Scheduling Trees in Parallel/Pipelined Processing Environments

Publisher Website
Google Scholar
Abstract
Scheduling task trees to be executed in parallel and/or pipelined processing systems are examined under individual situations. Processor structural requirements at task nodes are also included in the model of consideration. While simple techniques can serve as heuristics, counterexamples are constructed in some crucial cases. Simple optimal algorithms are presented in two important cases: 1) unistructure, multipipe, uniform latency, and flush time; and 2) vector loops. Finally, the complexity of the remaining cases is scrutinized with different structural parameter combinations.