Multi-objective hardware–software partitioning of embedded systems: A case study of JPEG encoder

• HW–SW partitioning of embedded system is formulated as a multi-objective problem.
• Various cost terms of the problem are taken as objective functions.
• A binary GA is developed for solving the partitioning problem.
• Results are statistically compared for showing the performance of the GA.

Embedded systems have become integral parts of today's technology-based life, starting from various home appliances to satellites. Such a wide range of applications encourages for their economic design using optimization-based tools. The JPEG encoder is an embedded system, which is applied for obtaining high quality output from continuous-tone images. It has emerged in recent years as a problem of optimum partitioning of its various processes into hardware and software components. Realizing pairing and conflicting nature among its various cost terms, for the first time the JPEG encoder is formulated and partitioned here as a multi-objective optimization problem. A multi-objective binary-coded genetic algorithm is proposed for this purpose, whose effectiveness is demonstrated through the application to a real case study and a number of large-size hypothetical instances.

Figure optionsDownload as PowerPoint slide