Use this URL to cite or link to this record in EThOS:
Title: Interactive global illumination on the CPU
Author: Dubla, Piotr
ISNI:       0000 0004 2748 4703
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Access from Institution:
Computing realistic physically-based global illumination in real-time remains one of the major goals in the fields of rendering and visualisation; one that has not yet been achieved due to its inherent computational complexity. This thesis focuses on CPU-based interactive global illumination approaches with an aim to develop generalisable hardware-agnostic algorithms. Interactive ray tracing is reliant on spatial and cache coherency to achieve interactive rates which conflicts with needs of global illumination solutions which require a large number of incoherent secondary rays to be computed. Methods that reduce the total number of rays that need to be processed, such as Selective rendering, were investigated to determine how best they can be utilised. The impact that selective rendering has on interactive ray tracing was analysed and quantified and two novel global illumination algorithms were developed, with the structured methodology used presented as a framework. Adaptive Inter- leaved Sampling, is a generalisable approach that combines interleaved sampling with an adaptive approach, which uses efficient component-specific adaptive guidance methods to drive the computation. Results of up to 11 frames per second were demonstrated for multiple components including participating media. Temporal Instant Caching, is a caching scheme for accelerating the computation of diffuse interreflections to interactive rates. This approach achieved frame rates exceeding 9 frames per second for the majority of scenes. Validation of the results for both approaches showed little perceptual difference when comparing against a gold-standard path-traced image. Further research into caching led to the development of a new wait-free data access control mechanism for sharing the irradiance cache among multiple rendering threads on a shared memory parallel system. By not serialising accesses to the shared data structure the irradiance values were shared among all the threads without any overhead or contention, when reading and writing simultaneously. This new approach achieved efficiencies between 77% and 92% for 8 threads when calculating static images and animations. This work demonstrates that, due to the flexibility of the CPU, CPU-based algorithms remain a valid and competitive choice for achieving global illumination interactively, and an alternative to the generally brute-force GPU-centric algorithms.
Supervisor: Not available Sponsor: Engineering and Physical Sciences Research Council (EPSRC) (EP/D069874/2)
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: QA76 Electronic computers. Computer science. Computer software ; T Technology (General) ; TK Electrical engineering. Electronics Nuclear engineering