The parallel generation of 2-D Hilbert Space-filling Curve on GPU

摘要

In this paper, we propose two parallel Hilbert Space-filling Curve(HSFC) generation algorithms BMIMp and SDDMp based on block matrix iteration method(BMIM) and state diagrams driver method(SDDM) in the CUDA parallel programming mode. Numerical results show that both of them obtain high parallel speedup. Especially, the speedup of BMIMp and SDDMp can reach 207 and 290 respectively for the 14-order HSFC. Furthermore, BMIMp outperforms SDDMp when considering the total computation time.