期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2015
卷号：3
期号：6
DOI：10.15680/ijircce.2015.0306136
出版社：S&S Publications
摘要：Game tree Search algorithm is used to search the best move in game tree. GTS is a combinatorialproblem in which it is hard to find optimal solution from huge possible solutions. Focus of the system is to takeadvantage of GPU’s massive parallelism capability to accelerate the speed of game tree algorithm and propose aconcise and general parallel game tree algorithm on GPUs. GPU computing is getting popular among scientificcommunity because of cheap and high performance computational power. Proposed system is implemented forConnect4 and Connect6 game using CUDA and MPI programming environment. It is found that parallelization taskson SIMD processors of graphics cards perform better during searching and evaluating a GTS. In this CPU isresponsible for maintain tree structure of game tree and GPU is responsible for evaluating node simultaneously. Thuschoice is to use combination of CPU-GPU solution with DFS-BFS search respectively. Comparison is done with serialimplementation for Connect4 and Connect 6 games
关键词：Connect4/Connect6; CUDA; Game Tree Search; GPU; Parallel Computing
