Graphical Ray Casting Representation of Directed Acyclic Graph in Critical Path Detection

• Item type: SIG Technical Reports(1)
• 公開日: 2025-07-28
• タイトル / 言語: ja
• タイトル: Graphical Ray Casting Representation of Directed Acyclic Graph in Critical Path Detection
• タイトル / 言語: en
• タイトル: Graphical Ray Casting Representation of Directed Acyclic Graph in Critical Path Detection
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: Graph / Training
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: RIKEN
• 著者所属: RIKEN
• 著者所属: RIKEN
• 著者所属: RIKEN／Institute of Science Tokyo
• 著者所属: RIKEN
• 著者所属: Kyoto University
• 著者所属: RIKEN／Institute of Science Tokyo
• 著者所属: RIKEN
• 著者所属: RIKEN
• 著者所属(英): en
• 著者所属(英): RIKEN
• 著者所属(英): en
• 著者所属(英): RIKEN
• 著者所属(英): en
• 著者所属(英): RIKEN
• 著者所属(英): en
• 著者所属(英): RIKEN / Institute of Science Tokyo
• 著者所属(英): en
• 著者所属(英): RIKEN
• 著者所属(英): en
• 著者所属(英): Kyoto University
• 著者所属(英): en
• 著者所属(英): RIKEN / Institute of Science Tokyo
• 著者所属(英): en
• 著者所属(英): RIKEN
• 著者所属(英): en
• 著者所属(英): RIKEN
• 著者名: Zhengyang,Bai; Peng,Chen; Lingqi,Zhang; Chen,Zhuang; Tenindra,Abeywickrama; Jing,Xu; Du,Wu; Emil,Vatai; Mohamed,Wahib
• 著者名(英): Zhengyang Bai; Peng Chen; Lingqi Zhang; Chen Zhuang; Tenindra Abeywickrama; Jing Xu; Du Wu; Emil Vatai; Mohamed Wahib
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Directed acyclic graphs (DAGs), where all edges are directed and the graph is cycle-free, play an essential role in various domains, such as phylogenetic networks in biology, causal inference in social science, and cutting-edge technologies like task scheduling in large-scale computer systems. However, most traditional DAG algorithms are complex and difficult to accelerate because of their irregular data access patterns and control-flow divergency that do not align well with GPUs' SIMT (Single Instruction, Multiple Threads) architecture. However, recent advances in GPU architecture offer new opportunities to overcome these issues by ray tracing cores, which were originally designed for rendering tasks but have enhanced GPU's ability to perform divergences that can be efficiently applied to DAG-related problems. In this work, we present a novel approach to modeling DAG problems as rendering tasks. We employ GPU ray tracing cores to efficiently explore the complex relationships within DAGs. In addition, we propose a general workflow for processing this graphical ray casting representation and demonstrate its practical implementation via a case study on critical path detection.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Directed acyclic graphs (DAGs), where all edges are directed and the graph is cycle-free, play an essential role in various domains, such as phylogenetic networks in biology, causal inference in social science, and cutting-edge technologies like task scheduling in large-scale computer systems. However, most traditional DAG algorithms are complex and difficult to accelerate because of their irregular data access patterns and control-flow divergency that do not align well with GPUs' SIMT (Single Instruction, Multiple Threads) architecture. However, recent advances in GPU architecture offer new opportunities to overcome these issues by ray tracing cores, which were originally designed for rendering tasks but have enhanced GPU's ability to perform divergences that can be efficiently applied to DAG-related problems. In this work, we present a novel approach to modeling DAG problems as rendering tasks. We employ GPU ray tracing cores to efficiently explore the complex relationships within DAGs. In addition, we propose a general workflow for processing this graphical ray casting representation and demonstrate its practical implementation via a case study on critical path detection.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10463942
• 書誌情報: 研究報告ハイパフォーマンスコンピューティング（HPC）; 巻 2025-HPC-200, 号 15, p. 1-7, 発行日 2025-07-28
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8841
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会