Parallel Algorithms of Well-Balanced and Weighted Average Flux for Shallow Water Model Using CUDA

Table 4

Running time and speedup for Srinakarin Dam break simulation.

	Time (ms)	Speedup compared with previous program	Speedup compared with baseline
Serial: cell-based	29,156,017	Baseline	Baseline
Serial: edge-based	15,845,661	1.84	1.84
Parallel V1	1,689,703	9.38	17.26
Parallel occupancy	647,045	2.61	45.06
Parallel memory pattern	621,090	1.04	46.94
Parallel unroll	508,607	1.22	57.33