Hardy David J, Choi Jaemin, Jiang Wei, Tajkhorshid Emad
University of Illinois at Urbana-Champaign, Beckman Institute for Advanced Science and Technology, Urbana, Illinois, USA.
University of Illinois at Urbana-Champaign, Department of Computer Science, Urbana, Illinois, USA.
Int Workshop OpenCL. 2022 May;2022. doi: 10.1145/3529538.3529560. Epub 2022 May 10.
HPC applications have a growing need to leverage heterogeneous computing resources with a vendor-neutral programming paradigm. Data Parallel C++ is a programming language based on open standards SYCL, providing a vendor-neutral solution. We describe our experiences porting the NAMD molecular dynamics application with its GPU-offload force kernels to SYCL/DPC++. Results are shown that demonstrate correctness of the porting effort.
高性能计算(HPC)应用对利用异构计算资源并采用与供应商无关的编程范式的需求日益增长。数据并行C++是一种基于开放标准SYCL的编程语言,提供了一种与供应商无关的解决方案。我们描述了将带有GPU卸载力内核的NAMD分子动力学应用移植到SYCL/DPC++的经验。结果表明了移植工作的正确性。