ソフトコンピューティングを応用したVLSI自動フロアプランの研究

Abstract

This paper proposes to apply fuzzy inference to automate the floorplanning design which decides a macroscopic placement of the top layer of LSI physical implementation. (1) Positioning of major blocks to be inferred by fuzzy inference based on the knowledge and experience of experts. (2) Genetic algorithms are employed to optimize the placement of such blocks that are not suitable for fuzzy inference because the fitness value is lower than certain level, to avoid the mess of fuzzy rules. (3) The result of fuzzy inference will be embedded into the initial population of genetic algorithms so as to realize simpler cost function and faster computation time than the conventional genetic algorithms which employ the initial population generated by using random numbers.