� Suitability of Topological Data Structures for Data Parallel Operations in Computer Cartography

High performance computing is rapidly evolving towards parallel processing . Among various issues in applying parallel processing to computer cartography, a major concern is the compatibility between parallel computational models and existing cartographic data structures as well as algorithms . This paper assesses the suitability of topological data structures for Data Parallel Processing in computer cartography. By mapping selected cartographic operations to the data parallel model, we found that topological data structures, with minor extensions, can support data parallel operations . The topological information stored in the cartographic database can be used as index vectors for segmented-scan and permutation . To demonstrate this approach, the paper describes data parallel implementations of three cartographic operations are described, including line generalization, point-in­ polygon search, and connectivity matrix construction . INTRODUCTION The motivation to conduct this research is three-fold . First, recent studies have shown the great potential of data parallel processing in computer cartography and GIS (Mower, 1992, 1993; Mills et al ., 1992a, 1992b ; Li,1993) . While most studies have been conducted in the raster domain, data parallel processing in the vector space has received less attention. Due to their spatial irregularity, cartographic operations in the vector domain are more difficult to adopt to the data parallel computational model. To perform data parallel processing, the vector space must be transformed to a regular one through specific decompositions such as the uniform grid (Franklin et al ., 1989; Fang and Piegl, 1993) . Although the uniform grid is a very efficient data structure for many vector operations, one of its drawbacks is the overhead to convert the data from and to their original structures . The conversion is necessary in the real world situations because few software packages for GIS and computer cartography use the uniform grid . It is therefore worthwhile to study the suitability of existing vector data structures for data parallel processing . Topological data structures are selected in this project because they are the most commonly used data structures in GIS software packages . Second, large scale applications of the parallel computing technology require "enabling technologies" that are based on existing hardware and software environment and allow flexible evolutions . Set aside other technical and economic conditions for using parallel computing, software developers would not adopt strategies that require fundamental changes in existing data