Parallel and GPU Based Strategies for Selected CFD and Climate Modeling Models

In recent years we have observed a huge increase in natural disasters, such as earthquakes, tornadoes and floods. Even Poland as a relatively stable region has experienced many big natural disasters, in particular three floods early this year with the estimated cost around 3 billion Euro. To address the dramatic changes in our climate, hydro-meteo scientists have to work together to share important data, tweak existing meteorological models, even couple them, ultimately achieving the goal of preventing such disasters in the future. The computing power and data capacity grow every day, and scientists are willing to run the hydro-meteo simulations at greater model size as well as use new archiving and back up services for historical analysis. However, to benefit from modern architectures, applications and data structures have to be adapted properly. In this chapter we discuss various ideas to improve the performance of Eulag - a numerical solver for all-scale geophysical flows using innovative computing technologies. We present preliminary results of applying the GPGPU and OpenMP shared memory model to Eulag, so that the application can be well scaled on cluster of GPUs or cluster of SMPs nodes respectively. Moreover, we present additional useful solutions to visualize the hydro-meteorological data at speed, as well as to share data in a secure way among end-users