Embedded processor characteristics specification through multiobjective evolutionary algorithms

The design of a superscalar microprocessor for a given workload is a tremendous task by itself due to the numerous parameters involved and the ranges of their possible values. If power consumption and area are also to be considered then the problem is even more complicated and requires a suitable framework and methodology for exploring the vast multidimensional space for such a problem. In this paper we propose such a framework based on multi-objective evolutionary algorithms and demonstrate its use on a significant size example.