A graph-theory-based method for parallelizing the multiple-flow-direction algorithm on CUDA compatible graphics processing units | IEEE Conference Publication | IEEE Xplore