期刊
出版社
IEEE
DOI: 10.1109/SC41404.2022.00078
关键词
Numerical Weather Prediction; Python; Data-Centric Programming
类别
资金
- European Research Council under the European Union [101002047]
- EuroHPC-JU [955606, 955513]
- Horizon 2020 programme
- PASC program (Platform for Advanced Scientific Computing)
- Swiss National Science Foundation [185778]
- Swiss National Supercomputing Centre (CSCS) [s1053]
- Allen Institute for Artificial Intelligence (AI2)
- Vulcan Inc.
- European Research Council (ERC) [101002047] Funding Source: European Research Council (ERC)
This study presents an optimization approach for weather and climate applications, achieving significant performance improvements by abstracting hardware details and utilizing optimization techniques.
Earth system models are developed with a tight coupling to target hardware, often containing specialized code predicated on processor characteristics. This coupling stems from using imperative languages that hard-code computation schedules and layout. We present a detailed account of optimizing the Finite Volume Cubed-Sphere Dynamical Core (FV3), improving productivity and performance. By using a declarative Python-embedded stencil domain-specific language and data-centric optimization, we abstract hardware-specific details and define a semi-automated workflow for analyzing and optimizing weather and climate applications. The workflow utilizes both local and full-program optimization, as well as user-guided fine-tuning. To prune the infeasible global optimization space, we automatically utilize repeating code motifs via a novel transfer tuning approach. On the Piz Daint supercomputer, we scale to 2,400 GPUs, achieving speedups of up to 3.92x over the tuned production implementation at a fraction of the original code.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据