Data Structures and Basic Algorithms

Abstract

VLSI chip design process can be viewed as transformation of data from HDL code in logic design, to schematics in circuit design, to layout data in physi-cal design. In fact, VLSI design is a significant database management prob-lem. The layout information is captured in a symbolic database or a polygon database. In order to fabricate a VLSI chip, it needs to be represented as a collection of several layers of planar geometric elements or polygons. These ele-ments are usually limited to Manhattan features (vertical and horizontal edges) and are not allowed to overlap within the same layer. Each element must be specified with great precision. This precision is necessary since this information has to be communicated to output devices such as plotters, video displays, and pattern-generating machines. Most importantly, the layout information must be specific enough so that it can be sent to the fab for fabrication. Symbolic database captures net and transistor attributes. It allows a designer to rapidly navigate throughout the database and make quick edits while working at a higher level. The symbolic database is converted into a polygon database prior to tapeout. In the polygon database, the higher level relationship between the objects is somewhat lost. This process is analogous to conversion of a higher level programming language (say FORTRAN) code to a lower level program-ming language (say Assembly) code. While it is easier to work at symbolic level, it cannot be used by the fab directly. In some cases, at late stages of the chip design process, some edits have to be made in the polygon database. The major motivation for use of the symbolic database is technology independence. Since physical dimensions in the symbolic database are only relative, the de-sign can be implemented using any process. However, in practice, complete technology independence has never be reached.