摘要： 區(qū)域氣候模式CWRF(Climate-Weather Research and Forecasting model)是國(guó)家氣候中心區(qū)域氣候預(yù)測(cè)系統(tǒng)的重要組成部分，也是系統(tǒng)最耗時(shí)的程序。高性能計(jì)算是提高CWRF數(shù)值預(yù)報(bào)計(jì)算性能的關(guān)鍵技術(shù)，開(kāi)展CWRF模式在國(guó)產(chǎn)神威眾核架構(gòu)上的移植和優(yōu)化，提高模式的模擬效率，對(duì)模式的擴(kuò)展、開(kāi)發(fā)能力和可持續(xù)發(fā)展具有重要意義?；趪?guó)產(chǎn)眾核SW26010處理器，完成了CWRF區(qū)域氣候模式的移植、性能分析和深入性能優(yōu)化，采用訪存優(yōu)化、Cache命中率優(yōu)化及眾核加速優(yōu)化等方法，對(duì)CWRF模式動(dòng)力過(guò)程、物理過(guò)程和I/O過(guò)程計(jì)算代碼進(jìn)行重構(gòu)及眾核加速。結(jié)果表明：優(yōu)化技術(shù)可使CWRF動(dòng)力過(guò)程平均加速2倍，最高加速6.4倍，物理過(guò)程平均加速1.7倍，最高加速5.4倍，I/O過(guò)程加速1.2倍，程序整體最高加速1.4倍，計(jì)算誤差在合理范圍內(nèi)。
中圖分類號(hào)： TP391 文獻(xiàn)標(biāo)識(shí)碼： A DOI：10.16157/j.issn.0258-7998.212397 中文引用格式： 呂小敬，劉釗，蔡蕙伊，等. 面向國(guó)產(chǎn)神威眾核架構(gòu)的區(qū)域氣候模式CWRF性能優(yōu)化技術(shù)[J].電子技術(shù)應(yīng)用，2022，48(1)：31-38. 英文引用格式： Lv Xiaojing，Liu Zhao，Cai Huiyi，et al. Optimization technology for regional climate model-CWRF based on domestic Sunway many-core architecture[J]. Application of Electronic Technique，2022，48(1)：31-38.
Optimization technology for regional climate model-CWRF based on domestic Sunway many-core architecture
Lv Xiaojing1，2，Liu Zhao2，3，Cai Huiyi2，Li Jinwei2
1.China Ship Scientific Research Center，Wuxi 214000，China； 2.National Supercomputing Center in Wuxi，Wuxi 214000，China；3.Tsinghua University，Beijing 100080，China
Abstract： CWRF(Climate-Weather Research and Forecasting model) is a component of the regional climate prediction system built in the National Climate Center, and consumes the largest proportion of time. High performance computing is a key technology used to improve the compactional performance of CWRF. Carrying out the configuration and optimization of the CWRF model based on the domestic Sunway many-core system, improving the simulation efficiency are of great significance for the speedup, as well as the development capability and sustainable development of the model. This paper completed the configuration and performance evaluation of CWRF based on the SW26010 many-core architecture. Memory access optimization, Cache hit rate optimization, many-core acceleration models are introduced to speedup CWRF relating to the dynamic-core process, physical process and I/O process. The results show that the average speed of the dynamic process is 2 times and the highest speed is 6.4 times, the average speed of the physical process is 1.7 times and the highest speed is 5.4 times, the I/O process speeds up 1.2 times, the overall program speeds up to 1.4 times, and the calculation error is reasonable.
Key words : CWRF；high performance computing；Sunway；SW26010
