Scholarlink

Abstract <p>Implementing complex numerical simulation programs in astrophysics, such as particle-in-cell methods or applications like Kulikov et al.’s parallel code for relativistic hydrodynamics (Mathematics <b>10</b>, 1865 (2022)), presents significant challenges requiring both domain expertise for correctness and systems programming skills for performance. Programming tools and automation techniques play crucial roles here. Universal solutions remain unsatisfactory due to the problem’s complexity, prompting exploration of diverse approaches. One such approach is the Active Knowledge Concept—a methodology for automatic parallel program construction developed at Institute of Computational Mathematics and Mathematical Geophysics, Siberian Branch of the Russian Academy of Sciences. This approach enables creation of a subject domain knowledge base containing formal descriptions of domain specifics, including performance-critical characteristics and manually optimized solutions developed using existing parallel programming tools. The presence of such an active knowledge base significantly simplifies automatic construction of parallel programs from high-level specifications. For users, this means describing a problem at high abstraction levels while obtaining automatically generated efficient parallel solver. As a methodology, practical application of the Active Knowledge concept requires developing domain-specific technological facilities, system algorithms, and tools. This paper presents one such facility—the Didal (Distributed Data Library) system—applied to implement an astrophysical simulation based on Kulikov et al.’s piecewise parabolic method (Mathematics <b>10</b>, 1865 (2022)). This approach yielded substantially higher program efficiency compared to existing implementations.</p>

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

High Performance Implementation of a Relativistic Flows Simulation Application Using Active Knowledge Concept

  • 期刊论文

  • 发表日期：

    2026年1月1日

查看全文
  • G. A. Schukin,
  • V. E. Malyshkin,
  • V. A. Perepelkin

摘要

Abstract

Implementing complex numerical simulation programs in astrophysics, such as particle-in-cell methods or applications like Kulikov et al.’s parallel code for relativistic hydrodynamics (Mathematics 10, 1865 (2022)), presents significant challenges requiring both domain expertise for correctness and systems programming skills for performance. Programming tools and automation techniques play crucial roles here. Universal solutions remain unsatisfactory due to the problem’s complexity, prompting exploration of diverse approaches. One such approach is the Active Knowledge Concept—a methodology for automatic parallel program construction developed at Institute of Computational Mathematics and Mathematical Geophysics, Siberian Branch of the Russian Academy of Sciences. This approach enables creation of a subject domain knowledge base containing formal descriptions of domain specifics, including performance-critical characteristics and manually optimized solutions developed using existing parallel programming tools. The presence of such an active knowledge base significantly simplifies automatic construction of parallel programs from high-level specifications. For users, this means describing a problem at high abstraction levels while obtaining automatically generated efficient parallel solver. As a methodology, practical application of the Active Knowledge concept requires developing domain-specific technological facilities, system algorithms, and tools. This paper presents one such facility—the Didal (Distributed Data Library) system—applied to implement an astrophysical simulation based on Kulikov et al.’s piecewise parabolic method (Mathematics 10, 1865 (2022)). This approach yielded substantially higher program efficiency compared to existing implementations.