A scheduling scheme for communicating tasks

Heuristic methods are proposed for assigning a set of soft tasks to a set of PEs (processing elements) in such a way that the set of tasks has a minimum execution time. The objective is to try to meet the performance requirements of applications, especially of real-time applications. Emphasis is placed on specific applications where there are more PEs than active tasks. Two criteria are proposed to measure the quality of an interconnection for these applications. A simple grain-size decision algorithm that groups tasks that communicate with each other into one node is presented.<
>