SC²S Colloquium - January 28, 2015
|Date:||January 28, 2014|
|Time:||3:00 pm, s.t.|
|Invited by:||Univ.-Prof. Dr. Michael Bader|
Nicolai Schaffroth: A Vectorizable Riemann Solver for the Shallow Water Equations
The goal of this projecet was to implement a vectorizable Riemann solver for the shallow water equations for use in tsunami modelling. The solver is based on the HLLE approach and can handle variable bathymetry and inundation. The simulation quality and performance is compared to other solver implementations in different artificial scenarios as well as simulations of the 2011 Tohoku Earthquake Tsunami. The quality of the results competes well with slower, more complex and not auto-vectorizable solvers for most practical applications. Automatic vectorization on Intel compilers significantly speeds up the execution. Additionally, a CUDA version for parallelized execution on NVIDIA GPUs is implemented and tested, leading to further gains in performance on suitable hardware.