繁荣和福利，为国家利益服务。维拉C。鲁宾天文台将拥有前所未有的时域能力。然而，LSST将产生大量需要检查的数据，以实现许多科学目标。该项目的重点是在太阳系科学的背景下支持网络基础设施（CI）的LSST。需要快速异常值检测，以便其他设施能够快速跟进，以确保发现太阳系中的瞬态事件和具有内在不寻常特性的物体。 为了确保快速检测能力，异常值检测算法将利用异构CPU和GPU架构。此外，将采用异构计算，其中工作分布在CPU和GPU之间。此外，该项目还研究了在现代GPU硬件上使用专用集成电路，例如适用于比矩阵乘法和光线跟踪更广泛应用的张量和光线跟踪核心。利用这些异构的处理器，因此，需要一个统一的框架开发模型的性能，这些算法上执行这些架构。这个框架和新颖的并行和可扩展的算法提供了基础的CI，这将使LSST能够成功地探索太阳系，了解其起源，并确定潜在的危险小行星，以及其他科学目标。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

项目成果

The Solar System Notification Alert Processing System (SNAPS): Design, Architecture, and First Data Release (SNAPShot1)

太阳系通知警报处理系统 (SNAPS)：设计、架构和首次数据发布 (SNAPShot1)

• DOI: 10.3847/1538-3881/acac7f
• 发表时间: 2023
• 期刊: The Astronomical Journal
• 作者: Trilling, David E.;Gowanlock, Michael;Kramer, Daniel;McNeill, Andrew;Donnelly, Brian;Butler, Nat;Kececioglu, John
• 通讯作者: Kececioglu, John

Leveraging GPU Tensor Cores for Double Precision Euclidean Distance Calculations

利用 GPU 张量核心进行双精度欧几里德距离计算

• DOI: 10.1109/hipc56025.2022.00029
• 发表时间: 2022
• 期刊: and Analytics (HiPC
• 作者: Gallet, Benoit;Gowanlock, Michael
• 通讯作者: Gowanlock, Michael

CUDA-DClust+: Revisiting Early GPU-Accelerated DBSCAN Clustering Designs

CUDA-DClust：回顾早期 GPU 加速的 DBSCAN 集群设计

• DOI: 10.1109/hipc53243.2021.00049
• 发表时间: 2021
• 期刊: and Analytics (HiPC 2021
• 作者: Poudel, Madhav;Gowanlock, Michael
• 通讯作者: Gowanlock, Michael