Scholarlink

While the FA-BSP model provides significant performance improvements in large-scale graph applications, its single-threaded execution model may limit the performance of certain graph applications. This paper explores the potential benefits of enabling user-level asynchronous tasking with async and finish in FA-BSP programs using distributed triangle counting. The initial results from a generic HPC cluster show that a version using asynchronous tasking leads to a performance increase of 3% to 32%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enabling User-Level Asynchronous Tasking in the FA-BSP Model Case Study: Distributed Triangle Counting

  • 章节

  • 发表日期：

    2025年

查看全文
  • Akihiro Hayashi,
  • Shubhendra Pal Singhal,
  • Youssef Elmougy,
  • Jiawei Yang

摘要

While the FA-BSP model provides significant performance improvements in large-scale graph applications, its single-threaded execution model may limit the performance of certain graph applications. This paper explores the potential benefits of enabling user-level asynchronous tasking with async and finish in FA-BSP programs using distributed triangle counting. The initial results from a generic HPC cluster show that a version using asynchronous tasking leads to a performance increase of 3% to 32%.