SearchPublisher and Institutes
You are here: Home :: Area NEM :: Computer science
Ursula Fissgus
Scheduling and Data Distribution in Parallel Machines – a Combined Decision Based on Genetic Algorithms
Distributed memory machines provide a large computing power, but the development process for a specific parallel algorithm on a specific machine is complex due to the complicated runtime behaviour. We consider a powerful multi-dimensional scheduling embedded into a tool for generating parallel programs with mixed task and data parallelism.
Our scheduling is based on the genetic algorithm paradigm and it takes not only decisions on the execution order (independent tasks can be executed consecutively by all processors available or concurrently by independent groups of processors) and on the mapping of processors to tasks, but also on appropriate data distributions and task implementation versions (for each task there are several implementation version available, e.g., taken from a predefined set of library functions). Data redistribution operations and communication domain management operations are added, if necessary.
it – Information Technology (vormals it+ti), Oldenbourg Wissenschaftsverlag
Print ISSN: 1611-2776
Volume: 44, 03/2002
Pages: 160