异构环境下的空间分析并行映射策略  

Parallel Processing Mapping Strategy of Spatial Analysis under the Heterogeneous Environment

作　　者：张剑波[1] 周斯波[1] 袁国斌[1] 石阳[1] 朱建勃[1] 

机构地区：[1]中国地质大学(武汉)信息工程学院,武汉430074

出　　处：《上海交通大学学报》2013年第1期70-75,共6页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金项目资助(41001225)

摘　　要：针对传统的地图代数局部算子实现方法用于海量栅格数据计算时效率低下的问题,从串行算法的并行化映射、计算机图形处理器资源的自适应参数调整等角度分析地图代数空间并行算法的实现机制,采用数据分割策略对空间分析算子的计算速度进行分析,将空间分析算子分割成若干子任务,并映射到图形处理器(GPU)中进行运算,通过运算与数据传输的重叠隐藏数据传输时间,借助异构环境计算能力的强大而加速算子的运算.理论分析与实验结果表明,该策略能够明显提高空间分析算子的运算速度.Aiming at the low efficiency when traditional realization methods of local arithmetic operators of map algebra apply to calculations for gigantic raster data, the realization mechanism from surrounding par- allel mapping of serial algorithms and adaptive parameter adjustments on computer graphic processor re- sources was discussed. The data partition strategy was used to the acceleration of spatial analysis opera- tors. The strategy divides operators into subtasks and maps these subtasks to the graphic processing unit （GPU） for calculation. It overlaps the calculation with data transmission to hide the data transmission time and accelerate operators by using the powerful computation ability of heterogeneous environment. The the- oretical analysis and experimental results show that this strategy can significantly improve the processing speed of space analysis operators.

关 键 词：异构环境 空间分析 并行映射 地图代数 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] P208[自动化与计算机技术—计算机科学与技术]