Code Modernization Optimization of Ocean General Circulation Model NEMO

The ocean general circulation model (OGCM) is the key tool for ocean environment simulation and forecast. With the ocean resolution finer, the demand for improving computational performance is more and more urgent. To improve the calculation performance of OGCM by taking full advantage of modern computers, a code modernization optimization scheme is carried out in this paper using an OGCN named NEMO as an example. The Intel performance analysis tool is used to evaluate the computing performance of the model at first. Then, several optimization steps, which are compiler options, serial and scalar optimization, SIMD, memory bandwidth optimization and extending to multicores, are applied to hotspot functions. After optimization, the model^ʼs overall performance can be improved by 31% without increasing any hardware cost and load balance have a good performance in multiprocess. The results indicate that the optimization strategy used in this study is very effective and useful. Furthermore, the problems that significantly affect the computational efficiency in the model, such as the heavy use of pointers that can prevent vectorization, multiple loop nesting, high memory bandwidth usage, are discussed in this paper, to provide reference and suggestion for the OGCM in the future design and improvement.