Date of Award

Spring 2022

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Puri, Satish

Second Advisor

Madiraju, Praveen

Third Advisor

Ahamed, Sheikh Iqbal

Abstract

Geo-Spatial computing and data analysis is the branch of computer science that deals with real world location-based data. Computational geometry algorithms are algorithms that process geometry/shapes and is one of the pillars of geo-spatial computing. Real world map and location-based data can be huge in size and the data structures used to process them extremely big leading to huge computational costs. Furthermore, Geo-Spatial datasets are growing on all V’s (Volume, Variety, Value, etc.) and are becoming larger and more complex to process in-turn demanding more computational resources. High Performance Computing is a way to breakdown the problem in ways that it can run in parallel on big computers with massive processing power and hence reduce the computing time delivering the same results but much faster.This dissertation explores different techniques to accelerate the processing of computational geometry algorithms and geo-spatial computing like using Many-core Graphics Processing Units (GPU), Multi-core Central Processing Units (CPU), Multi-node setup with Message Passing Interface (MPI), Cache optimizations, Memory and Communication optimizations, load balancing, Algorithmic Modifications, Directive based parallelization with OpenMP or OpenACC and Vectorization with compiler intrinsic (AVX). This dissertation has applied at least one of the mentioned techniques to the following problems. Novel method to parallelize plane sweep based geometric intersection for GPU with directives is presented. Parallelization of plane sweep based Voronoi construction, parallelization of Segment tree construction, Segment tree queries and Segment tree-based operations has been presented. Spatial autocorrelation, computation of getis-ord hotspots are also presented. Acceleration performance and speedup results are presented in each corresponding chapter.

Share

COinS