In Partial Fulfillment of the Requirements for the Degree of
Master of Science
Will defend his thesis
Heterogeneous computational platforms consisting of CPUs and one or more GPUs are becoming common and widely available these days. Such availability has brought up the need to develop techniques that allow the effective use of the computational resources offered by heterogeneous platforms. In this work we investigate efficient load balancing strategies for ray tracing on heterogeneous (CPU + Multi-GPUs) platforms. The main difficulty in achieving a high efficiency for ray tracing in this context has to do with the fact that although an embarrassingly parallel problem, ray tracing also exhibits a highly irregular workload. We propose ray traversal cost estimation to improve the balancing efficiency of commonly used balancing strategies by accurate initial distribution of tasks among processors. We compare regular and cost-based versions of popular task queuing systems for workload balance on a heterogeneous platform, and show that using the traversal cost-based initialization measurably improves the efficiency of workload balancing of ray tracing on the heterogeneous platform.
Date: Tuesday, November 29, 2011
Time: 2:30 PM
Faculty, students, and the general public are invited.
Advisor: Dr. Zhigang Deng