0 Project Outline

Motivation

Adapting a GPU-accelerated weather simulation will allow for the opportunity to commercialise space complex CPU intensive work that would otherwise be extremely costly through the means of a super computer. However, this comes with a myriad of challenges. Chunking the computation for the GPU such that it may be able to process, and to parallelise its functions to greatly speed up computation. The success of this project could improve current weather simulation technologies, and allow for greater and more accurate weather simulations at a finer cost.

Aims

This project will be to amend current implementation of the existing CPU weather simulation, and incorporate GPU-accelerated weather simulation through OpenCL. To be able to chunk arrays in such a way that it is efficiently fed to the GPU, and be able to perform calculations on each of the chunks where a calculation may be different depending from where the chunk came from. Implement the C/C++ implementation onto the FORTRAN template, and test if there were improvements in new approach