A Dynamic and Adaptive Quiescence Detection Algorithm
Authors:
A. Sinha and L.V. Kale, and B. Ramkumar
Parallel Programming Laboratory, Department of Computer Science, University
of Illinois at Urbana-Champaign
September 1993. [Internal Report #93-11]
A large number of quiescence detection algorithms with good theoretical upper-bounds have been proposed before. However the metric used to measure the performance of these algorithms is not suitable. We propose a new metric for measuring the performance of a quiescence detection algorithm. We also present an algorithm to detect quiescence in an asynchronous and dynamic model of parallel computation. The algorithm has been implemented for a machine independent parallel programming system, Charm. Quiescence detection is provided as a feature in Charm to conduct a variety of operations like collecting statistics about user computation, initiating new phases of computation, or just terminating the user computation.