当前位置: X-MOL 学术arXiv.cs.MS › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Preparing Ginkgo for AMD GPUs -- A Testimonial on Porting CUDA Code to HIP
arXiv - CS - Mathematical Software Pub Date : 2020-06-25 , DOI: arxiv-2006.14290
Yuhsiang M. Tsai (1), Terry Cojean (1), Tobias Ribizel (1), Hartwig Anzt (1 and 2) ((1) Karlsruhe Institute of Technology, (2) University of Tennessee, Innovative Computing Lab)

With AMD reinforcing their ambition in the scientific high performance computing ecosystem, we extend the hardware scope of the Ginkgo linear algebra package to feature a HIP backend for AMD GPUs. In this paper, we report and discuss the porting effort from CUDA, the extension of the HIP framework to add missing features such as cooperative groups, the performance price of compiling HIP code for AMD architectures, and the design of a library providing native backends for NVIDIA and AMD GPUs while minimizing code duplication by using a shared code base.

中文翻译:

为 AMD GPU 准备 Ginkgo —— 关于将 CUDA 代码移植到 HIP 的证明

随着 AMD 加强其在科学高性能计算生态系统中的雄心,我们扩展了 Ginkgo 线性代数包的硬件范围,为 AMD GPU 提供 HIP 后端。在本文中,我们报告并讨论了 CUDA 的移植工作、HIP 框架的扩展以添加缺失的功能(例如协作组)、为 AMD 架构编译 HIP 代码的性能价格,以及为 AMD 架构提供本地后端的库的设计。 NVIDIA 和 AMD GPU,同时通过使用共享代码库最大限度地减少代码重复。
更新日期:2020-06-26
down
wechat
bug