摘要
在粒子输运Monte Carlo程序中广泛采用线性同余法随机数发生器。为了改善该发生器产生的随机数序列的统计品质,该文基于洗牌法思想设计了一种新的随机数发生器。在新算法中,以原线性同余法随机数发生器的中间整型变量为基础,通过位操作来构造实现洗牌效果所需的随机整型变量。由于位操作的计算量较少,使得新算法在计算量上不会有明显的增加。而统计检验的结果表明,该方法对随机数序列统计品质的改善是显著的。同时,该文还阐述了新算法在并行环境中的应用。
Linear congruential random number generators (LCGs) are used in many Monte Carlo transport codes. An improved shuffle algorithm was developed to improve the statistical qualities of these random number sequence. In this algorithm, the random integral variable required for shuffling is obtained by bit manipulation of the integral variable generated in the LCGs. Thus, this algorithm is very fast and statistical tests indicate that the statistical quality is significantly improved. Application in a parallel environment is also discussed.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2009年第2期191-193,共3页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目(10775081)
国家"九七三"重点基础研究项目(2007CB209800)