Polygon-based hidden surface elimination algorithms: serial and parallel
thesisposted on 18.05.2018 by Julian C. Highfield
In order to distinguish essays and pre-prints from academic theses, we have a separate category. These are often much longer text based documents than a paper.
Chapter 1 introduces the need for rapid solutions of hidden surface elimination (HSE) problems in the interactive display of objects and scenes, as used in many application areas such as flight and driving simulators and CAD systems. It reviews the existing approaches to high-performance computer graphics and to parallel computing. It then introduces the central tenet of this thesis: that general purpose parallel computers may be usefully applied to the solution of HSE problems. Finally it introduces a set of metrics for describing sets of scene data, and applies them to the test scenes used in this thesis. Chapter 2 describes variants of several common image space hidden surface elimination algorithms, which solve the HSE problem for scenes described as collections of polygons. Implementations of these HSE algorithms on a traditional, serial, single microprocessor computer are introduced and theoretical estimates of their performance are derived. The algorithms are compared under identical conditions for various sets of test data. The results of this comparison are then placed in context with existing historical results. Chapter 3 examines the application of MIMD style parallelism to accelerate the solution of HSE problems. MIMD parallel implementations of the previously considered HSE algorithms are introduced. Their behaviour under various system configurations and for various data sets is investigated and compared with theoretical estimates. The theoretical estimates are found to match closely the experimental findings. Chapter 4 summarises the conclusions of this thesis, finding that HSE algorithms can be implemented to use an MIMD parallel computer effectively, and that of the HSE algorithms examined the z-buffer algorithm generally proves to be a good compromise solution.
- Computer Science