This thesis outlines a method of scheduling for signal processing applications onto heterogeneous multiprocessor systems using a combination of techniques. The problem is to efficiently schedule an application in the form of direct acyclic graphs (DAG), onto a heterogeneous processor environment, defined as a NP-complete problem. The solution is to use a variety of techniques based on greedy methods and meta-heuristic methods to solve the problem. Using aspects from DAG scheduling the application can be analyzed, using deadline like scheduling some tasks of different periods can be properly placed, and using meta-heuristic methods near optimal placements can be arranged for the tasks in the DAG. Results obtained indicate that the system is efficient in placing tasks to create a schedule. Results also show that the optimization engine consisting of a variety of meta-heuristic methods can generate proper processor assignments depending on how they are applied.
展开▼
