Automated specification and task allocation methods for single and multimode embedded applications

This thesis considers the design problem of specification and task allocation for single and multiple mode embedded applications. The thesis presents research into two different aspects of this problem.; The first aspect is the substitution of path ordering for the traditional task size ordering in the existing bin packing solution for single mode applications. Path ordering presents the bin packing algorithm with sequences of tasks that pass data to one another. This seeks to reduce the number of communicating tasks assigned to different processors and thus reduce bus traffic. Different heuristic algorithms are developed and explored to perform the path discovery process. The use of path ordering was found to be an effective way to reduce bus utilization.; The second aspect researched was solving the multiple mode application problem. The goal is to first specify an architecture and then assign the tasks of each distinct operating mode to a single specified architecture. Four types of algorithms with a total of 34 variations were developed and evaluated. The best variation resulted in solutions with average costs within 7% of lowest cost found for each example problem. A combination of two of the variations brought the average cost to within 3% of the lowest cost found.; Finally, the combined use of path ordering and multiple mode methods was explored. Path ordering reduced bus utilization by 3% to 25% depending on the multiple mode algorithm in which it is applied.