论文部分内容阅读
传统的蒙特卡罗粒子输运计算程序在粒子每一步模拟结束后,通过遍历所有计数器来判断当前粒子所在栅元是否需要进行计数,该过程耗时随计数器数量的增加近似线性增长,当计数器数量较大时,计数耗时远高于输运耗时。本文发展了一种基于计数辅助树的大规模计数加速方法,建立了与几何栅元一一对应的树形结构,并在节点中存储了相应栅元的计数信息,通过当前粒子所在栅元的几何信息从树中快速读出对应的计数器。为了验证该方法的有效性,基于Hoogenboom全堆基准例题测量了不同计数器数量下的计算耗时。测试结果显示本文方法能有效地提高大规模计数问题的计算效率。
The traditional Monte Carlo particle transport calculation program traverses all the counters to judge whether or not the cell of the current particle needs to be counted after the simulation of each step of the particle is finished. The process takes approximately linear increase with the increase of the number of the counters. When the number of the counters Larger, the count takes much longer than the transport time-consuming. In this paper, we develop a large-scale counting acceleration method based on counting auxiliary tree, establish a tree structure corresponding to the geometric cells one by one, store the counting information of the corresponding cells in the nodes, Geometry Information Reads the corresponding counter quickly from the tree. In order to verify the effectiveness of this method, the computational time-consuming under different numbers of counters was measured based on the Hoogenboom full stack benchmark example. The test results show that the proposed method can effectively improve the computational efficiency of large-scale counting problems.