Date of Award
Master of Science
Dr. Ajay Gupta
Dr. Elise de Doncker
Dr. Don Nelson
Masters Thesis-Open Access
The adaptive integration algorithm is· effective in numerically solving integration problems. It is able to focus the application of integration rules on the portion of the integration region where the integrand is the least well-behaved. Parallel implementations must use dynamic load balancing or performance suffers.
Dynamic local load-balancing techniques allow each processor to maintain its own pool of work in a local priority queue and balance the workload based on local criteria. However, the use of locally controlled priority queues is known to be inefficient (in terms of the number of integration rule applications needed to reach an answer) as the number of processors increases.
This thesis implements and analyzes the use of a distributed priority queue for managing the current pool of work and performing global load balancing in the parallel adaptive integration algorithm. Initial experimental results show that the use of a distributed priority queue does not provide any clear benefits over the use of multiple local priority queues. However , analyzing the circumstances where the results are better has led to greater understanding of the basic algorithm.
Zanny, "Efficiency of Distributed Priority Queues in Parallel Adaptive Integration" (1999). Master's Theses. 4237.