Comparison of Binary Rain Prediction on HIMAWARI using MPI and CUDA

Abstract

Predictions of rainfall events are very important to determine the preventive steps needed to avoid adverse effects with climate change that slowly changing the impact of rain events in Indonesia's area. One way to calculate the prediction of rain events is to use the Lookup Table and Probability of Rain method. To perform that, parallel programming is conducted to help solve large and complex computational problems. In this study, CPU (Central Processing Unit) programming and GPU (Graphical Processing Unit) programming are done to see each performance where computation design and Lookup Table design are implemented on both programming. The results show that CPU programming is more effective for the given case and data rather than GPU programming. Meanwhile, non-ranged Lookup Table design in CPU programming gives more accurate predictions even though it takes more time in computation.