November 30, 2011
Title: Ray Tracing Large Distributed Datasets Using Ray Caches
Abstract: Ray tracing is extremely useful for visualization due to its high quality and high accuracy. Most large scale simulations now produce datasets that can be significantly larger than can typically be stored in memory on a visualization system. Visualizing datasets of this size with a typical ray tracing algorithm becomes ineffective and stalls since the data must be paged to disk. Recently, in-situ visualization has received renewed attention for visualizing large datasets that are distributed among many processors during a simulation. In-situ visualization takes advantage of the fact that each processor taking part in the simulation will already have a portion of the dataset loaded into memory. To render this distributed dataset, however, then requires communication between the processors which can be just as expensive as disk access if it is not managed efficiently. The goal of this thesis was to develop an in-situ visualization technique using distributed out-of-core ray tracing. This technique assumes that each processor in a cluster contains a subset of the simulation dataset. To ray trace the dataset, rays traverse the processors which perform a local ray tracing algorithm with the data subset that it stores. The number of rays that pass between processors is often quite large, which causes significant communication overhead. To alleviate this, ray caches are placed at the boundaries between processors to capture and reuse rays, thereby replacing communication with a significantly less expensive cache search operation. Through testing of a simple implementation based on ray casting it was found that ray caching can significantly improve overall performance at a small cost to image quality.
Biography: Christopher Little is currently pursuing his MSc in Computer Science at UOIT under the supervision of Dr. Mark Green and Dr. Faisal Qureshi. He is expected to graduate in the next few months. Christopher received his BSc (Hons.) in Computing Science from UOIT as part of the first ever Computing Science graduating class in 2009.