SOLUTION SPACE REDUCTION IN TASK SCHEDULING FOR HETEROGENEOUS COMPUTING SYSTEMS

摘要

A new approach to address the mapping and scheduling problem in hardware/software codesign is presented. This approach is based on heuristic search and two techniques are proposed to enhance the quality of mapping/scheduling solutions: (1) a task clustering technique to reduce the solution space and (2) a multiple neighbourhood function strategy to better explore the solution space. It is evaluated by case studies involving 2880 randomly generated task graphs, as well as four real applications. Experimental results show that the proposed approach outperforms previous approaches in terms of speedup by up to 36.2% for a system with a microprocessor, a digital signal processor, and an FPGA.