Date of Award

5-2021

Document Type

Thesis open access

Department

Computer Science

First Advisor

Mark C. Lewis

Second Advisor

Britton Horn

Third Advisor

Berna L. Massingill

Abstract

This project creates a ray-tracing and geometry distribution framework through an actor model of parallelism, which is then expanded onto a cluster of machines to show effective data distribution across a network. This is shown to be feasible, but due to problems internal to the actor framework, as well as design failures, fails to effectively and consistently increase usable memory and generate larger ray-traces, though generally scaled well. Despite this, it compares several methods of ray organization across the geometry and shows that more complex methods generally scale better with the amount of geometry. A photometric renderer was added with very little modification, showing the generality of the geometry distribution framework, and the performance benefits of alternative serialization methods are shown to outweigh the drawbacks of more difficult implementation.

Download

Share

COinS