Efficacy of a GPGPU-Acceleration to Inundation Flow Simulation in Tonle Sap Lake in Cambodia
A new 2D numerical model is developed for a rapid computation of the seasonal inundation phenomena in Tonle Sap Lake in Cambodia. In order to overcome a huge computational cost for a prolonged analysis over an extensive area, the General-Purpose computing on Graphics Processing Units (GPGPU) technology is applied to the model. The developed model is applied to a solution of seasonal inundation process for the 154 days in 2002. Calculated result is compared with observational data and satellite remote sensing. It is found that the developed model seems to successfully reproduce reasonable progress/regress of inundation. A breakdown of the total elapsed time for the numerical analysis is considered in a detail. It is found that the GPGPU technology can accelerate the solution more than one hundred times faster by employing a simple rectangular mesh and coding to reduce a memory access overhead.