POSE: Phase I: Toward a Task-Parallel Programming Ecosystem for Modern Scientific Computing

POSE：第一阶段：迈向现代科学计算的任务并行编程生态系统

基本信息

批准号：
2229304
负责人：
Tsung-Wei Huang
金额：
$ 29.88万
依托单位：
University of Utah
依托单位国家：
美国
项目类别：
Standard Grant
财政年份：
2022
资助国家：
美国
起止时间：
2022-09-15 至 2023-10-31
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=2229304&HistoricalAwards=false
关键词：
POSE Phase Toward Task Parallel

项目摘要

Open-source software systems designed for task-parallel programming have become central to a wide range of modern scientific computing applications, such as machine learning and quantum computing. While decades of research has yielded many open-source task-parallel programming systems, most of them are led by a handful of developers and their impacts do not sustain in the long run. To overcome this challenge, this project proposes scoping activities to establish a route to a long-term sustainable ecosystem for task-parallel programming. These activities build atop the open-source software, Taskflow, a high-performance task-parallel system to streamline the building of complex scientific computing applications.This project starts by discovering an ecosystem based on three increasing applications of Taskflow, quantum computing, circuit design automation, and multimedia. The discovery effort consists of designing showcase software products and organizing workshops to pursue the optimal ecosystem for Taskflow. Then, the project designs a series of developer training programs to engage potential content contributors who can help develop and maintain Taskflow in a community-driven fashion. Finally, the project establishes a transparent and publicly visible governance model to formalize the decision-making process of Taskflow for various elements, including technical contributions, partnership, and security policies. This project lays the foundation for Taskflow to transition to a robust and sustainable system asset for the scientific computing community to quickly respond to emerging parallelism using scalable task-based approaches. Many scoping activities in this project can instill confidence in commercial adoption of Taskflow and grow its partnership with other open-source scientific computing projects to further broaden its impact and enhance sustainability. Also, results produced in this project can be used in classrooms to renovate existing learning materials for high-performance computing and software practice, engaging a diverse group of students in open-source development.The project website is available at https://taskflow.github.io/ which contains the latest news, source code, step-by-step learning materials, and benchmarks. The authors are committed to developing and maintaining the project in the long term.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.

专为任务并行编程而设计的开源软件系统已成为各种现代科学计算应用的核心，例如机器学习和量子计算。尽管数十年的研究已经产生了许多开源任务并行编程系统，但其中大多数都是由少数开发人员领导的，从长远来看，它们的影响不会持续。为了克服这一挑战，该项目提出了范围界定活动，以建立一条通往任务并行编程的长期可持续生态系统的路线。这些活动建立在开源软件 Taskflow 之上，Taskflow 是一个高性能任务并行系统，可简化复杂科学计算应用程序的构建。该项目首先发现一个基于任务流、量子计算、电路设计这三个不断增加的应用程序的生态系统自动化、多媒体。发现工作包括设计展示软件产品和组织研讨会，以追求任务流的最佳生态系统。然后，该项目设计了一系列开发人员培训计划，以吸引潜在的内容贡献者，他们可以帮助以社区驱动的方式开发和维护任务流。最后，该项目建立了一个透明且公开可见的治理模型，以正式化任务流各个要素的决策过程，包括技术贡献、合作伙伴关系和安全政策。该项目为任务流过渡到强大且可持续的系统资产奠定了基础，使科学计算社区能够使用可扩展的基于任务的方法快速响应新兴的并行性。该项目中的许多范围界定活动可以增强对任务流商业采用的信心，并加强与其他开源科学计算项目的合作伙伴关系，以进一步扩大其影响并增强可持续性。此外，该项目产生的结果可以在课堂上使用，以更新高性能计算和软件实践的现有学习材料，让不同的学生群体参与开源开发。该项目网站可访问 https://taskflow。 github.io/ 其中包含最新新闻、源代码、分步学习材料和基准测试。作者致力于长期开发和维护该项目。该奖项反映了 NSF 的法定使命，并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，被认为值得支持。

项目成果

期刊论文数量（3）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

GenFuzz: GPU-accelerated Hardware Fuzzing using Genetic Algorithm with Multiple Inputs

GenFuzz：使用具有多个输入的遗传算法进行 GPU 加速的硬件模糊测试

DOI：
发表时间：
2023-07
期刊：
ACM/IEEE Design Automation Conference (DAC
影响因子：
0
作者：
Dian
通讯作者：
Dian

qTask: Task-parallel Quantum Circuit Simulation with Incrementality

qTask：增量任务并行量子电路仿真

DOI：
10.1109/ipdps54959.2023.00080
发表时间：
2022-10-03
期刊：
2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)
影响因子：
0
作者：
Tsung
通讯作者：
Tsung

Fast STA Graph Partitioning Framework for Multi-GPU Acceleration

用于多 GPU 加速的快速 STA 图形分区框架