GPU Acceleration of the Horizontal Diffusion Method in the Weather Research and Forecasting (WRF) Model

Abstract

The Weather Research and Forecasting (WRF) is a next-generation mesoscale numerical weather prediction system. It is designed with a dual purpose, forecasting and research. The WRF software infrastructure consists of a number of components such as dynamic solvers and physical simulation modules. Dynamic solvers are intensive computational components of the WRF model. In this paper, the Horizontal Diffusion method, which is part of the ARW (Advanced Research WRF) dynamic solver, is accelerated using GPUs. The performance of the GPU-based method was compared to that one of a CPU-based single-threaded counterpart on a computational domain of 433x308 horizontal grid points with 35 vertical levels. Thus, the achieved speedup is 19x on a NVIDIA Tesla M2090, without considering data I/O.