三角晶格Ising模型Monte Carlo模拟的GPU加速算法

摘要

In the statistical model, the efficiency of most Monte Carlo algorithm reduces quickly near the critical point. In the analysis of traditional local algorithms, a GPU-based parallel simulation algorithm on the triangular lattice Ising model, which greatly improves the efficiency of the Monte Carlo simulation, is raised. For the model with the size of 1 024 X 1 024, a speedup of 69 is achieved. Besides, the critical behavior is analyzed, a high-precision critical point (/Jc = 0.274 66( 1) ) and critical exponents (y, = 1.01(2), yh= 1. 875 6(3) ) of triangular lattice Ising model are obtained, which implies the effectiveness of the GPU algorithm.%在分析传统Monte Carlo算法的基础上,针对三角晶格Ising模型提出了一种基于GPU的并行模拟方法,大大提高了算法的效率.对1 024×1 024的模型,实现了69倍的加速比.通过该算法所得数据分析模型的临界行为,获得了高精度的临界点βc=0.27466(1)和临界指数y1=1.01(2),yh=1.875 6(3).