The Source-Sink Model

Weather forecasting is based on the use of numerical weather prediction (NWP) models that are able to perform necessary calculations that describe/predict major atmospheric processes. Numerical weather prediction advances have been always very closely related with advances in computing infrastructure. Consequently, new computational technologies and facilities are exploited to boost the development of numerical weather prediction applications on different computing platforms. This paper describes the Source-sink model implementation that is significantly accelerated using accelerator based on DataFlow architecture. In our implementation, the algorithm execution time exploiting Maxeler MaxCard is 50 times shorter than the execution time on the Intel i3 based PC.