标题： 使用CUDA和OpenACC加速不连续伽辽金浅水方程求解器（DG-SWEM）的GPU加速 显示英文标题

标题： GPU-acceleration of the Discontinuous Galerkin Shallow Water Equations Solver (DG-SWEM) using CUDA and OpenACC

摘要： 本文介绍了将DG-SWEM（一种用于海岸洋流，特别是风暴潮的不连续伽辽金求解器）移植到GPU的两种不同方法：CUDA Fortran和OpenACC。 显式时间不连续伽辽金方法由于单元之间的耦合较弱，表现出大量的数据并行性，因此自然适合映射到GPU架构。 对于每种移植方法，我们讨论了代码设计、编程的难易程度以及在实际使用案例中的性能。 特别是对于OpenACC版本，我们还旨在通过使用统一内存在同一代码库中保持可维护性。 我们在NVIDIA的Grace Hopper芯片上测试了这些代码，并将每个节点的GPU性能与单个CPU节点（144个核心）上的MPI版本进行了比较。 显示英文摘要

摘要： This paper presents a porting of DG-SWEM, a discontinuous Galerkin solver for coastal ocean circulation, and in particular storm surge, to GPU using two separate approaches: CUDA Fortran and OpenACC. Time-explicit discontinuous Galerkin methods have been shown to exhibit a large amount of data parallelism due to the loose coupling between elements, and thus are naturally mapped to the GPU architecture. For each porting approach, we discuss the code design, ease of programming, and performance when running on realistic use cases. Specifically for the OpenACC version, we also aim to preserve maintainability within the same codebase through using Unified Memory. We test the codes on NVIDIA's Grace Hopper chip and compare the GPU performance on each node to the MPI version on a single CPU node (144 cores).