EXPERIMENTS IN THE IDENTIFICATION OF TERRAIN FEATURES USING A PC-BASED PARALLEL COMPUTER

This paper describes the process through which an existing serial algo rithm that classifies terrain features in a digital elevation model (D EM) is recast into a version that executes in a parallel computer envi ronment. The transformation process is guided by a formal analysis of dependency relationships that exists among the different steps in the algorithm. The resulting program is tested using a small parallel comp uter system. The results demonstrate that run times are reduced, and t hat the processors are used efficiently. The general approach to the d evelopment and implementation of parallel algorithms that is presented in this paper is extensible to a wide range of geographical computing problems.