Analog integrated circuit (IC) floorplan
automation due to its importance has been intensively studied in
the last decades. The problem is complex as multiple
requirements which appear in the form of device symmetry,
matching and proximity constraints must be dealt simultaneously
for a robust floorplan. Absolute coordinates is the most intuitive
manner of implementing these layout constraints, as it allows for
the optimization kernel to move the cells explicitly and represent
any possible floorplan. A complete study of previous approaches
shows that illegal overlaps have been improperly modeled in a
single-objective (SO) cost function for optimization along with
other objectives. In this paper, the problem of analog floorplan
automation in absolute coordinates is re-formulated, and, since it
is impracticable to determine a single best floorplan for all of the
design objectives, a multi-objective optimization (MOO)
algorithm is applied to solve it. In order to reduce the problem
complexity, Pareto optimal fronts (POFs) of placements
representing the tradeoffs between the design objectives are
combined bottom-up through the partitions of the circuit’s
hierarchy. The approach is demonstrated for the UMC 0.13μm
design process and compared with most recent representations.