Cohoon et al.in [19]offered the classical work on using genetic algorithms (GAs)for floorplanning.The arrangement of rooms on the layout surface was represented in the genotype by a postfix expression that was not normalized of the corresponding slicing tree. Schnecke et al.in [20]used a GA to manipulate the binary slicing tree directly for floorplanning.However,the crossover involved complex repair mechanisms simply to ensure that the offspring represented a legal slicing floorplan.Valenzuela et al.in [21]presented a GA that used a slicing tree construction process.They used an order-based representation that encoded rectangles and binary operations into a simple permutation of structures and a decoder that converted the permutation of structures into a normalized postfix expression. Generally speaking,applications of EAs to floorplanning problems are much fewer than those of SA.Although some researchers [19]-[21]used EAs,their works were all based on the slicing structure instead of the nonslicing structure.This restricted the popularization of these methods to some extent.In our opinion,the reason for fewer applications of EAs is that the nonslicing floorplan representations are generally very complex and solution spaces are constrained.In this case,EAs with traditional evolutionary operators (e.g.,crossover)tend to create infeasible individuals during the search.To boost the development of EAs in the field of floorplanning,it is important to design new nonslicing floorplan representations that do not exert extra constraints on solution spaces and facilitate effective crossover operators. B.Floorplanning Problems with Arbitrarily Shaped Rectilinear Blocks In the simplest situation of floorplanning,all blocks are rectangular.However,in real design,as some of the circuit blocks come from design re-use,they are not necessarily rectangular.To fully optimize some predefined cost metric,for example,area,wirelength,or7 Cohoon et al. in [19] offered the classical work on using genetic algorithms (GAs) for floorplanning. The arrangement of rooms on the layout surface was represented in the genotype by a postfix expression that was not normalized of the corresponding slicing tree. Schnecke et al. in [20] used a GA to manipulate the binary slicing tree directly for floorplanning. However, the crossover involved complex repair mechanisms simply to ensure that the offspring represented a legal slicing floorplan. Valenzuela et al. in [21] presented a GA that used a slicing tree construction process. They used an order-based representation that encoded rectangles and binary operations into a simple permutation of structures and a decoder that converted the permutation of structures into a normalized postfix expression. Generally speaking, applications of EAs to floorplanning problems are much fewer than those of SA. Although some researchers [19]-[21] used EAs, their works were all based on the slicing structure instead of the nonslicing structure. This restricted the popularization of these methods to some extent. In our opinion, the reason for fewer applications of EAs is that the nonslicing floorplan representations are generally very complex and solution spaces are constrained. In this case, EAs with traditional evolutionary operators (e.g., crossover) tend to create infeasible individuals during the search. To boost the development of EAs in the field of floorplanning, it is important to design new nonslicing floorplan representations that do not exert extra constraints on solution spaces and facilitate effective crossover operators. B. Floorplanning Problems with Arbitrarily Shaped Rectilinear Blocks In the simplest situation of floorplanning, all blocks are rectangular. However, in real design, as some of the circuit blocks come from design re-use, they are not necessarily rectangular. To fully optimize some predefined cost metric, for example, area, wirelength, or