基于CUDA的泥石流模拟计算研究  

CUDA-based simulation of debris flow

作　　者：杨升[1] 管群[1] 

机构地区：[1]四川大学计算机学院,四川成都610064

出　　处：《计算机工程与设计》2011年第12期4231-4236,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(40771026);国家自然科学基金与俄罗斯基础研究基金合作基金项目(40911120089)

摘　　要：针对泥石流仿真过程中海量数据计算问题,采用CUDA技术即结合CPU与GPU的优点研究了一种协同计算方法以提高数据计算效率和仿真性能。同时,搭建了基于GPU的泥石流仿真计算平台,对优化的CUDA并行计算方法进行验证。实验结果表明,该方法对海量数据的计算具有快速准确、低成本、低功耗的特点,能为灾害预测提供及时准确的决策支持,满足了高密集型计算的需求。In order to solve the mass data computing problems in the simulation process of debris flow,a collaborative computing method,employing the CUDA technology which combined the advantages of CPU and GPU is proposed to enhance the computing efficiency and the simulation performance.Besides,the debris flow simulation computing platform is built based on GPU to valid the optimized CUDA parallel method.And the simulation results show that the optimized CUDA parallel method has the characteristics of fast,accurate,low cost and low power consumption and it provide timely and accurate decision support for disaster prediction and meet the requirements of high intensive computing.

关 键 词：泥石流 灾害数据仿真 统一计算设备架构 数据运算 并行计算 

分 类 号：TP338[自动化与计算机技术—计算机系统结构]