Although the full custom approach to the design of integrated circuits offers many advantages over other approaches, it is the most time-consuming design of all. Much of this time is spent during the debug cycle, making changes to the layout of the circuit and then running a circuit extractor prior to simulating the design. This thesis introduces two new computer-aided design tools that drastically reduce the time spent in this debug cycle: a fast, new circuit extractor, and an operation called plowing for making changes to mask level layout. Both tools have been implemented as part of the Magic IC layout system. The circuit extractor is both incremental and hierarchical. It computes circuit connectivity and transistor dimensions, both internodal and substrate parasitic capacitance, and parasitic resistances. It is parameterized to work across a wide range of MOS technologies. The keys to its speed are a new mask-level extraction algorithm based on corner-stitching, and its ability to extract cells incrementally. The mask-level extractor is 3-5 times faster than the fastest previously published extractor, and computes significantly more information. Because the extractor is incremental, only a few cells must be re-extracted after typical changes to a layout. The above facts make it possible to re-extract incrementally a 36,000-transistor chip in under 10 minutes, an operation that used to take previous extractors hours to perform. Plowing is a new operation for stretching and compacting parts of an IC layout. It allows designers to make topological changes to a layout while maintaining connectivity and layout rule correctness. Plowing can be used to rearrange the geometry of a subcell, compact a sparse layout, or open up new space in a dense layout. Unlike traditional compactors, plowing works directly on the mask-level representation of a layout. It uses a novel edge-based algorithm that works from a corner-stitched layout. This algorithm applies a collection of rules, parameterized by a technology file, to determine when edges must move.
Recommendations
Graph-Optimization Techniques for IC Layout and Compaction
This paper describes a new approach for IC layout and compaction. The compaction problem is translated into a mixed integer-linear programming problem of a very special form. A graph-based optimization algorithm is used to solve the resulting problem. ...