REU: Parallelization of Monte Carlo Algorithms in Semi- conductor Device Physics

用蒙特卡罗粒子法模拟二维半导体器件非平衡输运数值模型的并行算法。这些模型目前使用矢量串行计算机进行计算。然而，通过使蒙特卡罗适应并行系统，例如NCUBE或Sequent，可以大大提高速度，并大大提高此类模型的规模和复杂性。这里的方法将采用基于“谷物包装”的通用算法分析，这是俄勒冈州立大学开发的一种技术，以确定多处理器系统中问题的最佳分配。研究重点是通过并行化现有的GaAs体输运顺序算法和蒙特卡罗方法模拟MESFET获得的并行与顺序加速。

Parallel algorithms for numerical models used in semiconductor device simulation of nonequilibrium transport in two dimensions using Monte Carlo particle methods. Such models are currently used and computed using vector serial computers. However, by adapting Monte Carlo to parallel systems, e.g. NCUBE or Sequent could result in substantial speedup and greatly enhance the scale and sophistication of such models. The methodology here will be to employ a general algorithmic analysis based on 'grain packing', a technique developed at the Oregon State University, in order to determine the optimal parti- tioning of the problem in a multi-processor systems. The research focuses on the parallel versus sequential speedup obtained through parallelization of existing sequential algo- rithms for bulk transport in GaAs and MESFET simulation using Monte Carlo methods.