Improving Divide-and-Conquer Ray-Tracing Using a Parallel Approach

Abstract

This paper presents a new Divide-and-Conquer Ray-Tracing (DACRT) algorithm that is designed to perform on multi-core processors. This new algorithm proposes a parallel and generic scheme that, without the use of any data structure for spatial subdivision, maintains memory management minimal and deterministic. Initially, the scene is divided into sub-scenes and those uniformly distributed across available hardware resources, processing each sub-scene individually. After, an iterative step to ensure the correct results is performed until the final frame is obtained. Results show that our algorithm is up to 2.4x times faster than the original DACRT in a common quad-core processor setup, allowing very high interactive frame rates in well-known benchmark scenes.