Abstract
Numerical simulations of bed deformation as well as vegetation dynamics in river engineering generally necessitate high-load computations because lots of numerical meshes would be incorporated for high-accuracy calculation. In this study, we investigated performances of GPGPU (General-Purpose GPU) for numerical simulations with the shallow-water model used in river engineering. Especially, we compared the simulation results of GPGPU with those of CPU. Then, we examined the numerical accuracy and time efficiency of GPGPU simulations for the quasi-steady flows in a natural river with movement of the water's edge. Results showed that the numerical accuracy of GPU-accelerated numerical simulations was sufficient for river engineering tasks, although results of GPGPU simulations are not completely consistent with those of CPU simulations. Additionally, it was demonstrated that the numerical efficiency in GPGPU simulations is considerably improved using more numerical meshes within a range of the present numerical conditions.
