Hybrid evolutionary algorithm of planning VLSI

Abstract EN

Hybrid evolutionary algorithm and technique of planning VLSI are described.

Process of slicing floorplan synthesis is broken into two connected stages- synthesis of the slicing tree structure and floorplan formation by aggregation method.

During genetic search the decisions specifying structure of slicing tree are exposed to evolutionary changes.

On each step of genetic search the plan is under construction by application to this decision of aggregation procedure constructed on the basis of mechanisms of collective alternative adaptation.

Two approaches to reduction of total area of plan are used in work.

At the first approach the sizes of modules are fixed.

It is possible to change modules orientation.

At the second approach the sizes of modules can change according to restrictions.

The linear estimation of time complexity of decoding procedure, and also structure, at which chromosomes and the genes located in the same loci, are homologous, simplifies genetic operators performance procedures and the genetic search organization.Experimental research has shown that algorithms at teamwork within the framework of the common approach give high quality.