Scholarlink

A parallel OpenMP–CUDA software for simulating multicomponent surface water and sediment transport dynamics in a shallow-water model has been developed, targeting hybrid CPU \( + \) multi-GPU platforms without GPUDirect technology. We compare the computational efficiency of two implementations on this platform: one using GPUDirect for direct GPU-to-GPU data transfer and another exchanging data via the host (CPU \( + \) DRAM). Parallelization efficiency is evaluated for different problem types. The model is applied to study riverbed morphodynamics in selected river systems.

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

Efficiency of Parallel Simulations of Coupled Surface Water and Sediment Dynamics in Riverbeds

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Alexander Khoperskov,
  • Ekaterina Agafonnikova,
  • Anna Klikunova,
  • Sergei Khrapov

摘要

A parallel OpenMP–CUDA software for simulating multicomponent surface water and sediment transport dynamics in a shallow-water model has been developed, targeting hybrid CPU \( + \) multi-GPU platforms without GPUDirect technology. We compare the computational efficiency of two implementations on this platform: one using GPUDirect for direct GPU-to-GPU data transfer and another exchanging data via the host (CPU \( + \) DRAM). Parallelization efficiency is evaluated for different problem types. The model is applied to study riverbed morphodynamics in selected river systems.