GPU加速模拟混合自旋的Ising模型  

GPU Accelerate Simulation of Mixed Ising Model

作　　者：崔沛东[1] 蔡静[1] 魏亮[1] 张伟[1] 

机构地区：[1]暨南大学理工学院物理学系,广州510632

出　　处：《科学技术与工程》2013年第26期7720-7724,7735,共6页Science Technology and Engineering

摘　　要：基于传统的Monte Carlo方法,利用CUDA(compute unified device architecture)技术,在GPU上分别模拟了正方晶格,三角晶格上较为复杂的不同浓度比的混合(自旋S=1和S=1/2)Ising模型,获得了不同浓度比的临界温度K c以及临界指数y t、y h。与同时期的CPU算法相比,基于GPU的并行模拟算法大大提高了算法效率,对于尺寸为1 024×1 024的混合自旋的Ising模型,在正方晶格上实现了40倍的加速比,在三角晶格上实现了41倍的加速比。Compute unified device architecture （CUDA） is used as the programming approach on GPU to Monte Carlo simulations of the mixed spin Ising model（ S = 1/2 , S = 1 ）. The critical temperature, critical exponent of different concentration ratio for the square and the triangular lattice are obtained. Comparing with the traditional algorithm on CPU, the GPU-based parallel simulation algorithm greatly improves the efficiency. As to the mixed spin Ising model sized at 1 024 × 1 024, it has achieved 40 times speedup on square lattice and 41 times on triangular one.

关 键 词：GPU 加速 混合自旋Ising模型 

分 类 号：O414.2[理学—理论物理]