Abstract: In the field of game theories and artificial intelligence Game tree search is a classical problem. The general use of GTS algorithm is in real time applications having much higher complexity like video games, chess, connect6, Go etc. Different algorithms for game tree are used to find the player's next best move on the game tree in minimum time. Main focus of the system is on increasing massive parallelism capabilities of GPUs to accelerate the speed of game tree algorithms and propose a general parallel game tree algorithm on GPUs. In game tree search, GPU surpasses a single CPU if high level of parallelism is achieved because of its searching is in BFS manner and CPU is in DFS manner so that CPU did not produce improvement. Here combination of DFS and BFS is main focus and selection will be the depth-first search on CPU and use breadth-first search on GPU. CPU can be responsible for generating number of choices of players' moves as a tree structure and parallel evaluation of these moves can be perform using GPU. It intends to look into a hybrid CPU-GPU solutions.

Keywords: SIMD, GPU, GTS, SUDOKU, Parallel Computing.